The Library

The fuller mechanism

Volatility is just the standard deviation of returns (how far results typically stray from their average) — a ruler, not a verdict. The Sharpe ratio lays that ruler against the return: (return − safe rate) ÷ volatility. The safe rate is what T-bills — short-term U.S. government IOUs — pay, and subtracting it matters: beating it by 2% with wild swings is a worse achievement than beating it by 2% calmly, and a return below the safe rate was never worth the danger at all. Professionals compare Sharpe ratios across strategies the way you'd compare fuel efficiency across cars — the raw speed tells you nothing about the cost of getting there.

A close cousin, the Sortino ratio, counts only downside volatility (the bad wiggles). Nobody ever complained about upside swings, so Sortino argues they shouldn't count against you. Same idea, fairer ruler: reward per unit of danger that actually hurt.

Where it breaks

Volatility treats a 5% jump and a 5% plunge as identical. That blinds it to the most dangerous shape in markets: strategies that win small and steadily, then lose catastrophically all at once — selling insurance against rare disasters. Their volatility looks tiny right up until the disaster, and their Sharpe ratios look brilliant. Nassim Taleb's image: a turkey fed every day, getting statistically more confident in the farmer, until Thanksgiving. Any ratio built on past volatility inherits this blindness — it measures the danger you've seen, not the danger that exists.

It also breaks across time: a three-year Sharpe says nothing about the next crisis, and leverage (borrowed money) can cosmetically inflate risk-adjusted numbers while quietly multiplying the real risk. A ratio is a summary of history, not a forecast.

A real episode

Long-Term Capital Management ran one of the great Sharpe ratios of the 1990s — a Nobel-backed bond-arbitrage fund that harvested tiny, "safe" price gaps with enormous leverage. When Russia defaulted in August 1998, the gaps didn't close; they exploded, correlations spiked, and liquidity vanished — the exact risks volatility never measured. The fund lost about $4.6 billion in months, and fourteen banks injected $3.6 billion to wind it down without detonating the system. Brilliant risk-adjusted returns, right up to the moment the adjustment was wrong.

The usual mistakes

Comparing raw returns without asking "per what risk" — the mistake the whole chapter exists to kill. Trusting a sparkling Sharpe from a short track record (two good years prove almost nothing). And the subtlest: assuming low volatility means safe. A savings account has near-zero volatility and a near-zero real return; safety and stillness aren't the same thing.

What to do

Always divide by the danger: ask for the Sharpe or Sortino, the period it covers, and whether that period included a real crisis. Prefer measures that punish downside specifically — maximum drawdown (the worst peak-to-trough fall, see Idea 02) tells you what the pain actually felt like. And remember the hierarchy: volatility is a ruler, permanent loss of capital is the risk, and no ratio survives contact with leverage it didn't account for.

The fuller mechanism

The formula from the short read — needed gain = loss ÷ (1 − loss) — comes straight from the arithmetic: after a loss L you hold (1 − L), and you need a gain G such that (1 − L) × (1 + G) = 1. Solve for G and the table writes itself: a 10% loss needs 11.1%, 20% needs 25%, 30% needs 42.9%, 40% needs 66.7%, 50% needs a full 100%, 60% needs 150%, 80% needs 400%, and a 90% loss needs 900% — a ten-bagger — just to break even.Computed.

Notice the cruelty curve: the first 10 points of loss cost you 11.1 points of required gain; the last 10 (from 80% to 90%) cost 500. Drawdowns don't scale — they accelerate. That's why professionals think in drawdowns first and returns second: the return is the reward, but the drawdown is the constraint that decides whether you survive to collect it.

Where it breaks

The table's silent assumption is survival. It has two quieter ones. A diversified index usually survives; an individual company can go to zero, and no recovery math rescues −100%. The table also assumes you can hold through the fall — but margin calls (Idea 03), fund redemptions, job losses, and plain panic force selling at the bottom, converting a paper drawdown into a permanent one. And recovery takes time, which the table never shows: the math says nothing about how many years the climb back requires.

A real episode

The Nasdaq Composite peaked at 5,048 in March 2000, then fell 78% to 1,114 by October 2002. The table says a 78% loss needs roughly a 350% gain to recover — and the market delivered it, eventually: the index didn't reclaim 5,048 until April 2015, fifteen years later. The S&P 500's 2007–2009 fall was kinder — 57% down, needing about 132% back — reclaimed by 2013. Japan is the warning: the Nikkei peaked at 38,915 in December 1989 and spent more than three decades below it in nominal terms. Same math, three different clocks — and the clock is the part nobody controls.

The usual mistakes

"Buying the dip" without checking the table — a stock "down 50%, half price!" is actually a bet that it can double from here. Averaging down: it lowers your average cost but commits more dollars to the same thesis — the hole gets wider even as it gets shallower. And confusing volatility with drawdown: a stock that swings ±5% daily feels terrifying but may never draw down much; a stock that slides 60% quietly over two years feels fine until you do the math.

What to do

Run the table before adding to any loser — write down the required gain and ask if you truly believe it's coming. Size every position so its plausible worst drawdown can't end you (this is what Idea 05's 25% cap is for). Keep cash outside the portfolio for the recovery — the investors who bought the 2002 and 2009 bottoms weren't geniuses, they were solvent. Survival isn't the boring part of investing. It's the whole game the table is describing.

The fuller mechanism

Three numbers govern every margin account. Initial margin (50% under Regulation T — the Fed's stock-borrowing cap) sets the starting line. Maintenance margin — the FINRA minimum is 25% of the position's value, though most brokers demand 30–40% as a house requirement (their own stricter cushion) — sets the tripwire. And the loan itself never shrinks when prices fall, which is the entire trap.

The call price has a formula. With loan L and maintenance requirement m, the call fires when the position value P falls below L ÷ (1 − m). The short read's example: L = $100,000, m = 25% → the call at $100,000 ÷ 0.75 = $133,333, exactly the 33.3% fall. Raise the house requirement to 35% and the same loan gets called at $153,846 — a mere 23% fall. Know your broker's number, not just the regulator's.Computed.

Then the part nobody dwells on: the broker liquidates without asking you. They choose what to sell, they sell into the falling market, and they charge you for the privilege — while interest on the loan kept accruing daily the whole way down.

Where it breaks

Gap risk: prices don't always slide politely through your tripwire — overnight news can gap a stock 30% past it before you can act, and the call arrives when adding cash is hardest (weekends, crises, your own cash crunch). Correlation risk: your "diversified" margin portfolio falls as one position in a panic, so five small cushions become one big margin call. And futures margin plays by different, harsher rules — daily cash settlement, no waiting.

A real episode

In 1929, 10% margin was common — $1 of your money controlled $10 of stock. When the market fell, the calls cascaded: forced selling pushed prices lower, triggering more calls, in the loop that turned a crash into the Crash. Nearly a century later, Archegos Capital replayed it with modern plumbing: Bill Hwang's family office used total return swaps (contracts that mimic owning the stock) to build enormous leveraged positions invisibly — the banks held the shares while he held the exposure. When a few holdings wobbled in March 2021, prime brokers — the banks that finance hedge funds' trades — all sprinted for the same exit.

Roughly $20 billion of positions unwound in days, and Credit Suisse alone lost $5.5 billion. Same lesson, new instruments: leverage manufactures fragility, and the call always comes at the worst moment.

The usual mistakes

Maxing out initial margin on day one — starting with zero buffer. Not knowing the house maintenance percentage (it's in your agreement; read it). The classic: "I'll just add cash if I get called" — in a crisis, cash is precisely what's scarce, including yours. And using margin for buy-and-hold: at 8% margin interest, a $100,000 loan costs $8,000 a year before the market moves at all — a guaranteed negative return hunting an uncertain positive one.Illustrative.

What to do

Compute your call price before you borrow — P below L ÷ (1 − m) — and write it down. Keep your equity at least double the maintenance minimum, so normal volatility can't reach you. Never meet a margin call with more borrowing. And size the loan so that a 50% market fall still leaves you uncalled: if that constraint makes the trade too small to matter, the trade was too dangerous to make.

The fuller mechanism

The short read's four tolls punish trading; the deep read adds two more — the ones that punish holding. Commissions, the bid-ask spread, market impact, and borrow fees are the trading tolls; fund expense ratios (the yearly skim, e.g. 0.03% vs 1.10%) and taxes are the holding tolls. A "free" trade in a taxable account that you close in eleven months can easily cost more in short-term capital gains tax (taxed as ordinary income) than every other toll combined.

Now the compounding cruelty. Take $10,000 growing at 7% for 30 years: about $76,100. Shave one point off for fees — 6% — and you get about $57,400. That single percentage point, compounded, confiscated roughly a quarter of your terminal wealth: $18,700 gone, not to the market, but to the fee schedule. Costs don't subtract from your returns. They compound against them, on the same curve, for the same decades.Computed.

Where it breaks

Not every cost is pure waste. Market impact is the unavoidable price of size — moving millions costs more than moving thousands, and no broker can fix that. "Commission-free" trading isn't cost-free either: brokers route your orders to wholesalers who pay them for the flow (payment for order flow — the broker sells your order stream), and the spread you cross is still the market's toll. The honest question is never "are there costs" but "what am I getting per dollar of cost" — and for most active trading, the answer is: less than the dollar.

A real episode

Every year, S&P's SPIVA scorecard asks a brutal question: what fraction of professional fund managers beat their benchmark? Over 10- and 15-year horizons, the large majority of U.S. large-cap managers underperform the S&P 500 — and costs explain a large share of the persistent gap between active managers and their benchmarks. Jack Bogle's summary was arithmetic, not philosophy: the market's return minus costs equals the investor's return, so minimizing costs is the one reliable way to keep more of it.

The hedge-fund version is starker: "2 and 20" — a 2% management fee plus 20% of profits (usually measured net of the management fee). On $1 million earning 10% gross, that's $20,000 off the top plus 20% of the remaining $80,000 gain — $16,000 — for $36,000 total, or 36% of the year's profit, before you've seen a dime.Illustrative.

The usual mistakes

Churning a "commission-free" account — the spread and taxes don't care about your broker's marketing. Ignoring the expense ratio because 0.90% "sounds tiny" next to 0.05% (run the 30-year math above). Day-trading against professionals with structural cost advantages is bringing a knife to a gunfight. And the meta-mistake: optimizing for the lowest visible fee while ignoring the biggest one, taxes.

What to do

Add up every cost before any strategy — commissions, spreads, impact, borrow, expense ratios, taxes — and demand the strategy earn back its total toll with room to spare. Build the core of your portfolio from broad, low-cost index funds and touch it rarely. Hold winners past a year where the tax code rewards patience. And tattoo the hierarchy somewhere visible: returns are uncertain, costs are certain, and certainty is the only edge you get for free.

The fuller mechanism

Diversification works through correlation — the −1 to +1 measure of how two assets move together. Pair assets that zig while others zag and the portfolio's wobble shrinks even though each holding still wobbles alone. Finance splits risk in two: idiosyncratic risk (one company's bad quarter — diversifiable) and systematic risk (the whole market falling — not diversifiable). Spreading across twenty stocks kills the first almost completely. It does nothing to the second, ever. The 25% single-issuer cap from the short read is this idea as a rule: no single −100% can ever take more than a quarter of you.

Where it breaks

In 2008, correlations shot toward 1. Stocks, REITs (real-estate investment trusts — companies that own property), commodities, high-yield bonds — the "diversified" toolkit — fell together as forced sellers sold what they could, not what they should. A classic 60/40 stock/bond mix still lost about 20% that year — 0.6 × −37% stocks plus 0.4 × +5.2% bonds; Treasuries, cash, and little else held their ground. The lesson: diversify across return drivers (growth, safety, cash), not across labels that all quietly depend on the same driver (credit, liquidity, optimism). And there's a second failure mode in the other direction — diworsification: owning so many overlapping funds that you've built an expensive index fund with extra steps, paying active fees for market returns.Computed.

A real episode

The quant crisis of August 2007 was the dress rehearsal. Dozens of "market-neutral" equity funds — diversified across hundreds of long and short positions, hedged against market direction — all started losing at once. Their models assumed independent, diversifiable risks; in reality they'd all crowded into the same factors (shared return drivers like value and momentum), and when one fund deleveraged, the selling hit everyone's book simultaneously. Diversification across positions had masked concentration in strategy. 2008 then ran the same film at economy scale.

The usual mistakes

Owning twenty tech stocks and calling it diversified (that's one bet with twenty tickers). Home-country bias — Americans overweighting America, Canadians overweighting Canada — concentrating in the economy you already depend on for your job. Adding assets you don't understand because a listicle said "diversify" (complexity isn't diversification). And never rebalancing: winners grow back into the concentration you diversified away from, which is why the fix has to be scheduled, not felt.

What to do

Spread across drivers, not names: productive assets (stocks), safety assets (high-quality bonds, T-bills), and cash — each behaves differently when it matters. Cap any single position (25% is a sane ceiling). Rebalance on a calendar — yearly is plenty — selling a little of what's grown and buying what's shrunk, which mechanically forces you to sell high and buy low. And accept the residual truth: the one risk you can never diversify away is the market itself. That's what the return is paying you for.

The fuller mechanism

The short read's 7,846 tested rules hide a piece of arithmetic worth seeing plainly. Test one rule at the standard 5% significance bar and there's a 5% chance it looks good purely by luck. Test 7,846 and you'd expect about 392 lucky winners — 7,846 × 0.05 — before any real edge exists at all. This is the multiple-testing problem (the more guesses, the more false hits), and it's why serious researchers apply corrections with names like White's Reality Check and Hansen's superior predictive ability test — statistical penalties for how many rules you tried.

The honest workflow has three stages. In-sample is finding the pattern; out-of-sample is testing it on data the rule never saw; forward paper-trading is watching it on data nobody has seen yet. Most "proven systems" for sale never survive stage two.Computed.

Where it breaks

Even honest backtests rot. Markets adapt: once an anomaly is published, money piles in and the edge decays — one well-known study found anomaly returns fall by more than half after publication (McLean & Pontiff, 2016). Databases quietly drop dead funds and delisted stocks (survivorship bias — the losers vanish, flattering every rule). Look-ahead bias sneaks in when a test accidentally uses information that wasn't available at the time (earnings restated later, index members known only in hindsight). And costs are almost always understated: the backtest trades at the mid-price while you pay the spread, the slippage, and the commissions from the short read's Idea 04.

A real episode

The January effect — small stocks outperforming in January — was a genuine, published anomaly for decades, with a tidy story (tax-loss selling in December, rebound in January). Then it got famous. As money chased it and markets changed, the effect faded to a ghost of its former self. Or take the Super Bowl indicator: for years, an NFC win "predicted" a market rally with an impressive hit rate — pure coincidence wearing a story's clothes, and it eventually failed loudly. Both teach the same lesson: without a mechanism (a reason the pattern should exist), a pattern is just a coincidence that hasn't been embarrassed yet.

The usual mistakes

Trusting a beautiful equity curve (the chart of a strategy's cumulative profits) that never deducted costs. Optimizing parameters until the past looks perfect — with enough dials, you can fit any squiggle, including pure noise (this is overfitting: memorizing the past instead of learning the future). Confusing a good story with a tested edge. And paying for "proven systems," where the only proven edge belongs to the seller.

What to do

Interrogate every pattern with five questions: tested on data the rule never saw? After all costs? How many variations were tried before this one worked? Is there a mechanism, or just a squiggle? And would I still believe it if the chart were ugly? Then paper-trade it forward for six months before sizing a dollar. The skill from the short read stands: demanding proof is the edge. The deep-dive version is just the checklist for what proof actually looks like.

The fuller mechanism

The VIX isn't a survey of anyone's feelings. It's a formula run over the prices of S&P 500 index options — specifically, a wide strip of out-of-the-money puts and calls (options that only pay off on big moves) — that backs out the market's expectation for volatility over the next 30 days, annualized. When traders bid up options, the formula reads higher expected swings; when options are cheap, it reads calm. Crucially, it measures implied volatility (what traders expect) not realized volatility (what actually happens) — and implied runs hotter than realized most of the time. That gap has a name, the variance risk premium: option sellers demand extra pay for insuring against the unknown, so protection is systematically overpriced.

Which is exactly why buying protection after a spike — the short read's warning — is the worst deal on the menu.

One more mechanical trap: you cannot buy the VIX. The index is a calculation, not an asset. The ETFs that track "volatility" actually hold VIX futures, and futures usually cost more than the spot index — a condition called contango. Rolling from expiring futures into pricier new ones bleeds value every single month, in calm markets most of all.

Where it breaks

The VIX measures expected wobble, not direction and not tail risk — it can read a placid 13 while the cliff edge approaches, because nobody's buying protection yet. It's also S&P 500-only: your small-caps, your crypto, your single stock can be in turmoil while the gauge sips tea. And it has a volatility of its own — the gauge can double in a day, which means hedging with VIX products is itself a volatile trade, the opposite of what a hedge should be.

A real episode

February 5, 2018 — "Volmageddon." After more than a year of record-low volatility, the VIX more than doubled in a single session, closing near 37. That should have been a good day for volatility products. Instead, XIV — a popular note designed to short volatility — lost the overwhelming majority of its value after hours as its mechanics broke under the spike, and the product was liquidated. In March 2020 the gauge itself hit 82.69, its highest close ever, as the pandemic crash repriced everything at once. Both episodes teach the pair of lessons: volatility arrives as a jump, not a trend — and products built on calm markets are the first casualties of the storm they were sold as shelter from.

The usual mistakes

Buying VXX or UVXY as a long-term hedge — the contango bleed means the long-run chart slopes relentlessly down; these are day-trading tools wearing investment costumes. Treating a low VIX as an all-clear (complacency is what low VIX measures — it's a sentiment reading, not a safety certificate). And the mirror error: panic-buying protection at VIX 40, paying the maximum variance premium for insurance against the crash that already happened.

What to do

Read the VIX the way the short read teaches: as a sentiment dial and a position-sizing input. Let the gauge set the posture — calm means protection is cheap, crisis means survival mode: raise cash, cut size. Hedge with the only tools that don't decay: smaller positions, diversification, and cash. If you ever buy tail protection, do it when the gauge is quiet and the premium is thin — insurance is for buying before the fire, and the VIX will tell you, every day, exactly how cheap the fire insurance is right now.

The fuller mechanism

In 1961, Modigliani and Miller proved the dividend irrelevance theorem: in a frictionless market, dividends don't create value — a $2 payout simply moves $2 from the share price into your pocket, and your total wealth doesn't twitch. The short read's ex-date drop is this theorem in work clothes. Buybacks are the mirror image: instead of mailing you cash, the company buys its own shares and retires them, shrinking the share count so each remaining share owns a bigger slice of the same company. Same pizza logic as the short read, running in reverse — fewer slices, same pizza, each slice worth more.

Why bother splitting at all, then? Real frictions: lower share prices make options contracts and employee stock grants more granular, keep the price in a range where retail investors feel comfortable, and generate a drumbeat of attention. None of it changes value — but markets run on humans, and humans have feelings about round numbers.

Where it breaks

Frictions are real, and they're where the theorem meets the tax code. Dividends are taxed when paid — qualified dividends (most U.S. large-cap payouts, held long enough) at lower capital-gains rates, ordinary dividends as income — while buybacks defer your tax until you sell. Reinvesting a dividend costs you a spread each time; a buyback does it for everyone with no per-shareholder transaction. Then there's signaling: a raised dividend is management publicly committing to cash it must keep producing (cutting it later is punished brutally), while buybacks can be quietly paused — which is why buybacks sometimes signal confidence and sometimes just dress up earnings-per-share by shrinking the share count.

A real episode

Apple is the buyback era's monument: a 7-for-1 split in 2014, a 4-for-1 in 2020 — and, more importantly, hundreds of billions of dollars of its own shares repurchased and retired, shrinking the split-adjusted share count by about 40% between 2013 and 2024. Each remaining share now owns a meaningfully larger piece of Apple, with no dividend tax drag on that portion. Contrast the reverse split: when a struggling company does a 1-for-10 reverse split to stay above a dollar and avoid delisting, the slice count shrinks but the pizza is still shrinking — the market reads it, correctly, as a distress signal. Same arithmetic, opposite message.

The usual mistakes

"Dividend stocks are safer" — they fall on the ex-date like everything else, and a high yield often means a crashing price (yield is dividend ÷ price; when the denominator collapses, the yield "looks" generous). Chasing yield into companies that can't sustain the payout — the dividend gets cut and the price falls twice. Buying a stock "for the split," as if the slice count were the pizza. And shorting through an ex-date without pricing in the payment owed (the short read's trap).

What to do

Judge total return — price change plus dividends — never yield alone. Check sustainability: the payout ratio (dividends ÷ earnings) tells you whether the dividend survives a bad year; anything persistently over 100% is eating the furniture. Choose buybacks vs dividends by your tax situation, not by vibes. And when a company announces a split, enjoy the cheaper options and move on — the only number that changed is the one on the slice count.

The fuller mechanism

Compounding has a pocket calculator built in: the Rule of 72. Divide 72 by your annual rate and you get roughly the years to double — 72 ÷ 7 ≈ 10.3 years. Four doublings in forty years turns each dollar into about fifteen, which is why the curve looks boring for a decade and then bends like a hockey stick: most of the growth happens in the final doublings, when the returns are earning returns on returns.

Now the adjustment the short read only hints at: inflation compounds too. At 2.5% a year for 40 years, prices multiply by about 2.69 — so the short read's $262,481 arrives with the purchasing power of roughly $97,700 today. Still slightly more than double your $48,000 of contributions in real terms (approximately — those contributions were nominal dollars) — compounding beats inflation comfortably — but the quarter-million headline is nominal (face-value dollars), and nominal is the currency of headlines, not groceries.Computed.

Where it breaks

Three cracks. First, the 7% is a long-run average wearing a costume — actual decades vary wildly, and retiring into a bad decade hurts far more than the average suggests — that's sequence-of-returns risk (crashes early in retirement, when you're selling): the average hides it because selling during those crashes locks in the losses. Second, inflation regimes: the math above assumed a calm 2.5%; the 1970s investor compounded nominal gains while inflation ate them faster. Third, taxes on the way out — the short read's numbers are pre-tax, and the government's share compounds against you just as faithfully.

The cost of waiting

The cost of waiting, in one comparison. Start at 25: $100 a month for 40 years at 7% → about $262,481 (the short read's number). Start at 35 — just ten years later, same $100 — for 30 years: about $122,000. That decade of delay cost roughly $140,000 — nearly four times the late starter's entire $36,000 of contributions. The early starter's first ten years of contributions ($12,000!) did most of the damage, because those dollars compounded the longest. "I'll invest when I earn more" is the most expensive sentence in personal finance: the dollars you invest youngest work hardest, and no later salary fully replaces them.Computed.

The usual mistakes

Waiting for "the right time" — the market's best days cluster near its worst days, and cash waiting for clarity usually just waits. Keeping long-term money in savings "until things settle" (the hurdle from the short read: every idle dollar is actively choosing ~4%). Checking the balance daily, which converts the volatility — the price of admission — into suffering. And lifestyle creep spending the raise before the automatic investment gets it: compounding needs the contributions to actually happen.

What to do

Automate a monthly investment on payday — before you see the money, not after you've spent it. Raise the amount with every raise (half the raise to investing is painless and bends the curve). Keep the hurdle visible: label your cash account "choosing 4%" if that's what it takes. Never interrupt compounding to time the market — the bend at the end only happens if you're still in the seat. And measure in real terms occasionally, so the nominal headlines don't fool you into thinking you're richer than your groceries say.

The fuller mechanism

The short read named the disposition effect; here are its four accomplices. Overconfidence: most investors rate themselves above average — which can't be true of most of us — and the classic finding (Barber & Odean) is that the more people trade, the worse they do: activity feels like skill.

Loss aversion: losses sting roughly twice as much as equal gains feel good (Tversky & Kahneman, 1992) — which is why investors sell winners to "feel" a win and nurse losers to avoid feeling the loss.

Anchoring: your purchase price becomes a mental reference point, so a stock "down from $100" feels cheap at $70 regardless of what the business is worth.

Recency: projecting the last year forward forever — buying high after rallies, swearing off stocks after crashes.

The herd: social proof hijacks the whole apparatus, because the pain of missing out feels like any other loss.

Where it breaks

Here's the uncomfortable part: knowing the bias doesn't remove it. Professional fund managers, shown the same research, exhibit the same patterns — expertise moves the biases around more than it erases them. "Debiasing" training has weak, fading effects; the heuristics exist because they were useful shortcuts for most of human history, and the market is precisely the environment where shortcuts get punished. Kahneman's framing stands: fast, intuitive thinking (System 1) runs the show, and slow, deliberate thinking (System 2) is lazy, easily tired, and often just rationalizes what System 1 already decided. You don't fix the hardware. You build guardrails around it.

A real episode

Take the dot-com peak: in 1999, recency bias said tech only goes up, overconfidence said you could pick the winners, the herd said everyone was getting rich, and anchoring kept buyers holding "discounted" stocks as the index fell 78% from its March 2000 peak (see Idea 02's deep dive — that needed roughly a 350% recovery). Or January 2021's meme-stock mania: GameStop's vertical chart was pure social-proof fuel — FOMO (fear of missing out) drove buying at prices no earnings could ever justify, and loss aversion then froze the late buyers in place for the long slide down. Different decades, identical wiring.

The usual mistakes

"I know about biases, so I'm immune" — the bias blind spot (spotting flaws in others, never yourself) is itself a bias. Checking prices constantly: myopic loss aversion means the more often you look, the more losses you feel, and felt losses drive exactly the panic selling the short read warns about. Confusing a bull market with brains — everyone's a genius when everything rises, and the bill arrives when it doesn't. And journaling your brilliance but never your errors, which turns the record into a highlight reel instead of a diagnostic.

What to do

Build systems, not willpower. A written investment policy: maximum loss per trade decided before entry, position caps (Idea 05's 25%), and the conditions under which you'll sell — written when you're calm, followed when you're not. A decision journal: one paragraph per buy — the thesis, what would prove you wrong — reviewed yearly; it's the only honest mirror.

Cooling-off rules: no selling within 48 hours of scary headlines. Automatic investing for the core portfolio, so System 1 never gets a vote. And the short read's master rule, repeated because it's the whole game: decide the maximum loss before the trade, not after — after, you're negotiating with yourself, and yourself is a biased counterparty.