Win Rate vs Risk/Reward: 36 Tests, and Win Rate Predicted Nothing

Written by chwoo · Crypto trader, 7 years. Running an automated futures bot since spring 2026. · 2026-07-07 · updated 2026-09-08

"My strategy wins 80% of the time" sounds great, right up until you learn it risks $10 to make $1 and one bad trade erases a month of wins. Win rate is the number beginners fixate on and the number that means the least on its own.

Rather than argue that, we built a grid. One entry signal, RSI crossing above 30 on BTC hourly candles, run 36 times with nothing changed but the take profit and stop loss. That is enough to move the win rate from 7.5% to 87.3% without touching the entry rule once, and it lets us ask the only question that matters: which of those 36 made money?

The trade-off is mechanical, not a matter of skill

First hold the stop at 2% and widen only the take profit. Realistic costs throughout (0.06% fee plus 0.02% slippage per side), 1x leverage, 90% position size, data through 31 August 2026.

Open the strategy this guide measures in the backtester →

Nobody got better or worse at trading between those rows. Widening the target means fewer trades reach it, so the win rate falls; the ones that do reach it are bigger, so the payoff ratio rises. Win rate is a setting, not an achievement. Anyone can hand you a high one.

A 73% win rate that loses 46%

Now hold the take profit at 3% and widen only the stop.

This is the cheapest trick in trading. Move the stop further away and the win rate goes up, because price has more room to come back before the position is closed. The bottom row wins nearly three trades in four. It also loses 46.2% of the account, because each of those rare losses is 8.13% while each of the frequent wins is 2.86%.

All 36 cells at once

Here is the full grid. Each cell is the win rate and the total return for that take profit and stop combination.

Read the top right corner first. 87.3% win rate, -75.2% return. That is the highest win rate in the entire grid and one of the worst results in it. Now read the bottom right: 44.5% win rate, +60.7% return, the best cell here.

Count the signs and the conclusion is blunt. Thirteen of the 36 cells win more than half their trades, and not one of those thirteen makes money. Only two cells in the grid are profitable, and both win well under half the time.

Two more things about that corner, so it is not read as a discovery. Those two cells trade 282 and 191 times against 855 at the tightest stop, so part of what looks like an edge is simply having paid the round trip fewer times. And their profit factors are 1.01 and 1.05, with maximum drawdowns of 79.3% and 75.0%. They are not usable strategies. They are the two least-bad cells in a grid where the entry signal has no edge.

The break-even line, and what costs do to it

A strategy breaks even, ignoring costs and compounding, when W = 1 / (1 + R). At R = 1 you need 50%; at R = 2 you need 33%; at R = 0.5 you need 67%. That formula is worth knowing, and it is worth knowing that costs move the line it draws.

The pattern in the last column matters. Costs lift the required win rate by 4.7 points for the 1% target and only 1.4 points for the 8% target. A round trip of roughly 0.16% is a large bite out of a 1% win and a small one out of an 8% win.

So the high-win-rate style is punished twice. It needs a high win rate by construction, and then costs push that requirement higher still. Scalping is not merely harder psychologically; the arithmetic asks more of it.

Where the formula stops being enough

Being honest about our own tool: the break-even formula is arithmetic, and an account is not. It compares average win with average loss, and ignores the order trades arrive in and the compounding of position size. Two of this site's presets show the gap.

The Donchian breakout sits at 40.0% against a required 39.9%, one tenth of a point above the line, and returns -41.5%. The formula called it a coin flip; the account lost 41.5%. Being marginally above the break-even line is not the same as making money, because a run of losses shrinks the position that has to earn it back. The trend-filtered RSI tells the same story: 0.6 points above the line, a profit factor of exactly 1.00, and a return of -0.2%.

Now read the whole column. The highest win rate in the table does belong to a profitable strategy, so win rate is not useless. But the second highest, at 49.8%, returns -59.1%, and the best return in the table comes from the strategy with the second lowest win rate. Ranking by win rate does not rank by outcome.

The column that holds up is profit factor. Across the 36 grid cells, a win rate above 50% predicted a profit zero times out of thirteen. A profit factor above 1 occurred in two cells, and both were profitable. That is a small sample, but the direction is not subtle.

What to do with this

• Never read a win rate without the average loss. Ask for the payoff ratio in the same breath, or the number is unreadable.
• Check W against 1 / (1 + R). Below that line the strategy cannot work at any position size. Just above it is not yet a reason for confidence.
• Prefer profit factor and total return. They already contain both numbers, and in this grid they were the only ones that tracked the outcome.
• Suspect a widened stop. If a win rate is high and the average loss is several times the average win, the win rate was manufactured, not earned.
• Charge costs before comparing. Costs raise the bar most for the tight targets that produce flattering win rates.