In the introduction, two traders finished a year apart in fortune even though one of them was wrong four times as often as the other. That wasn’t a paradox and it wasn’t luck. It was arithmetic — a single line of it — and this lesson is where we write that line down and stare at it until it changes how you look at every bet you will ever take.
The line is this: the value of a bet is not how likely it is to pay, and it is not how big the payoff is. It’s the two multiplied together. Miss either half and you will confidently walk into the wrong bets — grabbing the ones that win often and pay nothing, dodging the ones that lose often and pay everything. The whole model of asymmetry lives inside that multiplication, so let’s earn it slowly, with real numbers, and then hand you a machine to drive.
Before you read — take a guess
Bet A pays +$1 with 90% probability (and nothing otherwise). Bet B pays +$9 with 10% probability (and nothing otherwise). Which bet is worth more to take, on average?
Probability times magnitude
Here is the entire engine of the course, and it fits on one line. The expected value of a bet — its EV — is the payoff you’d average out to if you could somehow play it thousands of times. You get it by weighting each outcome by how likely it is and adding up:
In words: the probability of winning, times the gain you get when you win, minus the probability of losing times the loss you eat when you lose. Every term matters, but notice the two things being multiplied on each side: an odds (a probability) and a magnitude (a dollar size). Neither one is the answer by itself. The odds tell you how often; the magnitude tells you how much; only their product tells you how good.
Start with the cleanest possible demonstration — two bets with the same expected value but opposite shapes. Bet A wins a small amount very often; Bet B wins a large amount rarely.
| Bet | Win prob | Gain | EV = | Feels like… |
|---|---|---|---|---|
| A — the frequent winner | 90% | $1 | 0.90 × $1 = $0.90 | Winning almost every time |
| B — the rare big win | 10% | $9 | 0.10 × $9 = $0.90 | Losing almost every time |
Read that table twice, because it is quietly outrageous. Bet A wins nine times as often as Bet B. If you judged them by win-rate — as almost everyone instinctively does — A looks like a landslide favorite. But their expected values are identical: ninety cents apiece. The extra frequency of A is exactly cancelled by the extra magnitude of B. Win-rate saw a blowout; the math saw a tie. That gap between what win-rate feels like and what the EV actually is — that’s the misconception this whole course exists to burn out of you: win-rate is not expected value.
When magnitude swamps probability
The tie above was a warm-up. Now watch what happens when the magnitudes stop being polite and get lopsided — when the payoff on one side dwarfs the payoff on the other. This is where asymmetry earns its name.
Take a bet that loses almost all the time. It wins only 1 time in 100 — a 1% shot — but when it hits, it pays $500, and when it misses (the other 99% of the time), it costs you just $1:
| Outcome | Probability | Payoff | Contribution to EV |
|---|---|---|---|
| Win | 1% | +$500 | 0.01 × (+$500) = +$5.00 |
| Lose | 99% | −$1 | 0.99 × (−$1) = −$0.99 |
| Total EV | +$4.01 per play |
A bet you lose ninety-nine times out of a hundred is worth four dollars every time you play it. Look at where the value came from: the win side contributed +$5, the loss side subtracted less than a dollar. The 99% probability of losing barely moved the needle, because the thing it multiplies — the $1 loss — is tiny. Meanwhile the 1% probability of winning dominated the whole calculation, because the thing it multiplies — the $500 gain — is enormous. When one magnitude is hundreds of times the other, the magnitude term runs the show and the probability term is almost a rounding error. The odds didn’t decide this bet. The shape did.
The one-sentence version
A bet’s value is probability × magnitude, summed over its outcomes — and when the magnitudes on the two sides are wildly lopsided, the big magnitude dominates the sum, so a bet can lose almost every single time and still be enormously worth taking. The odds alone never decide a bet.
The pitfall: reading only one column
The trap is seductive because it’s half right. People fixate on one column of the table — usually the probability — and treat it as the answer. “It loses 99% of the time, so it’s a terrible bet.” “It wins 90% of the time, so it’s a great one.” Both statements throw away the magnitude column entirely, and the magnitude column is where lopsided bets hide all their value (or all their danger). You cannot judge a bet from its win-rate any more than you can judge a business from its number of sales without knowing the price of each. Frequency is one factor. Size is the other. Only the product is the verdict.
When to use it
Reach for probability-times-magnitude the instant you catch yourself — or a headline, or a colleague — ranking options by how often they work. A trading strategy quoted by its win-rate, a startup pitch that leads with “9 out of 10 customers loved it,” a medical choice described only by its success percentage: each is showing you one column and hiding the other. The fix is a reflex — ask “and how big is the win versus the loss?” — and multiply before you decide.
The convex smile
Now we give the two shapes from the introduction their precise arithmetic. First the one you want to own: the convex payoff, the smile.
A convex payoff has a capped downside and an open-ended upside. Your loss is small, known, and bounded — a fixed cost you can absorb and shrug off. Your gain is large and, in the extreme, unbounded — there’s no ceiling on how good the good case can be. In one line: lose a little often, win big rarely. Planting ten seeds knowing nine will die, buying a cheap lottery-like stake in an early venture, running a $50 experiment that might discover a $50,000 idea — all smiles.
The cruelty and the beauty of the smile is that it feels awful while it’s working. You lose, and lose, and lose, because most trials are the capped small loss. The whole payoff sits in the rare win — and the arithmetic pays off handsomely even though the calendar is a graveyard of small defeats. Watch it in full:
| Trial outcome | Probability | Payoff | Contribution to EV |
|---|---|---|---|
| Win (the rare big one) | 20% | +$12 | 0.20 × (+$12) = +$2.40 |
| Lose (the frequent small one) | 80% | −$1 | 0.80 × (−$1) = −$0.80 |
| EV per trial | +$1.60 |
You lose four times out of every five. A win-rate scout would call this a disaster and walk away. But each trial is worth +$1.60 on average, because the $12 you win on the rare good trial is twelve times the $1 you lose on each bad one. Run it a hundred times and you’d expect to be up around $160 — having lost roughly 80 of the 100 trials to get there. That is the convex bargain in a single table: a long losing streak that is, in expectation, a money machine. The losing streak isn’t a bug in the strategy. It is the strategy.
Why 'convex'?
Plot how your payoff grows as things go your way and a convex curve bends upward — it accelerates, curving like the bottom of a smile. The capped loss is the flat left edge; the open-ended gain is the rising right edge that never stops climbing. We’ll make that curve literal and drivable in a moment.
The concave frown
Now the mirror image — the shape to fear, because it disguises itself as success. The concave payoff, the frown, has a capped upside and an open-ended downside. Your gain is small, steady, and reliable; your loss is rare but potentially catastrophic and unbounded. In one line: win a little often, lose big rarely.
This is the exact reverse of the smile, and it feels the exact reverse too — which is what makes it lethal. A concave strategy hands you a long, comfortable winning streak while it quietly loads the one loss that erases all of it. Nassim Taleb’s image for it is unforgettable: picking up pennies in front of a steamroller. Each penny is a real, pleasant little win. The steamroller is the rare day it isn’t.
Put numbers on it. Here is a bet that wins 90% of the time — a hit-rate that would make any pundit swoon — and yet is a guaranteed loser:
| Outcome | Probability | Payoff | Contribution to EV |
|---|---|---|---|
| Win (the frequent penny) | 90% | +$1 | 0.90 × (+$1) = +$0.90 |
| Lose (the rare steamroller) | 10% | −$20 | 0.10 × (−$20) = −$2.00 |
| EV per trial | −$1.10 |
Ninety percent wins, and it bleeds — −$1.10 every single time you play it. The steady stream of $1 wins adds up to +$0.90 of expected value, and the rare $20 loss single-handedly drags the whole thing to −$2.00. Play it a hundred times and you’d expect to be down about $110, even though roughly 90 of your 100 trials were winners. The win-rate was a magnificent 90% and the strategy was a slow-motion disaster. Same trap as the introduction’s Trader A: a gorgeous hit-rate sitting on top of a hidden, open-ended loss.
A high win-rate is not a safety certificate
The concave frown produces exactly the track record that fools people: months of smooth, reliable little wins with almost no losing days. That streak is not evidence the bet is safe — it can be the signature of a hidden steamroller that simply hasn’t arrived yet. When something wins almost every time, don’t relax. Ask how big the rare loss is. That’s the column the win-rate is hiding.
Fill in the two shapes:
Pick the right option for each blank, then check.
A convex (smile) payoff has a downside and an open-ended upside — you lose a little and win big rarely. A concave (frown) payoff is its mirror: a capped upside and an downside — you win a little often and lose big rarely. The shape, not the , decides where you finish.
Drive it yourself
Reading the tables is one thing; feeling the shape decide the outcome is another. Below is the PayoffExplorer — a live machine where you set the shape of a bet (the size of the win, the size of the loss, and the probability of winning) and then run it for a hundred trials to watch the total roll in. It has two one-click presets, Convex and Concave, that snap the sliders to a smile and a frown so you can see each shape’s signature at a glance.
Try this, in order:
- Snap to the Convex preset and run 100 trials. Watch the trial-by-trial log: you’ll lose most individual trials — red result after red result — and yet the running total climbs. That’s the smile paying off in slow motion: a graveyard of small losses funding a few big wins.
- Now snap to the Concave preset and run 100 trials. Reverse everything. The log fills with green wins, trial after trial — and the running total bleeds downward anyway. A 90%-ish win-rate marching you straight to a loss. That is the frown, and that is what it feels like to pick up pennies right up until the steamroller.
- Then hand-tune the sliders yourself. Start from a bet with a negative EV and try to flip it positive without touching the win-probability at all — just widen the gap between the win size and the loss size until the shape carries the day. Feel how magnitude, not frequency, is the lever that moves the sign.
Shape the payoff
Set the shape, then run a hundred trials
Set how often you win, how big a win pays, and how much a loss can take — or snap to a preset. Then run 100 trials and watch the total. A convex bet can lose most of its trials and still win big.
Lose a little, often; win a lot, rarely. Downside capped, upside open.
Press “Run 100 trials” to sample this bet and watch the total build up.
- Expected value / trial
- +1.20
- Trials run
- 0
- Actual win rate
- —
- Total profit / loss
- —
The single most important thing to notice while you play: the running total and the win-rate keep disagreeing with each other. On the smile you’re losing trials and gaining money; on the frown you’re winning trials and losing money. Every time your gut says “but it keeps winning!” the total is there to remind you that winning trials and winning money are two different scoreboards — and only one of them pays the bills.
Recap
Check yourself: the shape of the payoff
What is the expected value (EV) of a bet, in words?
Check your answer to continue.
Next up
You now own the line the whole course turns on — value is probability times magnitude — and you’ve watched, in a live machine, a bet lose most of its trials and still climb, and win most of its trials and still bleed. That’s the shape doing the deciding. In the next lesson, Losing Most Bets, Winning the War, we take the convex smile out of the abstract and run it through a full year of trades: lose small on 80% of tries, win big on the other 20%, and finish far ahead — then set it against its mirror, the concave “pennies in front of a steamroller,” and see why fat tails make the steamroller not just possible but inevitable. The arithmetic you just learned is about to become a survival strategy.