My win rate fell to 25% and I started making money

Most traders treat win rate as the scoreboard. You check it after a losing week, see 38% and feel like a fraud. You read that professionals win 70% of the time and conclude the gap between you and them is accuracy.

In June I changed how I trade. My win rate dropped from 40% to 25%, and that was the point at which my results turned positive. Not despite the lower win rate — partly because of what caused it.

What actually changed#

Three things, all at once, which is bad science and honest history.

I moved from the 15-minute chart to the 1-hour chart. I capped my risk at 1% of the account on every single trade, with no exceptions and no "this one looks good". And because the structure on a higher timeframe is further apart, my targets moved out: a minimum of 3R, usually closer to 5R, measured to the previous swing rather than to whatever felt safe.

The third change is the one that lowered my win rate. A target about three times further away is a target price reaches less often. That is not a flaw in the method; it is the arithmetic of where you put the exit.

The numbers#

Fifteen points of win rate, gone. Average return per trade, up roughly nineteen times.

Read the risk column before you read anything else. Before June my average risk per trade was 3.85%, and two positions carried 70% of the account. That is a strategy with a thin edge whose results were drowned by position sizing. The 1,900 trades behind that column came out of my old journal, so they are real trades, not a backtest.

I knew the 1% rule long before June. What finally made me hold to it was my own data. In R-multiples my trades said I should be profitable; my account said the opposite. I suspected an inconsistency somewhere, so I loaded the whole history into the journal and went looking. It was risk management, and overtrading: more than a hundred trades a month before June, fewer than ten a month on average after. Once the risk percentage was fixed, the results I wanted came almost straight away.

Why a lower win rate can earn more#

The number that decides whether a strategy makes money is expectancy: what one trade returns on average, including the losers.

Expressed in R, where 1R is the amount you risk per trade:

expectancy = (win rate × average win in R) − (loss rate × average loss in R)

Here is the old version of my trading, from the log rather than from memory. 1,900 trades: 750 wins, 1,112 losses and 38 breakevens. Leaving the breakevens out, that is a 40.3% win rate, an average winner of +1.52R and an average loser of −0.99R.

(0.403 × 1.52) − (0.597 × 0.99) = 0.613 − 0.591 = +0.02R per trade

Two hundredths of R. Positive, technically. Over that many trades that is an edge you could describe at a party and not be lying, and it is nowhere near enough to survive anything going wrong.

Now the current version. 36 trades: 9 wins, 27 losses, no breakevens, a 25% win rate, an average winner of +4.44R, an average loser of −1.00R.

(0.25 × 4.44) − (0.75 × 1.00) = 1.11 − 0.75 = +0.36R per trade

Three out of four trades lose and the account still grows, because the one that works is worth more than four of the ones that don't. Same trader, same markets, from +0.02R to +0.36R per trade, on a win rate fifteen points lower.

Notice what did not change: the average loser, −0.99R before and −1.00R after. My losers averaged a full 1R in both periods, so stops were not the problem, which is exactly why the win rate was never the number worth fixing.

This is why win rate on its own tells you nothing. A 90% win rate is a disaster if the tenth trade gives back everything. A 25% win rate is a business if the winners are large enough and the losers stay small and controlled. Win rate vs risk-reward ratio works out the breakeven win rate for any target.

The part about small, controlled losers is the part people skip.

The hidden variable is sizing, not direction#

Here is the same sequence of trades, scored twice. Four losers and one winner at 5R, which is roughly my current distribution. First with disciplined 1% risk:

Now the same five trades, same wins and losses, with the sizing I used to apply — bigger when I felt confident, smaller when I didn't:

Identical trades. Identical win rate. Identical R-multiples. One version ends the week up, the other down nine percent — and the difference is entirely in how much was risked on each, which had nothing to do with the strategy at all.

The uncomfortable version of this: for more than a year I believed I had a strategy problem. My R-multiples said the strategy was slightly positive, +0.02R per trade across 1,900 of them. The money said I was losing. Both were true, and the entire gap between them was sizing — an edge that thin cannot survive risking 5% on the trades you like and 1% on the ones you don't.

When I first saw that +0.02R, it did not surprise me. It confirmed a hunch. The strategy worked, in a way, but it was not consistent enough: not in how much I risked, and not in where I took profit. The sizing explained why the money and the R-multiples disagreed. The inconsistent profit-taking explained why the R-multiples were so thin in the first place. Fixing one without the other was not going to be enough, so I changed the strategy and my risk-taking together.

If you want to see what consistent risk actually looks like in money terms before you commit to it, the position sizing calculator turns an account balance, a risk percentage and a stop distance into the position size — no account needed.

What 36 trades cannot tell you#

The honest part, because the numbers above are small.

Thirty-six trades is not proof. At a 25% win rate you expect long stretches of losses as a matter of course, and a single outlier moves everything. In my case the +13R total rests on July, which produced +14R by itself; the other four months add up to −1R. September was −6R: one win, eleven losses, ten of them in a row. If I had stopped reading after September I would have concluded the change failed.

What 36 trades can tell you is whether your process changed: whether risk is actually fixed, whether targets are actually where you said they would be, whether you are taking the setup you defined rather than the one in front of you. Those are facts about behaviour, visible immediately. Whether the edge is real needs far more data — enough that one good month cannot carry the result. How many trades before you know a strategy works works through what "far more" actually means.

So: the process change is established. The edge is a hypothesis with encouraging early evidence and a sample too small to defend.

I do have a line, though it is a flag rather than an exit. I check the running total once a month. If it falls 10R below its best point since June, I stop and work out whether something has to change. Measuring from the peak rather than from zero keeps that line the same distance away however good the previous months were. A rule like "stop if the total turns negative" gets looser with every winning month.

That line is not hypothetical. Those ten September losses took the running total from +20R on 2 September to +10R on 30 September — exactly 10R below the peak. I set the rule afterwards, so it did not stop me then.

What to measure instead#

Four numbers, in the order they matter. All of them come out of a journal that records risk and R, which is the thing most broker exports leave out.

1. Average R per trade. Your expectancy. Positive over a large enough sample suggests the approach pays; negative means screen time makes it worse.
2. Risk consistency. Average risk per trade, plus the maximum. If the maximum is several times the average, your results are being decided by sizing rather than by setups. This was my biggest problem.
3. Average win in R versus average loss in R. Together with win rate, this is where expectancy comes from. A low win rate needs big winners; the ratio tells you whether you have them.
4. Win rate. Last. Useful for setting expectations — at 25% losing streaks come with the territory (I had ten in a row in September; how long a losing streak is normal has the odds) and you need to not take them personally — and nearly useless as a measure of quality.

Common mistakes with this#

Raising the target to raise expectancy, without the structure to support it. Moving your take-profit from 2R to 5R on the same chart does not create a 5R trade. The target has to be somewhere price actually goes — a prior swing, a level that matters. Otherwise you have simply made a losing strategy slower.

Comparing your win rate with someone else's. Scalpers taking 0.5R profits run high win rates by construction. Swing traders holding for 5R run low ones. The two numbers are not on the same scale and the comparison is meaningless.

Treating a lower win rate as a licence to be wrong. The approach only works if your losses average close to 1R and an oversized one stays rare. One oversized loss erases several winners, and that is sizing again.

Concluding too early. In both directions. Three good weeks is not proof, and neither is one bad month.

The takeaway#

Win rate measures how often you are right. Expectancy measures whether being right pays. They are different questions, and only the second one — together with how much you risk on each trade — decides whether the account grows.

If your own log says your win rate is respectable and your account is not growing, stop looking at setups and look at the size column. Compute your average R, your average risk, and your largest risk. The gap between the average and the largest is usually where the money went.