A high win rate can hide the tail
A win rate near two out of three looks great in a backtest. Here's a real SPX 0DTE strategy we tested, what its win rate left out, and the numbers worth checking first.
30 trades, nearly two winners out of three
HAB_BOT #9 is an SPX 0DTE strategy we put on probation in September. It trades once a day, and every fill is measured after fees. From Sept 4 to Oct 21 it took 30 trades. Nineteen made money, and it finished up $4,366 per 1-lot.
On a strategy sales page, that's the whole pitch: 63% winners and a profit. Here's what each trade actually looked like.
Most bars are modest. Three are not. On Sept 8, Sept 18 and Sept 30 the strategy lost $878, $1,580 and $1,320. Together those three days cost $3,776.68, more than its 10 smallest wins combined.
A win rate counts how often. It says nothing about how much.
The win rate you need just to break even
When the average loss is bigger than the average win, a strategy has to win more often just to stay even. The break-even win rate is:
So the real cushion wasn't "63% winners." It was 63% against a 52% hurdle: about 11 points, or roughly three trades out of 30 going the other way.
That's thin. Two more days like Sept 18 and nearly three-quarters of the profit is gone.
The running total shows what that felt like. The strategy spent most of September below zero. By its eighth trade, on Sept 18, it was down $1,815. It clawed back to about even, then Sept 30 took it to −$1,779 after 16 trades. It stayed above zero for good only from Oct 2, and most of the final profit came in October, after a month most people would have quit.
Judged at trade 16, it was down $1,779. Judged at trade 30, it was up $4,366. Same strategy, same rules. The only difference was which days had arrived yet.
Where the tail comes from in 0DTE
Many popular 0DTE strategies sell options. They collect a small, capped premium on most days and accept a larger loss on the days the market moves hard. That shape is built in: the win rate comes from the many quiet days, and the result depends on the few loud ones.
Loud days tend to share features: a big gap at the open, a scheduled event like CPI or a Fed decision, or a session where dealer hedging adds to the move instead of absorbing it. They're rare, which is exactly why a short backtest can miss them entirely, or catch only one.
Five things to check in any backtest
- Average win vs. average lossIf losses are bigger, work out the break-even win rate and compare it with the real one.
- Worst day vs. average winHow many normal wins does one bad day erase? Here, about two and a half.
- Largest drop from a highCould you sit through it without changing the plan halfway?
- Sample sizeThirty trades is a start, not proof. Did the test include an event week or a gap day?
- How fills were measuredResults priced at the midpoint look better than real fills after fees.
About the data: HAB_BOT #9 is a test strategy on probation, not a recommendation. 30 measured fills, Sept 4 – Oct 21, 2026, net per 1-lot after fees. Past results don't predict future results.