Behavioral Finance

Phase 6 · Week 55 · ~20 hours

Where it sits. Fourth module of the philosophy phase, and the one that decides whether any of the technical skill you have spent fifty modules building actually survives contact with a live market. Phases 1 to 5 taught you what a business is worth: how to read the statements (Phase 1), trust or distrust them (Phase 2), value them (Phase 3), judge their competitive quality (Phase 4), and apply the right sector lens (Phase 5). The first two modules of this phase gave you the great investors' approaches, and the third gave you the decision science: base rates, expected value, Kelly, Bayesian updating, inversion, ruin-avoidance. All of it assumes a calm, rational agent doing the analysis. You are not that agent, and neither is anyone setting the prices you trade against. What follows is the study of the systematic, predictable ways human minds misfire under uncertainty, greed, and fear. First as it happens inside one skull (Kahneman's two systems, prospect theory, the bias catalog), then as those errors aggregate into market-level forces (Munger's Lollapalooza, Soros's reflexivity, Marks's pendulum), and finally the practices that build a durable behavioral edge. That edge is very likely the largest and most defensible one you will ever have, because it is the hardest for anyone to copy.

Suggested hour budget. Core content: 8h · worked examples (the disposition-effect numbers, the Lollapalooza teardown, the reflexivity loop, the calibration drill): 2.5h · practice set (bias-identification vignettes + the self-audit): 4.5h · applied mini-project (the behavioral post-mortem): 3h · flashcards + mastery check: 2h.

App link. This module feeds the Philosophy & Behavior branch. It holds the bias-spotting drill (the gate for this material: a scenario or a footnote or a price chart in, the bias and its counter-practice out; ≥85% to pass), and the Lollapalooza case teardown (a bubble or fraud shown as a stack of compounding tendencies to be named one by one). It also holds the "take the market's temperature" gauge (a sentiment-and-valuation panel that locates the pendulum without pretending to time it), and the calibration tracker that shares its plumbing with the M9 decision journal and the calibration dashboard (predicted-vs-actual on your own theses, scoring reasoning over outcome). The red-flag drills you first met in Phase 2 get a behavioral overlay here: every great fraud worked because it exploited human tendencies, not just accounting ones, and you learn here to see both layers at once.

Learning objectives

By the end you can:

1. Explain why the market is structurally a machine for exploiting your psychology: how price is the crowd's consensus, why beating it requires being differently right (M6.02's second-level thinking), why the same biases that create the mispricing also live in your own head, and why the market presents maximal temptation at exactly the worst moments.
2. Describe Kahneman's System 1 and System 2 and locate precisely where in the analytical process investing errors are born, in the fast, intuitive snap judgment that System 2 then lazily rationalizes rather than checks, and explain why a written process is the only reliable fix.
3. Use prospect theory: reference dependence, loss aversion (~2×), diminishing sensitivity, and the reflection effect. Use them to explain the disposition effect (riding losers, selling winners) and reproduce the empirical numbers (Odean's PGR/PLR and the performance gap).
4. Name the mechanism, the specific investing failure mode, a real market example, and the counter-practice for each of the eleven core biases: anchoring, confirmation, loss aversion, recency, overconfidence, herding, sunk-cost, narrative fallacy, availability, hindsight, and endowment.
5. Work Munger's "Psychology of Human Misjudgment": group the ~25 tendencies sensibly, teach the six highest-leverage ones in depth (incentive-caused bias, social proof, commitment/consistency, deprival-superreaction, envy, authority), apply "show me the incentive and I'll show you the outcome," and explain the Lollapalooza effect with a worked case where four or more tendencies stacked.
6. Explain Soros's reflexivity, the two-way feedback loop in which perceptions change fundamentals which change perceptions, trace a boom-bust sequence, and identify the situations (financials, serial acquirers, anything where confidence is itself a fundamental) where it dominates.
7. Apply Marks's sentiment model, the pendulum and the three stages of both a bull and a bear market, and show how it is simply the individual biases summed to a market-level force you can take the temperature of.
8. Assemble the behavioral edge: map each bias to its specific counter-practice (checklist, decision journal, pre-commitment, environment design) and explain why process, not willpower or intelligence, is the only thing that reliably defeats a System-1 error.

Prerequisites & connections

Builds on. M6.03 supplies the rational baseline that human behavior gets measured against here. Base rates and the outside view are the cure for narrative fallacy and recency. Expected value and probabilistic thinking are what loss aversion and the reflection effect distort. Calibration and ruin-avoidance are what overconfidence attacks. Inversion and the pre-mortem are structured defenses against confirmation bias. Where M6.03 taught you how a good decision should be made, M6.04 explains why it usually isn't. From M6.01 come Graham's Mr. Market (the manic-depressive partner) and "the investor's chief problem, and even his worst enemy, is likely to be himself." Those were the founding behavioral insights of value investing, and they predate the academic field by decades. M6.02 showed that every school is, at bottom, a pre-commitment device against a specific bias. Klarman's cash discipline guards against greed and the institutional imperative. Fisher's hold-forever guards against overtrading, the Magic Formula's mechanical rank against discretionary panic. Marks's pendulum, also from M6.02, gets extended here. All of M2 matters too. Every forensic red flag you learned to spot exists because a human was incentivized to create it and other humans were biased enough to miss it, so the frauds in M2.08 are behavioral case studies as much as accounting ones.

Feeds forward. M6.05, the phase capstone, is where your personal checklist gets its behavioral lines, and where the behavioral self-audit you run in the practice set becomes a permanent instrument. The Mental-Model & Moat Checklist produced in M6.05 has a whole "behavioral self-audit" section, and its content is written here. The whole of Phase 9 (M9.01 to M9.04) is the behavioral edge put to work. Position sizing guards against overconfidence and hidden correlation. The five sell disciplines in M9.03 guard against the disposition effect, sunk cost, and endowment. The decision journal and calibration loop in M9.04 guard against hindsight and overconfidence. Together they take the closing material on building the behavioral edge and turn it into a running process. M7.05 takes cycles, bubbles, and crises. Its Minsky and Kindleberger anatomies are the macro-scale expression of the reflexivity and Lollapalooza dynamics taught here, and its 2008, dot-com, and Japan case studies are behavioral finance at national scale. In M8.04, the deep dive's Day 3 "steelman the bear" and "pre-mortem" are direct antidotes to confirmation bias and overoptimism. Competency C9 in the master map, "an investor's temperament and process," is certified partly here (the bias-spotting gate) and completed in M9.

If you keep one idea. The market is an adversary that has learned which buttons in your brain to press, and when. Your edge won't come from superior intelligence, because the crowd is plenty smart. It comes from a system that forces slow, disconfirming, pre-committed thinking at the very moments your fast, social, loss-averse, story-loving mind most wants to do the opposite.

Core teaching content

4.1 Why the market is a machine for exploiting your psychology

A valuation only makes you money if the market disagrees with you and later comes to agree. That's an uncomfortable premise, and it reframes everything in Phase 6. In Phases 1 through 5 you learned to answer "what is this business worth?" with growing precision. But the price on the screen isn't a number waiting to be discovered. It's the crystallized consensus of every participant, the weighted average of all the analysis, hope, fear, and forced buying and selling in the market right now. To profit, you have to be differently right. You need a view the consensus doesn't hold, and you need to be more correct than it is. Marks calls this second-level thinking, and you met it among the great investors' approaches. If you're right in a way the market already agrees with, you earn nothing. If you're differently wrong, you lose money.

Edge lives only in the narrow band where you are both non-consensus and correct.

Why does that band exist at all, and why is it a behavioral band? If every participant were the calm rational agent of textbook finance, prices would already embed all information and there'd be no band. That's the efficient-market case, and you'll weigh it properly later in the course. But the crowd that sets the price is made of human minds running the same buggy software yours runs. And the bugs don't cancel out.

If errors were random, a million of them would wash to zero and the average price would still be rational. They aren't random. They're correlated, because the same stimuli trigger the same biases at the same time. A rising price makes everyone more optimistic at once (recency, social proof, envy). A crash makes everyone fearful at once (loss aversion, availability). Correlated error doesn't cancel. It compounds, and it pushes price systematically away from value in booms and busts. That crowd-driven mispricing is the opportunity, and it's the reason a diligent analyst can beat an index at all.

Now the danger, which is the other half of the same coin. You're a member of that crowd, running the same software. The market isn't a neutral arena. Nobody engineered it. Evolution and structure have turned it into a machine that presents you with maximal temptation at the worst possible moments. It shows you a green, rising, everyone's-getting-rich ticker just when assets are most expensive and most dangerous, and that fires your recency bias, your social proof, your envy, and your fear of missing out all at once. It shows you a red, crashing, blood-in-the-streets tape just when assets are cheapest, firing your loss aversion, your availability bias, and your herd instinct to run. The financial media amplifies every one of these with vivid stories (availability), the latest number (recency), a compelling narrative (narrative fallacy) and a chorus of confident voices (social proof, authority). Even your brokerage app joins in. It pings your loss aversion in real time by showing you a red P&L.

Put it together and the stimulus environment is optimized to make you buy high and sell low. That's the exact inverse of the strategy that works.

Benjamin Graham had a name for this. He personified the market as Mr. Market, a manic-depressive business partner who shows up every day and offers to buy your share or sell you his. His price is driven entirely by his mood, euphoric and overpriced some days, despairing and cheap on others. Graham's genius was to note that Mr. Market is there to serve you, not to guide you. His mood is your opportunity, never your instruction. The behavioral extension is that Mr. Market isn't only "out there." He's also "in here," in your own System 1, and the manic-depressive partner you most need to ignore is the one inside your own skull. On average, the person on the other side of your trade is moving money from the impatient and emotional to the patient and disciplined. Buffett's version is that "the stock market is a device for transferring money from the active to the patient." Every behavioral bias you carry is a leak, and the market is a machine that finds and drains leaks.

So carry one idea through the next fifty pages. Your analytical skill and your behavioral discipline are two different competencies, and the second protects the first. A brilliant valuation run by a mind that panics at the bottom and capitulates at the top produces worse returns than a mediocre valuation run by a disciplined temperament. Buffett's most-repeated claim about investing makes the same point. It isn't a game where the person with the 160 IQ beats the person with the 130 IQ. What you need beyond ordinary intelligence is "a stable personality, a temperament that neither derives great pleasure from being with the crowd or against the crowd," and the ability to control the urges that get others into trouble.

What follows takes those urges apart and builds the systems that contain them, in four movements. First comes the machinery of the individual mind, then the specific leaks. After that you'll see how the leaks aggregate into market forces, and last, how to build the defenses.

Common mistakes & how experts think differently

Mistake 1. "Now that I know the biases, I'm immune to them." This is the most dangerous misconception in the whole field, and it's a bias in its own right. It's called the "bias blind spot": we readily see biases in others and not in ourselves. Knowing about anchoring doesn't stop you from anchoring. Kahneman, who spent his life studying these errors, was emphatic that he remained as susceptible as anyone. How experts think: they treat their own mind as a permanently unreliable instrument. They rely on external structure (checklist, journal, pre-commitment), never on the felt confidence that they've "seen through" the bias. Their humility isn't modesty. It's an accurate model of how little insight introspection gives into your own biased processing.

Mistake 2. Confusing a bias-label with an analysis. The beginner learns the catalog and starts slapping labels on everyone: "he's just anchoring," "that's confirmation bias." Naming a bias in someone else is cheap, especially someone who disagrees with you. It's often a confirmation-bias move in itself, a way to dismiss a disconfirming view without engaging it. How experts think: they use the catalog mainly on themselves, where it's useful and hard. When they analyze the market, they specify the mechanism and the evidence as well as the label. They don't say "the market is irrational." They say "here is the specific reflexive loop / the specific incentive / the specific sentiment reading, and here is what it implies about price versus value."

Mistake 3. Treating "the market is behavioral" as a license for arrogance. Once you see that crowd psychology drives prices, it's tempting to conclude the crowd is stupid and you're smart, and to bet aggressively against sentiment. But the crowd is usually right (the trend is your friend until it bends). Sentiment can stay irrational far longer than you can stay solvent, as Keynes warned. And "everyone else is biased, only I am clear-eyed" is overconfidence in its purest form. How experts think: they respect the crowd's price as the default. Before betting against it, they demand a specific, articulable reason: a mispricing they can name and a catalyst or a durability that lets them wait. They also size for the real possibility that the crowd is right and they're wrong. Being contrarian isn't a strategy. Being correctly contrarian, rarely and with a margin of safety, is.

Mistake 4. Fighting biases with effort instead of systems. The natural response to learning you're biased is to resolve to "think harder" and "be more disciplined." As the section on the behavioral edge argued, this fails. Effort is a System 2 resource, and it runs out just when biases are strongest. How experts think: they externalize discipline into structures that work when willpower doesn't. They design the ship so an ordinary, tired, frightened human can survive the Sirens, because that's who will be at the controls in the crisis. The question they ask isn't "will I be disciplined enough?" It's "what structure makes the disciplined action the default?"

Mistake 5. Using the disposition effect's logic while believing you're being prudent. "I'll take profits here to be safe, and I'll hold this loser until it recovers so I don't lock in a loss." It sounds like risk management. It's the disposition-effect error dressed as prudence. The tell is that both halves refer to your cost basis and ignore forward value. How experts think: every hold/sell decision runs through one forward-looking question, "would I buy this at today's price with fresh money?", and the cost basis gets treated as the irrelevant historical accident it is. Winners are held while they remain undervalued, which is how you let them run, as with Lynch's ten-baggers. Losers are sold when the thesis breaks, and getting back to even has nothing to do with it.

Mistake 6. Mistaking a good company or a good story for a good investment. Beginners fall in love with businesses and narratives and forget that price is half of every investment. The halo effect makes "I admire this company" bleed into "I should own this stock at any price." How experts think: they keep two questions rigidly apart, "is this a good business?" (Phase 4) and "is this a good price?" (Phase 3). They know the most dangerous investments are wonderful companies at wonderful-company-priced-for-perfection valuations. There a great business can still be a terrible investment, even with the narrative true, because the price already embeds a flawless future. Running a reverse DCF is how you catch it.

Mistake 7. Believing the behavioral edge is about predicting the crowd. The pop version of behavioral finance is "understand the crowd's psychology so you can predict its next move and front-run it." That's market timing, and it doesn't work reliably. How experts think: the edge comes from non-participation in the crowd's errors, plus readiness to act when those errors create mispricing. Prediction isn't part of it. Marks's "know where you are, not where you're going" is the rule. You can't time the pendulum, but you can decline to buy at the euphoric extreme, keep dry powder, and deploy it when panic creates bargains. It's a positioning edge, not a forecasting one.

Worked examples

Worked Example 1: The disposition effect, quantified (synthetic portfolio + the Odean logic)

Purpose: turn the reflection effect from a story into numbers you can compute, using the ratios Odean measured. A synthetic investor holds ten positions. Today they decide to raise some cash and sell four of them. The book looks like this. All figures are illustrative, and "position value" means current market value:

That's 5 winners and 5 losers available to sell. Our disposition-prone investor feels the pull to "lock in gains" and "not crystallize losses." So they sell winners C, D, E and loser F, the smallest loss and the least painful to realize. That's 3 winners and 1 loser.

Step 1. Compute Odean's two ratios.

• Proportion of Gains Realized: PGR = winners sold / winners available = 3 / 5 = 0.60.
• Proportion of Losses Realized: PLR = losers sold / losers available = 1 / 5 = 0.20.

Step 2. Measure the disposition tilt. The ratio is PGR / PLR = 0.60 / 0.20 = 3.0. This investor is realizing gains at three times the rate of losses. That's an even stronger tilt than Odean's population average of about 1.5× (PGR 0.148 / PLR 0.098), and a textbook disposition signature. A disposition-free investor selling on forward prospects would show PGR ≈ PLR.

Step 3. The wealth consequence, and why it isn't harmless. Odean found that the winners sold went on to beat the losers held by ~3.4 points over the next year. Apply that logic here. The investor has kept G, H, I, J, the deep losers. They sold the winners C, D, E. If the sold winners outperform the held losers over the next year, as Odean found on average, the investor has systematically sold the stronger horses and kept the weaker ones. Suppose over the next year the two sold winners' cohort returns +12% and the held-losers' cohort returns +5%. That's an illustrative ~7-point gap, consistent with the direction of Odean's result. The disposition behavior has cost this investor real return. It has also cost tax efficiency. In a taxable account they realized gains, which triggers tax, while deferring the harvestable losses, the exact inverse of tax-smart behavior.

Step 4. The fix, shown. Re-run the sell decision through the forward test. Ask of each position, regardless of cost basis, "would I buy this at today's price with fresh money?" Suppose the answers come out like this:

• A, B, C are still undervalued. Hold.
• D, E are now fully valued. Sell, even though they're winners.
• F, G are still sound and now cheaper. Hold, even though they're losers.
• H, I, J are broken theses. Sell, even though selling crystallizes losses.

The forward-looking investor sells D, E, H, I, J. That trims fully valued winners and cuts broken losers, nearly the opposite of the disposition-driven trade. Same portfolio, same day. The reference point (cost basis versus forward value) decided everything.

Worked Example 2: A Lollapalooza teardown: Satyam (India, 2009)

Purpose: practice decomposing a real fraud into the stack of Munger tendencies that let it run, the core skill of the Munger section. Until January 2009, Satyam Computer Services was one of India's four largest IT-services firms. It was NYSE-listed, a winner of governance awards, and audited by a Big Four firm. Then chairman Ramalinga Raju confessed in a letter to the board that he'd fabricated the accounts for years. Roughly ₹5,000 crore of the ₹5,361 crore of cash and bank balances on the balance sheet didn't exist. That was part of a total hole of about ₹7,100 crore, once fictitious accrued interest, understated liabilities, and overstated debtors were counted. Revenue and margins had been inflated for years. He described it as "riding a tiger, not knowing how to get off without being eaten." Now decompose why it worked. Leave aside the accounting mechanics (fictitious invoices, forged bank statements) and look at the psychology that let it persist and fool so many:

1. Incentive-caused bias (Munger #1). Raju's incentive was to keep reported growth and margins high to support the share price and his family's holding. By his own account, he also wanted to hide a gap between real and reported performance that had begun small and compounded. The auditors had a lucrative recurring engagement, and that incentive didn't reward the aggressive skepticism that would have caught forged bank confirmations. Every actor did roughly what their incentives paid them to do.
2. Commitment and consistency (Munger #5). Raju's own "riding a tiger" metaphor names this tendency. Once the first small misstatement was made, consistency demanded the next, and the next. Each period's fabrication had to be at least as large as the last to avoid an unexplained collapse. The fraud grew because abandoning it, admitting the gap, got more painful with every quarter of commitment. What drove the escalation each quarter wasn't fresh greed. It was consistency-avoidance, compounding.
3. Authority-misinfluence (Munger #22). The clean audit opinion from a Big Four firm, the governance awards, the NYSE listing, and the eminent board worked as authority blessings. In investors' and analysts' minds they substituted for independent verification. Almost no one re-derived whether over ₹5,000 crore of cash could plausibly exist alongside the company's borrowing and the implausibly low interest it earned. The authorities had signed off, so the mind stopped checking.
4. Social proof (Munger #15). Institutions owned Satyam widely, and it was included in indices. "Everyone" owned it and reputable everyone endorsed it, which told each new holder that owning it was safe. The consensus was self-reinforcing.
5. Doubt-avoidance and denial (Munger #4, #11). The final, failed act came in December 2008. There was an attempt to have Satyam buy two Raju-family infrastructure companies (Maytas), which would have let real assets "fill" the fictitious cash hole. It provoked a shareholder revolt that collapsed the scheme. For years before that, though, the simple pain-avoiding reluctance to look hard at a successful, award-winning national champion kept doubt at bay.

The Lollapalooza point. No single one of these would have sustained a ₹7,000-crore fraud. Stacked, they multiplied: incentives created it, consistency compounded it, authority and social proof blessed it, and denial protected it. The collapse wiped out shareholders and shook confidence in Indian corporate governance, and it drove the reforms you study in the ethics thread. Carry one habit away from this. When a company seems too good and everyone agrees, run the stack deliberately:

• Whose incentives?
• What has been publicly committed?
• Which authority am I trusting instead of checking?
• Is my comfort coming from social proof?

The tells were in the filing: implausible cash earning implausibly low interest, the related-party structure, promoter behavior. The biases are why the market didn't look.

Worked Example 3: Reflexivity, with numbers: the serial-acquirer feedback loop

Purpose: make Soros's abstract loop concrete and quantified, and show how the same mechanism runs in reverse. Take a synthetic serial acquirer, "RollUp Co," and watch the reflexive loop turn.

The boom phase. RollUp trades at a P/E of 40 (₹40 of price per ₹1 of EPS). It earns ₹100 crore of net income on 100 crore shares, so EPS is ₹1.00 and the share price is ₹40, for a market cap of ₹4,000 crore. It finds a target earning ₹25 crore of net income that it can buy for ₹250 crore. That's a P/E of 10, for a boring, low-growth business. RollUp pays with its own richly valued stock, issuing 250 / 40 = 6.25 crore new shares to fund the ₹250 crore purchase.

• New net income = 100 + 25 = ₹125 crore.
• New share count = 100 + 6.25 = 106.25 crore.
• New EPS = 125 / 106.25 = ₹1.176, up 17.6%, with no organic growth at all. All of that "growth" came from buying ₹1 of earnings at a P/E of 10 with stock valued at a P/E of 40. It's pure multiple arbitrage.

Now the reflexive turn. The market sees 17.6% EPS "growth," concludes RollUp is a great growth compounder, and maintains or raises the P/E of 40. New price = 40 × 1.176 = ₹47.06. The high multiple has produced the earnings growth that justifies the high multiple, which is Soros's loop to the letter. And with the stock even higher, the next acquisition is even more accretive, so management does more deals, faster. Perceptions (high multiple) changed fundamentals (EPS growth), which changed perceptions (validated multiple). The boom accelerates.

The bust phase. The loop runs on one condition: the market keeps paying a high multiple for acquired earnings. It can break on a recession, an accounting question, a short report, or simply the arithmetic getting too large to sustain. When it does, the reversal is violent and fast, as Soros's asymmetry predicts. Say the market re-rates RollUp to a P/E of 12, having realized the "growth" was financial engineering on low-quality earnings. Price collapses from ₹47 to 12 × 1.176 = ₹14.1, a 70% fall. Now the reflexive loop runs in reverse. The cheap stock is no longer a valuable acquisition currency, because issuing shares at a P/E of 12 to buy earnings at a P/E of 10 is barely accretive and dilutes existing holders. So the EPS growth stops. The halted growth confirms the "it was just financial engineering" thesis, and the confirmation pushes the multiple lower still. Meanwhile any debt taken on during the boom now stands against a shrunken market value. The machine that manufactured growth on the way up manufactures collapse on the way down.

The analyst's takeaways. (1) The tell that you're in a reflexive situation is that the company's growth depends on its own share price. In this case the stock is the acquisition currency. Whenever price feeds the fundamentals you're in Soros's world, and a snapshot "intrinsic value" that ignores the loop is unreliable. (2) Your defense is the reverse DCF and the quality-of-earnings checks from your forensic work. Strip out acquisition-driven EPS growth and ask what the organic business earns and is worth. The reflexive premium is the part that evaporates. (3) Put "confidence" and "funding access" in place of "share price" and "acquisition currency," and the same loop becomes the NBFC or bank run. Financials are the purest reflexive instruments for this reason, and "a cheap financial" can be a value trap, cheap for a reason once confidence cracks.

Worked Example 4: A calibration check on overconfidence

Purpose: let you feel your own overconfidence and see how calibration measures it. This is the raw material of the calibration process you'll run later. Answer each of these ten general-knowledge questions with a 90% confidence interval: a low and a high number you're 90% sure contains the true answer. You aren't trying to be right. You're trying to be calibrated. If you've set your intervals the way you were asked, about 9 of 10 should contain the truth. The questions are of the type used in classic calibration studies.

1. The year the Reserve Bank of India was established.
2. The length of the Ganges river, in kilometres.
3. The market capitalisation of TCS at its 2024 peak, in ₹ lakh crore.
4. The number of member countries in the United Nations.
5. The air distance from Mumbai to New York, in kilometres.
6. The year Warren Buffett took control of Berkshire Hathaway.
7. The number of listed companies on the NSE.
8. The boiling point of water at the summit of Mount Everest, in °C.
9. The gross domestic product of India in 2024, in US$ trillion.
10. The number of tendencies in Munger's revised "Psychology of Human Misjudgment."

How to score it. Check your intervals against the true values (RBI 1935; Ganges ~2,525 km; TCS peak ~₹15–16 lakh crore; UN 193; Mumbai–NY ~12,500 km; Berkshire 1965; NSE ~2,000+ listed; Everest boiling point ~71°C; India GDP ~US$3.9 trillion; Munger 25). Count how many of your ten intervals actually contained the truth.

The universal result. Almost everyone catches fewer than 9, typically 4 to 6, even though they were told to be 90% confident. That gap, feeling 90% sure while being right ~50% of the time, is overconfidence, measured. It doesn't mean you're ignorant of geography. It means your subjective confidence runs systematically higher than your actual accuracy, and that's the error that makes a DCF's "₹1,847 fair value" and a forecast's narrow range so dangerous. The fix is mechanical, and you'll turn it into a standing practice later. State explicit probabilities on real forecasts and track predicted against actual over dozens of decisions. Then let the scoreboard teach you how wide your intervals need to be, because your feeling won't. Well-calibrated humans are made, not born, and the making is done with a journal.

Worked Example 5: The rupee price of loss aversion (a declined +EV bet, and the disposition drag)

Purpose: put a number on loss aversion. You'll see why a profitable bet gets refused, and what riding a loser costs in cash, using the λ ≈ 2.25 coefficient from prospect theory. An Indian analyst is offered a one-year position with a 50/50 payoff: +₹1,50,000 on heads, −₹1,00,000 on tails (all figures synthetic).

Step 1. The rational expected value. EV = 0.5 × 1,50,000 + 0.5 × (−1,00,000) = 75,000 − 50,000 = +₹25,000. The bet is clearly worth taking, and a risk-neutral agent accepts it every time.

Step 2. The felt value under loss aversion. Prospect theory weights the loss by λ ≈ 2.25: felt = 0.5 × 1,50,000 − 0.5 × (2.25 × 1,00,000) = 75,000 − 1,12,500 = −₹37,500. The threatened loss feels like ₹2,25,000, which swamps the ₹1,50,000 upside. System 1 reports a losing proposition, and the analyst declines, giving up ₹25,000 of real expected value. That forgone ₹25,000 is the rupee cost of loss aversion on this one decision. This is a linear approximation. The true value function adds diminishing sensitivity, but λ is treated here as the order-of-magnitude driver.

Step 3. The breakeven upside, deriving the "2:1" rule. What heads-payoff G makes the 50/50 bet just tolerable, with felt value = 0? 0.5 × G = 0.5 × (2.25 × 1,00,000) → G = ₹2,25,000. The upside must be 2.25× the downside before the bet feels fair. That's the everyday "I need about 2:1" intuition, derived from λ instead of asserted.

Step 4. The same asymmetry as portfolio P&L, with the disposition drag priced. The identical force makes her hold a loser instead of realizing a loss. Suppose she clings to a position now worth ₹3,50,000. She bought it higher, so it carries a painful paper loss, and she expects it to return only ~2% next year. She also has a better idea she expects to return ~12%, but she won't switch the ₹3,50,000 into it. The one-year cost of refusing to "make the loss real" is the forward spread on the tied-up capital: (12% − 2%) × 3,50,000 = ₹35,000. Notice what doesn't appear in that calculation. Her purchase price. The disposition effect is expensive because it prices decisions off an irrelevant past number. The fresh-money test restores the only figures that matter, forward return and capital at work.

Worked Example 6: Anchoring to the 52-week high, in dollars ("down 40%, so it's cheap")

Purpose: quantify the most common anchoring trap in investing, mistaking a fallen price for a low value, and show how the anchor contaminates the valuation itself. A US investor is drawn to NovaTech (synthetic). Its 52-week high was $180, and it now trades at $108. "Down 40% from its high: a bargain." A from-scratch DCF built blind to the quote (Phase 3) puts intrinsic value at $85/share.

Step 1. The anchored perception. Discount from the high: (180 − 108) ÷ 180 = 40%. Anchored on $180, the $108 price reads as a 40%-off sale, and System 1 files "cheap."

Step 2. The from-scratch reality. Against the $85 intrinsic value, the $108 price is a premium: (108 − 85) ÷ 85 = +27.1%. Look at it blind to the high and the stock isn't cheap at all. It's 27% expensive. The $180 high was a price set at peak sentiment, not a value, and it may itself have been absurd.

Step 3. The dollar cost if value wins. Buy 500 shares at $108 = $54,000 for shares worth 500 × 85 = $42,500, and you've overpaid by $11,500 on day one. If price converges to value, the return is (85 − 108) ÷ 108 = −21.3%. In cash, the "40% discount" was a one-in-five loss waiting to happen.

Step 4. The anchor contaminates the estimate itself. Anchoring doesn't stop at the buy/sell decision. It poisons the valuation too. Classical experiments show we adjust insufficiently from a salient number. Say the analyst lets the $108 quote anchor her and adjusts only halfway to her model. Her stated "fair value" lands at 108 − 0.5 × (108 − 85) = $96.50, still 13.5% above the true $85. Now the $108 purchase looks like a rounding error, when it's really a 27% overpayment. Counter-practice: build the intrinsic-value range first, blind to the price. Treat the 52-week high (and your own cost basis) as psychologically loaded noise that tells you nothing. Then, and only then, compare to the quote.

Practice set

Work each problem completely before you read the solution. Name the bias (or biases) and the counter-practice, or do the computation. The set ramps up from guided single-bias vignettes to multi-bias and computational problems, then timed mixed identification. It closes with the self-audit. It's the same format as the app's gate (the bias-spotting drill) and the mastery check.

Problem 1 (guided, single bias). An investor says: "I bought TCS at ₹4,200. It's ₹3,600 now, but I'm not selling at a loss. I'll wait until it gets back to ₹4,200." Name the bias (there are two closely related), and state the counter-practice.

Solution. The primary bias is anchoring to the ₹4,200 purchase price. That's a fact about the investor's past, and it's irrelevant to TCS's future value. Loss aversion reinforces it: he refuses to "sell at a loss" because crystallizing the loss is ~2× as painful as the equivalent gain, which is the disposition effect. The market has no memory of what he paid and owes him nothing at ₹4,200. Counter-practice: the fresh-money test. "Ignoring what I paid, would I buy TCS at ₹3,600 today?" If yes, hold because it's a good buy (waiting to break even is no reason). If no, sell now, whatever the loss. Replace the cost-basis reference point with a forward, value-based one.

Problem 2 (guided, single bias). A fund manager privately thinks a popular index heavyweight is overvalued, but keeps it at index weight in her portfolio, reasoning: "If I'm underweight and it keeps going up, I'll underperform and lose clients. If I hold it and it falls, so does everyone else, and no one blames me." Name the bias and the underlying driver.

Solution. Herding, driven specifically by career/agency risk. It isn't simple social proof here, because she has an independent view and the pressure is professional. This is Keynes's "better to fail conventionally than to succeed unconventionally" and Buffett's institutional imperative in action. The incentive structure rewards tracking the crowd, even against her own analysis. Counter-practice (structural, mostly at the process/mandate level): an independent written thesis, and an investment process measured over multi-year horizons instead of monthly tracking error. For the individual investor, recognize that having no career risk or benchmark is a real structural edge (Lynch). Use it, and don't squander it by voluntarily hugging an index.

Problem 3 (guided, story vs numbers). An analyst writes: "The founder is a visionary, the TAM is $2 trillion, and they're literally changing how the world works. I don't care that it loses money and trades at 30× sales; you can't value a revolution with a spreadsheet." Name the bias and the discipline that answers it.

Solution. Narrative fallacy, with a dose of the halo effect: a visionary founder is taken to imply sound economics, without evidence. It's compounded by the classic confusion of a good story / good company with a good investment at this price. "You can't value a revolution with a spreadsheet" is what it sounds like when System 1 refuses the System 2 check. Counter-practice: Damodaran's narrative and numbers. You may believe in the revolution, but every element of the story has to become a number and be tested for plausibility. What revenue, what margin, what reinvestment does a $2 trillion TAM imply for this company's share? Then run the reverse DCF. It solves for the growth and margin the 30×-sales price already assumes, and asks whether that has to be true. A revolution can be real and the stock still a terrible investment if the price embeds a flawless future.

Problem 4 (independent, compute the disposition ratios). An investor holds 8 winners and 12 losers. Over a quarter she sells 6 of the winners and 3 of the losers. Compute PGR, PLR, and the disposition ratio, and state whether the behavior is disposition-consistent.

Solution. PGR = winners sold / winners held = 6 / 8 = 0.75. PLR = losers sold / losers held = 3 / 12 = 0.25. Disposition ratio = PGR / PLR = 0.75 / 0.25 = 3.0. She's realizing gains at three times the rate of losses, which is strongly disposition-consistent. A disposition-free investor selling on forward prospects would show PGR ≈ PLR ≈ 1.0-neutral. The behavioral risk is that she's systematically keeping her losers, which per Odean tend to keep underperforming the winners she sold. She's also paying tax on realized gains while forgoing harvestable losses.

Problem 5 (independent, the reflection effect). Present the same person with two choices. Choice X: a sure gain of ₹9,000, or a 90% chance of ₹10,000 (else nothing). Choice Y: a sure loss of ₹9,000, or a 90% chance of losing ₹10,000 (else nothing). Compute the expected value of each gamble, predict what most people choose in each, and explain what the pattern reveals.

Solution. Choice X gamble EV = 0.90 × 10,000 = ₹9,000, identical to the sure ₹9,000. Choice Y gamble EV = 0.90 × (−10,000) = −₹9,000, identical to the sure −₹9,000. Both pairs are EV-equivalent, so a risk-neutral agent is indifferent. Most people, though, take the sure ₹9,000 in X. They're risk-averse in gains and grab the certain win. And they take the gamble in Y, because they're risk-seeking in losses and gamble to avoid the certain loss. Same person, opposite risk attitude, flipped purely by the gain/loss frame. That's the reflection effect. Map it onto a portfolio and you get the disposition effect: risk-averse with winners (sell to lock the sure gain), risk-seeking with losers (hold and gamble on a recovery instead of crystallizing the loss).

Problem 6 (independent, multi-bias vignette). A retail investor in early 2000 (or 2021, or any mania) says: "My colleague made 8× on this stock last year. Everyone in my WhatsApp group owns it. It's been going straight up for months. I've read all the bullish threads and the bears are just jealous shorts. I'm putting in my savings." Name four distinct biases operating, and identify the market-level force they sum to.

Solution. At least four are at work. There's envy: "my colleague made 8×" is Munger's "not greed but envy," the specific pain of others' gains driving capitulation. There's social proof / herding, where "everyone in my group owns it" takes the crowd's behavior as evidence of correctness. There's recency bias, as "straight up for months" extrapolates the recent trend indefinitely. And there's confirmation bias in "read all the bullish threads... bears are just jealous shorts," which seeks support and dismisses disconfirmation without engaging it. You could add availability (the vivid, everywhere-in-the-headlines story) and overconfidence (betting savings).

Sum these individual biases and correlate them across millions of such investors, and they are the market-level force. It's Marks's pendulum, swung to the euphoric extreme: stage 3 of a bull market, where "everyone believes it will go up forever." That's where risk is highest and feels lowest. It's also a Lollapalooza in the making, four-plus tendencies stacking toward one outcome.

Problem 7 (independent, reflexivity recognition). For each situation, state whether reflexivity is likely to dominate (price/confidence feeds back into fundamentals) or be second-order (fundamentals largely independent of price), and why: (a) a debt-free consumer-staples company with negative working capital and 40% ROCE; (b) a rapidly-growing NBFC funding long-term loans with short-term commercial paper; (c) a serial acquirer using its high-multiple stock to buy companies; (d) a mature, cash-rich IT-services firm.

Solution. (a) Second-order. A debt-free, self-funding, high-return staples business generates cash whatever its share price does. Sentiment moves the multiple but doesn't threaten the fundamentals. It's the kind of business Buffett buys because it's robust to reflexive loops. (b) Reflexivity dominates. The NBFC's access to short-term funding depends on the market's confidence in its solvency. A confidence crack cuts off the funding the model requires, forcing asset sales and realizing the very losses that were feared, as in the IL&FS/DHFL run. Confidence is a fundamental here. (c) Reflexivity dominates. The high multiple is the acquisition currency that manufactures the EPS growth that justifies the high multiple (Worked Example 3). When the multiple compresses, the machine reverses. (d) Second-order. Like (a), a mature cash-generative firm's economics don't depend on its stock price. The transferable tell is that reflexivity dominates whenever the share price or the market's confidence feeds back into the company's actual results. That chiefly means leveraged financials and stock-as-currency acquirers.

Problem 8 (independent, incentive reading). You are handed three bullish research notes on a company. Note A is from the company's own investor-relations deck. Note B is from a sell-side analyst at the bank that just underwrote the company's bond issue. Note C is from an independent investor with a large disclosed long position who has published her full thesis and the disconfirming evidence she's watching. Using Munger's incentive lens, rank the notes by how much weight to give them and why, and state what you'd still do with Note C.

Solution. Ranked from most to least trustworthy as analysis: C, then A, then B, and you read all three critically. Note B gets the least weight despite the analyst's expertise, because the incentive-caused bias is severe. The analyst's firm has a banking relationship through the bond underwriting, and that's the reason sell-side "sell" ratings are perennially rare. The incentive pushes hard toward a favorable conclusion. Note A is openly promotional, since IR's job is to present the bull case, so at least its bias is transparent. Read it for the company's own framing and the facts you can verify. Don't read it for balance. Note C is the most useful because the author's incentive comes closest to being aligned with getting the answer right. She profits only if she's right, and she has exposed her thesis to scrutiny and named her own disproof triggers. But a large disclosed long also puts commitment/consistency and confirmation pressure on her. So still verify her numbers against the primary filing, and steelman the bear case yourself instead of adopting her long thesis wholesale. Munger's standing question, whose incentives shaped this, and which way do they push?, is the tool.

Problem 9 (timed, 6 minutes, rapid mixed identification). For each one-line vignette, name the single best-fit bias and its one-line counter-practice. (a) "It's down 60% from its 52-week high, so it must be cheap." (b) After a plane-crash headline, an investor dumps all his airline stocks though nothing changed in the businesses. (c) "I've spent 200 hours researching this position, I can't sell it now." (d) After the 2008 crash: "Obviously housing was a bubble, everyone knew it." (e) "I wouldn't buy this stock at today's price, but I already own it, so I'll keep it." (f) A DCF outputs "fair value ₹2,143.86 per share" and the analyst treats it as precise.

Solution. (a) Anchoring, to the 52-week high, which was a price not a value. Counter it by valuing the business from scratch, blind to the high. (b) Availability. A vivid, recent, emotionally charged instance hijacks the probability estimate. Look up the base rate of the risk, because the business economics are unchanged. (c) Sunk-cost fallacy. Past effort is spent whether you hold or sell. Apply the fresh-money test, since the research cost is irrelevant to the forward decision. (d) Hindsight bias, "knew it all along" after the outcome. The only cure is the decision journal written before the outcome, and almost no one acted on this "obvious" knowledge at the time. (e) Endowment effect. Owning it raises its felt value above what you'd pay. Use the zero-based portfolio test: if you wouldn't buy it here, it's a sell candidate. (f) Overconfidence, in the form of false precision, since the inputs can't support two decimals. State a range and then widen it. Calibrate with a track record.

Problem 10 (timed, 5 minutes, decompose a Lollapalooza). In two or three sentences each, name four Munger tendencies that stacked in the 2008 US housing/credit bubble, and state the single sentence that captures why decomposing a mania this way is useful.

Solution. (1) Incentive-caused bias. Originators were paid on loan volume not quality, rating agencies were paid by the issuers they rated, and bankers were paid on this year's originations, so each did what they were paid to do. (2) Social proof. Every bank participated because every other bank did ("as long as the music is playing, you've got to get up and dance"). (3) Authority-misinfluence. AAA ratings and regulators' calm were treated as verification, so buyers didn't look through to the loans. (4) Availability/recency. "US national house prices have never fallen" was an available recent truth extrapolated into false certainty. Denial/overoptimism is a valid fifth. The useful sentence: a mania is a multiplicative confluence of ordinary tendencies pulling the same way, with no individual stupidity required, so decomposing it into its named components is how you recognize the next one. The tendencies that stacked last time will stack again, in different clothes.

Problem 11 (the self-audit, written down and done candidly). This is the exercise everything here is really for. Take a real investment decision you've made. If you have none, take a significant non-investment decision, such as a job, a purchase, a commitment. Write answers to all of the following before reading the solution/model:

• (a) What was my actual reasoning at the time, and can I reconstruct it truthfully, or has hindsight already rewritten it?
• (b) Which two or three biases from the catalog was I most likely running? Give the specific evidence in my own behavior, not a generic label.
• (c) Did I seek the disconfirming case, or only support? Note too where my information came from and whose incentives shaped it.
• (d) What was my confidence, and, with the outcome now known, was it calibrated or overconfident?
• (e) What one structural practice (checklist line, journal entry, pre-commitment, environment change) would have most improved the process (regardless of the outcome)?

Solution / model of a good self-audit. There's no single right answer, because the deliverable is being specific and truthful about yourself. A strong self-audit does have recognizable features, though, and you should grade yourself against them.

On (a), it separates what you can document (a note, a message, a contemporaneous reason) from what you're reconstructing from memory. It flags the reconstruction as unreliable, because hindsight bias means your memory of your prior belief has probably drifted toward the outcome.

On (b), it names mechanisms with evidence. A weak answer says "I was overconfident." A strong one says "I sized it at 25% of the portfolio on the strength of one quarter's results (recency) and I followed only the three bullish accounts I already agreed with (confirmation)."

On (c), it traces where the information came from and whose incentives shaped it. A weak audit says "I did a lot of research." A strong one says "my research was 90% from the company's own materials and bullish forums (a confirmation echo chamber), and I never read a bear case or checked the related-party note."

On (d), it states the confidence you actually had, ideally from a contemporaneous record, and judges calibration separately from outcome. A decision that worked out can still have been overconfident and lucky. A decision that lost can have been well calibrated and unlucky. That's process over outcome.

On (e), it identifies one structural fix, because biases are beaten by systems, and resolve won't do it. That one fix should become a permanent line in your checklist or a standing journal prompt.

A mature self-audit is uncomfortable. If yours flatters you, you've run confirmation bias on your own self-audit. Redo it, looking specifically for what you got wrong in process, even where the outcome was fine.

Problem 12 (integration, bias to structure). For each of the four structural practices in the behavioral-edge section (checklist, decision journal, pre-commitment, environment design), name the two biases it most directly defeats and explain the mechanism in one sentence. Then explain in a single sentence why four structures can defend against eleven-plus biases.

Solution. The checklist defeats confirmation and narrative fallacy. It forces System 2 onto the "state the bear case" and "turn the story into testable numbers" questions System 1 skips. The decision journal defeats hindsight and overconfidence. An un-editable prior belief blocks "I knew it all along," and comparing stated probabilities to outcomes makes calibration possible. Pre-commitment defeats loss aversion/disposition and recency/herding. Rules set in the calm state bind the agitated one, forcing buy-low/sell-high mechanics against the mood of the moment. Environment design defeats availability and social proof by lowering trigger density: less media, fewer price-checks, primary sources first. The bias never fires, so it never has to be resisted. Why four suffice: the eleven biases share a small number of roots (reference-point dependence, System 1's story-building, social/authority forces). So a handful of well-aimed structures covers the field, plus the fresh-money test, which on its own defeats anchoring, endowment, sunk cost, and loss aversion. That's the practical hope of everything here. You don't need eleven disciplines. You need a small, well-engineered system.

Problem 13 (guided, loss-aversion arithmetic). An investor is offered a one-shot 50/50 bet: win ₹80,000 on heads, lose ₹50,000 on tails (synthetic). (a) Compute the expected value. (b) Using a loss-aversion coefficient λ = 2.25, compute the loss-aversion-weighted "felt" value and predict whether she takes the bet. (c) What minimum win would make the bet feel fair?

Solution. (a) EV = 0.5 × 80,000 − 0.5 × 50,000 = 40,000 − 25,000 = +₹15,000. It's positive, so a rational agent accepts. (b) felt = 0.5 × 80,000 − 0.5 × (2.25 × 50,000) = 40,000 − 56,250 = −₹16,250. Because the ₹50,000 loss is weighted 2.25×, the bet feels like a loser, and she declines a +₹15,000-EV bet. The ₹15,000 she forgoes is the cash cost of loss aversion here. (c) Set felt value to zero: 0.5 × G = 0.5 × (2.25 × 50,000) → G = ₹1,12,500. The win must be 2.25× the ₹50,000 downside. This is loss aversion. The counter-practice is to judge the bet on its EV and its survivability, never on the felt sting of the downside.

Problem 14 (guided, anchoring, with the tell). An analyst writes: "The sell-side target price is ₹1,500; the stock is ₹1,200, so there's 25% upside." She then builds her own DCF and arrives at a fair value of ₹1,480. Name the bias, identify the specific tell, and state the counter-practice.

Solution. Anchoring, to the sell-side target of ₹1,500, which becomes "a gravitational anchor for your own estimate," as the catalog warned. The tell is that her "independent" DCF landed at ₹1,480, within (1,500 − 1,480) ÷ 1,500 = 1.3% of the analyst's target. That's far too close to be a coincidence. She adjusted too little from the anchor when she should have valued from scratch. Notice too that the "25% upside" ((1,500 − 1,200) ÷ 1,200) is measured off the anchor, not off a value. Counter-practice: build the intrinsic-value range blind to both the market price and the analyst target, then compare. Treat sell-side targets as incentive-shaped noise, because the analyst's firm may want the banking mandate (Munger's incentive lens again). Don't use them as an input.

Problem 15 (independent, disposition P&L in dollars). A US investor holds two positions bought together for $20,000 each: a winner now worth $28,000 and a loser now worth $14,000. Feeling the disposition pull, she sells the winner and keeps the loser. Over the next year the type of stock she sold (the winner) rises 15% while the type she kept (the loser) falls a further 8%; assume sale proceeds sit in cash at 0%. (a) What gain did she realize on the sale? (b) Compare her year-end value under this disposition choice versus the thesis-based choice (sell the loser, keep the winner). (c) Name the effect and the fix.

Solution. (a) Realized gain on the winner = 28,000 − 20,000 = $8,000. (b) Disposition choice (keep the loser, winner to cash): 28,000 + 14,000 × 0.92 = 28,000 + 12,880 = $40,880. Thesis choice (keep the winner, loser to cash): 14,000 + 28,000 × 1.15 = 14,000 + 32,200 = $46,200. The disposition choice leaves her $5,320 poorer over the year (46,200 − 40,880). She sold the horse that ran and kept the one that stumbled, which is Odean's finding in miniature. (c) The disposition effect. The fix is the fresh-money / thesis-based sell rule: hold each name only if you'd buy it at today's price, whether it shows a gain or a loss.

Problem 16 (independent, the reflexive accretion loop). GrowthRoll Inc (synthetic, US) trades at a P/E of 25, earning $50m on 50m shares (EPS $1.00). It acquires a target earning $10m of net income for $150m, paid entirely in its own stock. (a) How many new shares are issued? (b) Compute the new EPS and the % accretion. (c) At an unchanged P/E of 25, what is the new price, and what has the market mistaken for growth?

Solution. (a) Share price = 25 × 1.00 = $25, so shares issued = 150 ÷ 25 = 6m. (b) New net income = 50 + 10 = $60m. New share count = 50 + 6 = 56m. New EPS = 60 ÷ 56 = $1.071, accretion of +7.1% with zero organic improvement. (c) New price = 25 × 1.071 = $26.79. The "growth" is pure multiple arbitrage: the target was bought at a P/E of 150 ÷ 10 = 15 using stock valued at a P/E of 25. It's Soros's reflexive loop from Worked Example 3. The high multiple manufactures the EPS growth that appears to justify the high multiple, and it reverses violently the moment the multiple compresses. Counter-practice: strip out acquisition-driven EPS growth and value the organic business, using quality-of-earnings checks and a reverse DCF.

Problem 17 (independent, the reflection effect, both frames). Present one person two choices. Gain frame: a sure $4,500, or a 50% chance of $10,000 (else $0). Loss frame: a sure −$4,500, or a 50% chance of −$10,000 (else $0). (a) Compute each gamble's EV. (b) Predict the typical choice in each frame. (c) Map the pattern onto a portfolio and name the behavior.

Solution. (a) Gain gamble EV = 0.5 × 10,000 = $5,000, and loss gamble EV = 0.5 × (−10,000) = −$5,000. (b) In the gain frame most people take the sure $4,500, giving up $500 of EV to lock in a certain win. That's risk-averse in gains. In the loss frame most take the gamble, accepting a $500-worse EV to avoid a certain loss. That's risk-seeking in losses. Same person, opposite risk attitude, flipped only by the frame. It's the reflection effect, and notice that it sacrifices EV in both directions. (c) On a portfolio this is the disposition effect: sell winners (grab the sure gain), hold losers (gamble on a recovery). The fix is the fresh-money test, which removes the reference point that creates the two frames.

Problem 18 (timed, 8 minutes, stretch: reflexivity + Lollapalooza + numbers). "InstaCred" (synthetic, India) is a fast-growing NBFC funding 3-year loans with 90-day commercial paper. Its stock has 5×'d in a year and retail investors are piling in because "everyone's making money and the founder is a visionary." (a) Answer in a single sentence: is this reflexivity-dominant, and why? (b) Name four biases/tendencies driving the retail crowd. (c) InstaCred funds ₹10,000 crore of loans yielding 13% with commercial paper costing 8%: compute the annual net spread income. (d) A confidence shock forces it to roll ₹4,000 crore of that funding at a penalty 15% (the other ₹6,000 crore stays at 8%); recompute the net spread and state, in Soros's terms, what just happened.

Solution. (a) Yes, reflexivity dominates. The NBFC's access to short-term funding depends entirely on the market's confidence in its solvency, so perception feeds the fundamental. (b) Any four of: social proof/herding ("everyone's making money"), envy (others getting rich), recency (5× extrapolated forward), narrative fallacy ("visionary founder"), overconfidence (piling in savings). (c) Loan income = 13% × 10,000 = ₹1,300 cr and funding cost = 8% × 10,000 = ₹800 cr, so net spread = ₹500 cr. (d) New funding cost = 8% × 6,000 + 15% × 4,000 = 480 + 600 = ₹1,080 cr. Net spread = 1,300 − 1,080 = ₹220 cr, a 56% collapse ((500 − 220) ÷ 500) from the funding shock alone, before any forced asset sales. In Soros's terms, the perception of weakness raised the cost of funding. That damaged the actual fundamental, spread income, which confirms and deepens the fear. It's a reflexive run, the IL&FS/DHFL failure mode, and it's why a "cheap" financial is cheap for a reason.

Problem 19 (guided: sort the bias, then choose moderate or adapt). For each of the following, name the bias from the exam taxonomy, classify it as a cognitive error or an emotional bias using the diagnostic question, and state whether an adviser should moderate it or adapt to it. (a) A client insists that Indian small caps "always recover within a year", having watched two such recoveries. (b) A client refuses to sell a holding that is 40% underwater, and can state clearly that the position no longer has a thesis. (c) A client's retirement contributions have been split across the same three funds since 2011, and the split no longer matches any stated goal. (d) A client compares two identical portfolios and prefers the one whose factsheet reports a 90% chance of meeting the goal over the one reporting a 10% chance of missing it.

Solution. (a) Representativeness, specifically the sample-size fallacy, and a cognitive error. Two observations aren't a base rate, and showing the actual frequency of one-year recoveries changes the answer. Moderate. (b) Loss aversion with regret aversion underneath, and emotional. The client already has the correct analysis and still can't act, which is just what the diagnostic question is for. Adapt, using a pre-committed sell rule written when no position is underwater. (c) Status quo, and emotional in origin, though it looks like an oversight. Nothing is stopping the client except that revisiting is a decision and inaction isn't. Adapt with structure, meaning a dated review and a rule-based rebalance in place of an intention. (d) Framing, and cognitive. Showing both frames side by side resolves it in seconds, which no emotional bias would allow. Moderate.

Problem 20 (independent: price the accommodation). A client with ₹500 lakh is advised, on risk tolerance and horizon, to hold 65% equity and 35% debt. Because the client mentally accounts, the adviser instead funds three goals: ₹120 lakh of essential spending at 0% equity, ₹230 lakh of lifestyle goals at 60% equity, and ₹150 lakh of aspirational goals at 90% equity. Expected returns are 11.0% on equity and 6.5% on debt. (a) Compute the aggregate equity weight the bucket structure produces. (b) Compute the expected return of the advised allocation and of the bucketed one, and the annual cost of the accommodation. (c) The client has panicked and sold once in the last ten years. State the case for and against the accommodation in one sentence each.

Solution. (a) Equity rupees = (120 × 0) + (230 × 0.60) + (150 × 0.90) = 0 + 138 + 135 = ₹273 lakh, which is 273 ÷ 500 = 54.6% equity and 45.4% debt. (b) Advised: 0.65 × 11.0 + 0.35 × 6.5 = 9.425%. Bucketed: 0.546 × 11.0 + 0.454 × 6.5 = 8.957%. The accommodation costs 9.425 − 8.957 = 0.468 percentage points a year, roughly 5.0% of the expected return given up. (c) For: one panic sale in ten years is a demonstrated behaviour, not a hypothetical one. A structure that keeps five years of essential spending visibly untouchable removes the trigger that produced it. Against: 0.468 points a year is a large price for a structure that changes nothing about the underlying exposure. The same adherence might be bought with a written rule and a review cadence, at no cost to expected return.

Problem 21 (timed: 6 min, the eight, applied). Name the bias each sentence describes, using only the eight the catalog leaves out, and give its category. (a) "I have read the downgrade, but I am keeping my estimate and taking it down a bit each quarter." (b) "I built the model myself, so I am comfortable making this a 25% position." (c) "These are my dividends, so I can spend them; selling shares would be eating capital." (d) "I know I should start the systematic investment plan, and I will, from next month." (e) "I would rather be wrong with everyone else than be the only one who bought it."

Solution. (a) Conservatism, cognitive. The speaker updates too little when they should rebuild once. (b) Illusion of control, cognitive. Effort on the model has turned into confidence about outcomes the model doesn't govern. The fix is the sizing rule you'll meet later, where size follows the range of outcomes and ignores the depth of the work. (c) Mental accounting, emotional. A dividend rupee and a sale rupee are the same rupee, and the distinction changes the spending rule without changing the wealth. (d) Self-control, emotional. The person agrees with the action and can't take it now, so the counter-practice is an automated instruction, not a better argument. (e) Regret aversion, emotional, in its herding form. A consensus loss carries less regret than a solitary one, so the discomfort is real even when the analysis isn't.

Applied mini-project: the behavioral post-mortem of a real episode

The task. Choose one real market episode where behavior clearly overwhelmed analysis. Write a 3 to 4 page behavioral post-mortem that decomposes it into its psychological drivers and pulls out a permanent lesson for your own process. Pick a bubble/mania, a fraud/blow-up, or a crash/panic, and work from primary and near-primary sources. A Wikipedia summary won't be enough. Over the program you'll want to build both muscles, Indian and US, so do at least one of each across this project and the phase capstone. Strong candidates:

• India: Satyam (2009); the IL&FS/DHFL NBFC crisis (2018) as a reflexivity case; the Harshad Mehta scam (1992) or Ketan Parekh (2001) as social-proof/authority manias; a recent SME-IPO or micro-cap frenzy as a recency/envy/herding case; a specific new-age IPO where the narrative outran the price.
• US/global: the dot-com bubble (1999–2000); the 2008 housing/credit bubble; Enron (2001), Valeant (2015), or Theranos as fraud/narrative cases; the 2021 meme-stock episode (GameStop) as envy/social-proof at internet scale; Long-Term Capital Management (1998) as overconfidence.

What to pull and where. For a fraud or blow-up, get the last clean annual filing before the collapse (SEC EDGAR for US 20-F/10-K, BSE/NSE or the company site for Indian ARs). Then look for the forensic tells you learned to spot in Phase 2 (related-party notes, cash that shouldn't exist, promoter behavior, auditor changes). For a bubble or crash, pull valuation data over the run-up and collapse, meaning the multiple against history (Damodaran's data library, screener.in, stockanalysis.com). Add issuance quality (IPO volumes and post-listing performance) and contemporaneous commentary from news archives and from memos and letters written at the time. Oaktree memos are gold for this. For any episode, find at least one contemporaneous bear voice (a short report, a skeptical journalist, a dissenting memo). It shows you what was knowable in real time, as opposed to only in hindsight.

The deliverable. Structure it like this:

1. The episode in one paragraph. What happened, the timeline, the magnitude (₹ crore or $bn of value created and destroyed).
2. The behavioral decomposition. Name the four-plus biases/tendencies that stacked (the Lollapalooza), with specific evidence for each from the sources. Generic labels don't count. If it's a reflexivity case, trace the feedback loop explicitly: what fed back into what.
3. What was knowable in real time. Cite the contemporaneous bear voice and the specific tells in the filing/data, and judge fairly how visible they were before the outcome. Guard against your own hindsight bias here. This is the hardest and most important section.
4. Where the sentiment pendulum was. Locate the episode on Marks's stages and take the temperature using the observable tells (valuation, issuance quality, media/amateur behavior, the language of "this time is different").
5. The transferable lesson and the structural fix. Write the one permanent checklist line or journal prompt this episode earns, phrased so it would help you catch the next instance wearing different clothes.

Scoring rubric. Score each criterion 0/1/2, for 16 points in total. You need ≥12 to pass.

Why this project. It welds everything here to a real case and to the skills of Phases 2, 3, and 5. It also produces two durable assets for your permanent system: a company/episode file you keep forever, and at least one new behavioral line for your checklist. And it's a direct rehearsal for the phase capstone and for the "steelman the bear / pre-mortem" work of the deep dive.

Case Lab: run case CL-6.1 in the app's Cases tab for this module.

Reading & resources

• Daniel Kahneman, Thinking, Fast and Slow (2011). The foundational text. Part I (Systems 1 & 2), Part III (overconfidence, WYSIATI, the illusion of validity), and Part IV, chs. 26 to 28 (prospect theory, the endowment effect, and the fourfold pattern) are the core reading here. Dense but essential; read it slowly. [Paid] [Intermediate]
• Charlie Munger, "The Psychology of Human Misjudgment", in Poor Charlie's Almanack (the revised 2005 version with all ~25 tendencies). The talk is also freely available as a transcript and audio online. It's the most important reading for the Munger material. Re-read it every year. [Book Paid / Transcript Free] [Intermediate]
• Howard Marks, The Most Important Thing (2011): chs. on "The Most Important Thing Is... Understanding Risk," "Being Attentive to Cycles," "Awareness of the Pendulum," and "Combating Negative Influences." And the Oaktree memos (oaktreecapital.com/insights): free, decades deep. For this topic, read "On the Couch" (behavior and cycles), "Yet Again?", and "The Most Important Thing." [Book Paid / Memos Free] [Beginner→Intermediate]
• George Soros, The Alchemy of Finance (1987): the Introduction and Part I on the theory of reflexivity; hard going (Soros is not a natural expositor) but the source. For a gentler route, Soros's later lectures and essays ("Fallibility, Reflexivity, and the Human Uncertainty Principle," 2013, free) restate it more clearly. [Book Paid / Essays Free] [Advanced]
• James Montier, The Little Book of Behavioural Investing (2010), the best practical single volume: short, concrete, investing-specific chapters mapping each bias to an investing failure and a fix. If you read one applied book after Kahneman, read this. His larger Behavioural Investing (2007) is the reference tome. [Paid] [Beginner→Intermediate]
• Richard Thaler, Misbehaving (2015), the history and logic of behavioral economics, including the endowment effect and mental accounting, told engagingly by the field's other Nobel laureate. [Paid] [Beginner]
• Terrance Odean, "Are Investors Reluctant to Realize Their Losses?" (Journal of Finance, 1998) and Barber & Odean, "Trading Is Hazardous to Your Wealth" (2000) and "Boys Will Be Boys" (2001) are the primary empirical papers behind the disposition-effect and overconfidence numbers; all findable free as PDFs. Read at least the abstracts and results tables to see the real magnitudes. [Free] [Intermediate]
• Philip Tetlock, Superforecasting (2015, with Dan Gardner), the constructive sequel to Expert Political Judgment. It shows how calibration and process produce forecasters who actually beat chance, and it bridges the diagnosis here to the calibration practice in M9. [Paid] [Beginner→Intermediate]
• Jason Zweig's commentary in The Intelligent Investor (revised ed.) and his Your Money and Your Brain (2007): behavioral finance woven into Graham; excellent for seeing how old the core insight is ("the investor's chief problem, and even his worst enemy, is likely to be himself"). [Paid] [Beginner]
• Guy Spier, The Education of a Value Investor (2014): the environment-design case study, and how one investor engineered his surroundings and habits to protect his behavior. [Paid] [Beginner]
• India ties. Parag Parikh, Value Investing and Behavioral Finance (2009): behavioral finance written directly for the Indian market, with Indian examples, and the natural first behavioral book for you. [Paid] [Beginner→Intermediate]. Sanjay Bakshi's BFBV ("Behavioural Finance and Business Valuation") lectures and blog (Fundoo Professor): free, deep, India-centric, and unusually good on Munger's latticework applied to Indian companies and frauds. [Free] [Intermediate]. Gautam Baid, The Joys of Compounding: a chapter-length synthesis of temperament and behavior for the Indian investor. [Paid] [Beginner].
• Free video/course: the Behavioural Finance module within any reputable free finance MOOC, plus Kahneman's and Munger's talks on YouTube; Sanjay Bakshi's lectures where available. Munger's USC and Harvard talks are free and short. [Free] [Beginner→Intermediate]
• Going deeper: Michael Mauboussin, Who Is on the Other Side? (Morgan Stanley Counterpoint Global, 2019; free PDF) [Free] [Intermediate]. A rigorous taxonomy of where edge actually comes from (behavioral, analytical, informational, technical) and how to identify the biased counterparty whose error is your opportunity. It's the natural sequel to "differently right" from the opening. Optional now; ideal alongside M9.02.
• Going deeper: Annie Duke, Thinking in Bets (2018) [Paid] [Beginner→Intermediate]. A poker champion's practical treatment of decision quality versus outcome ("resulting" is her name for outcome-driven hindsight), probabilistic belief-updating, and building a truth-seeking group. It bridges the diagnosis here to the M9.04 calibration and decision-journal practice.

The Modern Analyst's Addendum

Everything above teaches this skill from first principles, by hand. That's how you learn it, and the mastery check still tests it that way. This addendum shows how a working analyst amplifies the same skill today. It adds; it never replaces. (R1/R10)

AI-Augment this skill

{
  "skill": "Prospect theory (reference dependence, loss aversion, the reflection effect), the disposition effect quantified via PGR/PLR, the bias catalog, Munger's tendencies and the Lollapalooza effect, Soros's reflexivity, Marks's pendulum, and the structural counter-practices (checklist, journal, pre-commitment, environment design)",
  "use": "Two narrow jobs, and one absolute ban. (1) PRACTICE MATERIAL: ask it to draft short scenario vignettes that each exhibit exactly one NAMED bias, unlabeled and in random order, for you to diagnose blind and self-check — or to draft the strongest disconfirming ('steelman the bear') case against a thesis you already hold, as a structured adversarial list for you to go verify. (2) EXTRACTION: turn a real brokerage export (a trade-history CSV or a CAS/1099-B) into the clean cost-basis / current-value / sold-flag table the PGR/PLR calculator needs. THE BAN: never ask it to diagnose which bias YOU are exhibiting in a live decision, or whether your live thesis is sound — that self-diagnosis is exactly what this module trains, and a model asked to check your reasoning tends toward agreeable confirmation, which is uncomfortably close to the confirmation bias (§4.4/catalog) it would be checking for.",
  "tools": ["Chat assistants — Claude, ChatGPT for vignette generation and steelman-the-bear drafting", "A spreadsheet or notebook — where PGR/PLR and the loss-aversion sensitivity table actually get computed", "Zerodha Console or a CDSL/NSDL Consolidated Account Statement (India) and Schwab/Fidelity/E*TRADE trade-history or 1099-B exports (US) — real position data to run the calculator on"],
  "prompt": "Draft three short (3-4 sentence) investment scenario vignettes, each exhibiting exactly one of these biases without naming it: anchoring, disposition effect, overconfidence. Number them 1-3 in a random order and do not reveal which is which anywhere in your answer — I will diagnose each one myself and check afterward.",
  "verify": "Compute PGR and PLR by hand on one small extracted book to confirm the parser before trusting it on a full history; before repeating any AI-cited study or effect size — a lambda, a percentage, an 'X times more likely' — find the paper by name and check the number against the abstract yourself.",
  "diy": "Without a tool in the room: compute PGR and PLR from a position list and state what the ratio implies; compute the felt value of a gamble under a stated loss-aversion coefficient and the breakeven upside it implies; run the fresh-money test on a real holding and say, honestly, whether cost basis is influencing the answer; and name, from memory, at least four biases from the catalog together with the specific mechanism — not just the label — that produces each one."
}

Modern Data Analysis

By hand first. WE1 computes PGR and PLR from a given ten-row table you can eyeball. A real portfolio has dozens of positions across years, and one snapshot's ratio matters less than whether the tilt is getting worse or better over time.

Today's workflow. Export the real transaction and holdings history, compute PGR and PLR programmatically per quarter, and plot the ratio over time. A rising PGR/PLR is the disposition effect getting worse, and you can see it long before it shows up as underperformance. The same extracted table feeds the fresh-money test too. For every open position, log cost basis next to current value and your own forward return estimate, stated without spin. Then "would I buy this today" has a number attached, not just a feeling.

Tools & sources (IN + US). India: Zerodha Console's tradebook/P&L export (CSV) or the CDSL/NSDL Consolidated Account Statement (CAS), both giving cost basis and current holdings directly. US: brokerage trade-history or 1099-B exports (Schwab, Fidelity, E*TRADE) or a manually maintained blotter.

# recomputed in-session — R3
def pgr_plr(positions):
    """positions: list of dicts with cost, current, sold(bool). Mirrors Odean (1998)."""
    winners = [p for p in positions if p['current'] > p['cost']]
    losers  = [p for p in positions if p['current'] < p['cost']]
    win_sold = sum(1 for p in winners if p['sold'])
    los_sold = sum(1 for p in losers if p['sold'])
    pgr = win_sold / len(winners) if winners else float('nan')
    plr = los_sold / len(losers) if losers else float('nan')
    return pgr, plr

# WE1's own ten-position book, run through the function as the verification anchor
book = [
    {'cost': 100, 'current': 160, 'sold': False}, {'cost': 100, 'current': 135, 'sold': False},
    {'cost': 100, 'current': 120, 'sold': True},  {'cost': 100, 'current': 108, 'sold': True},
    {'cost': 100, 'current': 104, 'sold': True},  {'cost': 100, 'current': 96,  'sold': True},
    {'cost': 100, 'current': 85,  'sold': False}, {'cost': 100, 'current': 78,  'sold': False},
    {'cost': 100, 'current': 64,  'sold': False}, {'cost': 100, 'current': 55,  'sold': False},
]
pgr, plr = pgr_plr(book)
print(f"PGR={pgr:.2f}  PLR={plr:.2f}  ratio={pgr/plr:.1f}")

Verify. pgr_plr(book) on WE1's ten-row table returns PGR=0.60, PLR=0.20, ratio=3.0, which matches WE1 Steps 1 and 2. The function adds nothing WE1 didn't already do by counting. Its value is scale. It handles a real, hundred-position history without the counting errors that creep in past a dozen rows, and it turns a quarterly PGR/PLR trend line into one line of code, so you don't recount by hand every time.

Quantitative lens

The prospect-theory section gives the loss-aversion coefficient as λ ≈ 2.25 (Tversky & Kahneman, 1992), and it correctly hedges it as an order-of-magnitude constant that varies by person and context. WE5 then builds a specific verdict on that one number: declining a 50/50 bet of +₹1,50,000 / −₹1,00,000. That raises two questions worth running numbers on. How sensitive is WE5's verdict to the exact λ? And what does the actual replication literature say λ is?

How sensitive is the verdict? Re-run WE5's felt-value formula (felt = p·gain − (1−p)·λ·loss) across λ = 1.00 to 3.00 on the identical bet:

The objective EV (+₹25,000) doesn't depend on λ, since it isn't a felt quantity. The accept/decline verdict, though, flips at λ = 1.50 and stays "decline" for every λ above it. So WE5's conclusion isn't equally sensitive everywhere. It holds firmly for λ comfortably above 1.5, and it's undetermined near it.

What does the replication literature say λ actually is? Here the course's recurring caution, "verify against the paper," matters most. Walasek, Mullett & Stewart's meta-analysis of loss-aversion studies (Journal of Economic Psychology, 2024) pools the available choice-based estimates and reports λ = 1.31, 95% CI [1.10, 1.53]. That pooled estimate is barely above this bet's flip point, and its confidence interval straddles it. The paper's own qualification matters as much as the number. "Surprisingly few studies" cleanly estimate λ from mixed gambles, and much of the underlying data is "of poor quality," so don't treat the 1.31 as a fixed replacement for T&K's 2.25. It's one pooled estimate from a thin, heterogeneous literature.

Two other studies fill in the picture. Ruggeri et al.'s 19-country replication of prospect theory (Nature Human Behaviour, 2020) found that the qualitative patterns (the reflection effect, the direction of loss aversion) replicate broadly across cultures, while parameter magnitudes like λ vary by sample. And Yechiam's review "Acceptable losses: the debatable origins of loss aversion" (Psychological Research, 2018) argues that much of the apparent strength of the 2.25-scale estimate comes from hypothetical, low-stakes choice tasks. Those are methodologically close to the illustrative ₹8,000-vs-80%-of-₹10,000 pair in the prospect-theory section. On that reading, much of that estimate reflects how the question is asked, and it isn't a universal law that holds however you ask.

What survives and what is contested, said plainly. Three findings rest on revealed behavior in actual market or brokerage data, replicated across decades and markets. They're momentum (Jegadeesh & Titman, "Returns to Buying Winners and Selling Losers," Journal of Finance, 1993), post-earnings-announcement drift (Bernard & Thomas, Journal of Accounting Research, 1989), and the disposition effect measured on real trading records (Odean, 1998, cited in the prospect-theory section). The exact multiplier on loss aversion is a different kind of claim. It's a single parameter estimated mostly from hypothetical lab choices, and its modern pooled estimate (1.31) is close enough to WE5's flip point (1.50) that the verdict itself is live research. So is the number.

Know the limits of all this. None of it overturns loss aversion as a real phenomenon whose direction holds up. The reflection effect and the general tendency to weight losses more than equivalent gains show up across the cited replications. What's uncertain is the exact multiplier to use in a calculation like WE5's. The course states T&K's 2.25 because it's the number the field still cites and teaches. The addendum hasn't resolved the dispute. Treat any single λ, including 2.25 and including 1.31, as a point estimate from one study design. Give it the same scrutiny you'd give a single backtested Sharpe ratio, and never treat it as a physical constant to six-decimal confidence.

Where this goes next: galaxy cross-links

• G1 · QM1.02 (Hypothesis Testing & Regression): the machinery (confidence intervals, pooled estimates, heterogeneity) behind reading a meta-analysis like Walasek, Mullett & Stewart (2024) correctly rather than quoting its headline number alone.
• G2 · DA1.03 (pandas: Cleaning, Time Series & the Bridge): turning a real brokerage export into the clean, sold-flagged table the PGR/PLR function above needs.
• G3 · CN1.02 (Hypothesis-Driven Problem Solving): the disconfirming-evidence discipline behind "steelman the bear," generalized past investing to any analytical claim.

Teach it back & journal

Feynman prompt. Write a one-page explainer for a smart 15-year-old titled "Your own brain is the person you're really trading against." Keep it free of jargon, and get three ideas across. First, the price of a stock is just what a huge crowd of people currently believes, so to make money you have to be right in a way the crowd isn't. But the crowd runs the same buggy brain-software you do, and the software has predictable bugs. It hates losing about twice as much as it likes winning, it copies whatever everyone else is doing, it falls for a good story, and it thinks whatever just happened will keep happening. Second, the market is set up to trigger those bugs at the worst possible moments, showing you rising prices when things are expensive and crashing prices when they're cheap. Third, you can't beat the bugs by deciding to be smart. You beat them by building simple rules and habits ahead of time, the way Ulysses tied himself to the mast before the Sirens sang. If your page needs the words "prospect theory," "reflexivity," or "System 1," you don't own it yet.

Journal prompt. Four entries. (1) Of the eleven biases, which two do you most suspect you personally run? Give the specific evidence from your own past behavior, a real decision, and skip the flattering generality. (2) Do the calibration exercise in Worked Example 4 without fudging it, and record how many of your ten "90% confident" intervals actually contained the truth. Write one sentence on what that number means for how you should treat your own DCF outputs. (3) Take the last real decision you made under uncertainty (investment or not) and write down, now, what you think you believed beforehand. Then ask how much of that is memory and how much is reconstruction, and what a contemporaneous journal entry would have changed. (4) Name the one structural practice (a checklist line, a journal prompt, a pre-commitment, an environment change) you'll actually adopt this week, and write its exact wording. It becomes a line in your checklist v1 in the capstone.

This module's flashcards and mastery quiz are wired into the app: see the node's Quiz and Reviews.