Down this much
−50%
One bad, oversized stretch.
Position sizing, or how not to blow up.
Every other lesson is about being right. This one is about the thing that keeps you in the game long enough for being right to matter. You can win most of your trades and still go to zero. The gap between those two facts is called size, and the math of it is not intuition, it's arithmetic. Here's the arithmetic.
not advice · not a formula to follow blindly · size is personal
Why this lesson exists
Nobody blows up because they read one trade wrong. They blow up because they had too much on when they read it wrong. A great edge and a reckless size is still a blown account, and it happens faster than a bad edge and a careful size ever could. Sizing is the only part of trading that is fully in your control. The market decides the outcome. You decide how much rides on it.
A loss hurts more than the same-size gain helps.
Lose 10% and you need more than 10% to get back, because the 10% you have to make now comes off a smaller base. The formula is exact and it gets ugly fast:
gain to recover = drawdown / ( 1 − drawdown ) → −50% needs +100%, −90% needs +900%
| You are down | You need to make | What that really means |
|---|---|---|
| −10% | +11.1% | Annoying. A good month erases it. |
| −20% | +25.0% | Now the recovery is visibly bigger than the loss. |
| −33% | +50.0% | A third gone costs you a half to fix. |
| −50% | +100.0% | You have to double what's left just to break even. |
| −70% | +233.3% | Most accounts never come back from here. |
| −90% | +900.0% | You need a 10-bagger to undo one bad run. |
Gain required to recover a drawdown
Takeaway: the goal of sizing isn't to make more when you're right. It's to make sure a bad run leaves you at −15% instead of −60%, because −15% is a bad month and −60% is a new career.
The same edge, four position sizes.
Here is the experiment that makes sizing concrete. Take a genuinely good game: you win 55% of the time, and a win pays the same as a loss costs (1:1). That is a real, enviable edge. Most traders never have it. Now run it four ways, changing only one thing: what fraction of your account you put on each trade.
| Risk per trade | Chance of blowing up | Median ending account | Read |
|---|---|---|---|
| 1% | 0.0% | $16,080 | Bulletproof, but slow. Barely uses the edge. |
| 5% | 0.5% | $65,293 | The sweet spot. Grows hard, almost never dies. |
| 10% | 15.6% | $122,336 | Bigger median, but 1 in 6 is dead. |
| 20% | 70.8% | $1,912 | Same 55% edge. Most players go broke anyway. |
Read the last row again. Nothing about the edge changed. The win rate is still 55%. The only difference is that the 20% bettor is putting so much on each trade that a normal losing streak, the kind that is guaranteed to happen, takes them past the point of no return. The edge was real and it still went broke.
Median account over 500 trades · 10,000 simulated paths each
risk 1% · slow grind
risk 5% · compounds
risk 10% · fastest median
risk 20% · goes nowhere
The survivorship trap
Watch how the 20% bettor lies to you. Across 10,000 simulated runs, the average ending account was $436,000, which sounds incredible. But the median was below where they started, and 53% ended down. The average is a fantasy dragged up by a handful of lottery-winner paths (the top 1% ended above $6 million). The typical experience was slow bleeding. When someone shows you a monster return from big sizing, you're looking at a survivor. The graveyard doesn't post screenshots.
There's a mathematically best size. You should bet less than it.
In 1956 a Bell Labs physicist named John Kelly worked out the exact fraction of your bankroll that maximizes long-run growth for a given edge. Ed Thorp took it from a paper to the blackjack tables and then to a hedge fund that compounded for decades. The formula, for a bet that pays b-to-1 with win probability p:
f* = ( b·p − q ) / b where q=1−p, b=payoff ratio → 55% at 1:1 = 10% of bankroll
So for our 55% game, full Kelly says bet 10% per trade. Look back at the ruin table: 10% had the biggest median account and a 1-in-6 chance of blowing up. That is Kelly working exactly as advertised. It maximizes growth, and it is a white-knuckle ride that most humans cannot actually stomach.
Long-run growth rate vs fraction bet · 55% edge, 1:1
This is why almost nobody who uses Kelly bets full Kelly. The standard move is half Kelly or quarter Kelly. You give up a little growth and you cut the drawdowns roughly in half. The curve near the top is nearly flat, so shading down costs you almost nothing in return and buys you a lot of sleep. And that assumes you even know your true edge, which you never do, which is the next problem.
Kelly calculator
Your edge per trade (expected value)+10.0%
Full Kelly says bet10.0%
Your chosen size5.0%
Half Kelly on a 55%, 1:1 edge is 5% per trade. Sane and fast.
Payoff ratio is your average winner divided by your average loser, not your target. A tight-stop day trader might be 40% win rate at 2:1. A premium seller is often 80%+ at a much worse than 1:1 ratio, which is exactly why sellers have to watch size. If the edge comes out negative, there is no size that makes it work.
A 55% edge still hands you brutal losing streaks.
People size as if a 55% win rate means "win, lose, win, win, lose" in a tidy rhythm. It doesn't. Randomness clumps. Inside a real edge live streaks long enough to make you quit, revenge-trade, or double up at the worst possible time. The math is not a matter of opinion.
| In 100 trades at 55% wins | Chance it happens at least once | |
|---|---|---|
| A 4-loss streak | 91.5% | Basically guaranteed. |
| A 5-loss streak | 64.7% | More likely than not. |
| A 6-loss streak | 36.3% | Better than 1 in 3. |
| A 7-loss streak | 17.9% | 1 in 6. |
| An 8-loss streak | 8.4% | Rare, not impossible. |
A six-in-a-row losing streak feels like the strategy broke. It didn't. At a 55% win rate it shows up in more than a third of any 100-trade stretch. Even a strong 60% edge coughs up a 6-loss streak about 21% of the time. Your size has to assume the streak is coming, because over a real career it absolutely is.
Small N lies (the recurring theme)
This is the same warning every page on this site keeps making. A trader up big over 20 trades has told you almost nothing. Variance this large means a losing strategy can look brilliant for months and a winning one can look broken for weeks. Size for the edge you can prove over hundreds of trades, not the streak you're on. The math above is why the honest answer to "is it working?" is almost always "too early to say."
Takeaway: if a 6-loss streak at your position size would do real damage to the account or your decision-making, you are already too big. Size so the guaranteed bad streak is survivable and boring.
The whole lesson, compressed into one habit.
You do not need to run simulations before every trade. All of the math above collapses into a single discipline that pros have used forever: decide the most you'll lose on any one trade as a fixed percent of the account, and never break it. One percent is the textbook figure for directional traders. Two percent is aggressive. Whatever you pick, it is a rule, not a mood.
position size = ( account × risk% ) / loss per unit if your stop hits
That formula answers "how many" so you never have to guess. Risking 1% of a $10,000 account means $100 at risk. If your stop is $2 away, you buy 50 shares. If it's $0.50 away, you buy 200. The size floats so the risk stays constant. That is the entire trick.
Sizing & drawdown calculator
Most you risk on the next trade$100.00
Account after the losing streak$9,414.80
Drawdown from the streak−5.9%
Gain needed to recover it+6.2%
Survivable. A 6-loss streak barely dents you. This is the point.
Assumes each loss is a full risk-unit off the running balance (fixed-fractional). Real trades vary, but the shape holds: keep the per-trade risk small and even a bad streak stays a bruise, not a wound.
Sellers, this lands differently
If you sell premium, your risk per trade is not a neat stop, it's the max loss on the position, and it can be far bigger than the credit you collected. A short put on 100 shares risks the whole way to zero. Size off the worst case, not the premium. The options course walks through a real −$4,520.34 day that came entirely from being oversized, not from being wrong. See selling premium, start to finish → for the seller-side version of this exact rule.
The famous blowups were sizing, not stupidity.
The people in this table were not dumb. Some were the smartest in the room, with Nobel laureates on staff. They had real edges. What killed them was size and leverage, which is the same disease this whole page is about, scaled up to where it makes the news.
| Who | The number | What actually went wrong |
|---|---|---|
| LTCM 1998 |
−$4.6B | Two Nobel laureates, a genuine edge, and ~25:1 leverage on $1.25 trillion of notional. When the rare move came, the size turned a bad month into insolvency. The Fed had to organize a $3.6B rescue. |
| Barings Bank 1995 |
−£827M | One trader, unhedged and oversized on Nikkei futures, sank a 233-year-old bank. It was later sold for £1. No edge at all, just size with no cap. |
| Archegos 2021 |
≈−$10B | A family office ran ~$36B of concentrated bets on far less real capital through swaps. A few stocks dropped and it was gone in days, taking ~$5.5B of Credit Suisse with it. |
| Typical day trader 2020 study |
97% lose | Of retail day traders who stuck with it past 300 days, 97% lost money; barely 1% beat minimum wage. Same story, small scale: too big, too often, no survival math. |
The thread through all of them
Every one had the same cause of death, whether it was a $4 billion fund or a $4,000 account: a position large enough that one normal-sized bad outcome could not be survived. That is the definition of oversized, and it is the only real way to blow up. Get the size right and the market can be wrong about you for a long time without ending you.
Where this connects
← Back to the education portal
Sizing is downstream of a few things this portal covers elsewhere. Selling premium, start to finish puts these rules on real options positions. Keep what you make handles what taxes do to the gains you protect by not blowing up. The math here doesn't change with the asset. It's the same arithmetic on a memecoin, a covered call, or a futures contract.