Why Do I Keep Blowing Up My Account?
Blow-ups are a sizing failure far more often than a selection failure. Recovery is asymmetric — a 50% drawdown needs a 100% gain to undo — so one or two outsized positions can destroy an account whose method was fine, and a losing streak long enough to do it is an ordinary event.
Why is recovery so much harder than the loss?
Because a percentage down and a percentage up are computed against different bases. Recovering
from a drawdown d requires a gain of 1/(1-d) - 1:
| Drawdown | Gain required to get back to flat |
|---|---|
| 10% | +11.11% |
| 25% | +33.33% |
| 50% | +100.00% |
| 75% | +300.00% |
| 90% | +900.00% |
The curve is gentle until it is not. At 10% down you need an ordinary month; at 75% down you need to quadruple an account that is now a quarter of the size, using the same method that produced the hole. That is not usually recoverable, and the trader who reaches it did not decide to accept a 75% drawdown — they accepted several 25% ones in sequence.
And the drag is structural rather than personal: measured across a whole market, individual traders gave up 3.8 percentage points a year, almost all of it through aggressive orders.[1]
Is it really the sizing and not the picks?
Test it against a streak. At a 55% win rate, 45 of every 100 trades lose, and seven losses in a row is close enough to even money across 200 trades that you should plan on meeting one.
Now price that streak at two sizes. Seven losses at 2% risk is a 13.19% drawdown, needing +15.19% to repair. The same seven at 20% is a 79.03% drawdown, needing +376.84%. Same trader, same seven bad calls, and only one of them still has an account.
What does one oversized trade actually cost?
More than the arithmetic suggests, because it usually arrives in pairs. Risk 25% on a conviction trade, lose, and the natural response is to size the next one to make it back. Two 25% losses in a row are a 43.75% drawdown, and getting back to flat from there takes +77.78%.
Put differently: a single 20% risk deploys the equivalent of ten trades' worth of a 2% plan on one outcome. Whatever your edge is, it was measured across many trades and cannot express itself in one. Concentrating the stake removes the only mechanism by which an edge becomes money.
Why does it keep happening after you know all this?
Because the decision to size up is not made in the same state as the decision to have a rule. It is made mid-drawdown, with a specific trade in front of you that looks unusually good — and it looks unusually good partly because you need it to. No amount of agreeing with the arithmetic in advance binds the person who is about to override it.
What does bind is a constraint that exists outside the moment. A size stated before the entry and recorded where you cannot quietly amend it is a different object from a size you intended: the override becomes visible rather than silent, and a visible override is one you have to actually decide to make.
What is the fix, concretely?
- Derive size, never choose it.
size = risk budget / distance to stop. A $200 risk with a $2.00 stop is 100 shares. Conviction changes which trades you take, not the size of the ones you take. - Cap the ratio. Largest risk over median risk should sit near 1. Above 3, one decision outweighs your edge.
- Size for the streak you will meet, not the one you expect. Seven losers in a row is an ordinary event across a year of trading, so it is a planning input rather than bad luck.
- Record the intended size before the fact. A plan that is only in your head is a plan you will remember having followed.
The escalation after a loss is a documented effect rather than a character flaw. Thaler and Johnson found people accept gambles they would otherwise refuse when the gamble offers a route back to breaking even[2] — which is precisely the trade that follows a 25% loss. Kelly's 1956 result is the other half: growth is maximised at a finite fraction of capital, and betting above it lowers long-run growth even when the edge is unchanged.[3]
kappi is a trade recorder: you commit a trade before the fact, it is sealed on your device for a time-capsuled delay you choose, then kappi publishes it on a Merkle-anchored log. The seal is the whole mechanism — once a stop, a target and a thesis are sealed, revising them is no longer a private edit. $15/month, no free tier.
Sources
- Barber, Lee, Liu & Odean, 'Just How Much Do Individual Investors Lose by Trading?', Review of Financial Studies 22(2), 2009, 609–632 read 2026-08-16
- Thaler & Johnson, 'Gambling with the House Money and Trying to Break Even: The Effects of Prior Outcomes on Risky Choice', Management Science 36(6), 1990, 643–660 read 2026-08-16
- Kelly, 'A New Interpretation of Information Rate', Bell System Technical Journal 35(4), 1956, 917–926 read 2026-08-16
Frequently asked questions
Why is a 50% drawdown so much worse than a 25% one?
Because recovery is 1/(1-d) - 1. A 25% drawdown needs +33.33% to get back to flat; a 50% one needs +100%, and a 75% one needs +300%. The requirement accelerates while the account doing the work shrinks.
Is blowing up caused by bad analysis?
Rarely. Seven consecutive losses at a 55% win rate is an ordinary event — 0.45^7 = 0.3737% from any starting trade, which is 0.725 expected occurrences over 200 trades. At 2% risk that streak costs 13.19%; at 20% risk it costs 79.03%. The analysis was the same.
How much does one oversized trade cost?
Two 25% losses in a row leave 0.75^2 = 0.5625 of the account — a 43.75% drawdown needing +77.78% to recover. A single 20% risk also puts ten trades' worth of a 2% plan on one outcome, which is the opposite of how an edge turns into money.
How do I actually stop sizing up?
Derive size from the stop rather than from conviction: size = risk budget / distance to stop. Then record the intended size before the fact, so an override is a visible act rather than a private one.