A High Win Rate Doesn't Mean Profit: Check the Break-even Win Rate First

Win rate is the share of trades that close in profit. It sounds like the most important number, but on its own it doesn’t tell you whether you make or lose money. You have to compare it with the break-even win rate, which depends on how much a winner makes compared with how much a loser costs. A system that wins 95% can lose money, and a system that wins 39% can make money. This article works through the math step by step and looks at real results from our tests on gold.

What is the break-even win rate

RR (Risk:Reward) is the profit of a winning trade compared with the loss of a losing trade. RR 1:2 means you risk 1 to make 2. In this article we write it as a single number, for example RR = 2.

The break-even win rate is the win rate at which total profit exactly equals total loss.

Break-even win rate = 1 ÷ (1 + RR)

RR (gain ÷ loss)	Break-even win rate
3	25.0%
2	33.3%
1	50.0%
0.5	66.7%
0.25	80.0%
0.05	95.2%

The lower the RR, the more often you need to win. A system that makes a little and loses a lot has to win almost every trade just to break even. To try your own numbers, use the RR calculator at the end of this article.

Example: winning 95% and still losing

Say you trade gold at 0.01 lot, where a 1-dollar price move = 1 dollar of profit or loss. You set Take profit at +1.3 dollars and Stop loss at −26 dollars.

1. RR = 1.3 ÷ 26 = 0.05
2. Break-even win rate = 1 ÷ (1 + 0.05) = 1 ÷ 1.05 = 95.2%
3. Over 100 trades, 95 wins make 95 × 1.3 = 123.5 dollars
4. 5 losses cost 5 × 26 = 130 dollars
5. Net = 123.5 − 130 = −6.5 dollars

Check: with 96 wins you make 96 × 1.3 = 124.8, and 4 losses cost 104, so the net is +20.8 dollars. The break-even sits between 95% and 96%, matching the 95.2% we calculated. Just one trade in a hundred flips the result from a loss to a profit.

Spread pushes the break-even above the formula

The formula 1 ÷ (1 + RR) doesn’t include the cost of the spread (the gap between the buy and sell price, which you pay on every trade). Spread does two things at once: it makes winners smaller and losers bigger.

Break-even win rate after spread = (loss + spread) ÷ [(loss + spread) + (gain − spread)]

Example, with a spread of 2 dollars per trade:

• Gain 10, loss 10: (10 + 2) ÷ (12 + 8) = 12 ÷ 20 = 60%, not 50%
• Gain 50, loss 50: (50 + 2) ÷ (52 + 48) = 52 ÷ 100 = 52%

Check the first case: over 100 trades, 60 wins make 60 × 8 = 480 and 40 losses cost 40 × 12 = 480. Net is exactly 0.

Same RR, but the small target needs a much higher win rate, because the spread eats a large share of the distance. This gets even worse on small timeframes. Read more in Spread cost on small timeframes.

Back to the +1.3 / −26 example: with a spread of about 0.27 dollars (close to what we saw on gold in our tests, at 0.01 lot), the break-even becomes 26.27 ÷ (26.27 + 1.03) = 26.27 ÷ 27.3 = 96.2%.

A high win rate mostly comes from the shape of TP and SL

If price moved randomly with no direction, the chance of hitting the target before the Stop loss would be loss ÷ (loss + gain), exactly the same number as the break-even win rate. That means with a near TP (Take profit) and a far SL (Stop loss), anyone can win often, even picking direction with a coin flip. Random entries break even before costs, then lose by the amount of the spread.

In our backtest, we took apart an 8-candle box breakout system on gold, whose SL sits about 26 dollars away, and compared it with random entries. We used real 2026 tick data and measured in price distance, not tied to lot size or balance.

• Random entry, counting whether price hits +1.3 before −26: wins 95.3%
• The real system: wins 92.8–95.8% across several test variants

The real system’s win rate is close to random entry, so the breakout entry doesn’t guess direction any better than chance. Whatever edge is left is very thin, and it doesn’t come from the win rate at all. And the random entry’s 95.3% is still below the 96.2% break-even after spread we calculated, which is exactly winning 95% and still losing.

These are backtest results. Past results do not predict future results, and they are not returns you should expect.

Same entries, different exits, opposite results

Our clearest example uses the same Order block + FVG entries on gold M5, from 5 January to 24 July 2026, with real tick data, 0.02 lot per trade and a 3,000-dollar balance. Only the exit changes.

• arm/lock: once floating profit reaches 2 dollars, move the Stop loss to lock in profit and trail it
• Fixed TP: take profit at 10 dollars
• Far target: take profit at the Fibonacci 161.8 level of the swing

Exit	Trades	Win rate	Profit factor	Actual RR	Break-even	Net (dollars)
arm/lock at +2	1,695	78.5%	0.86	0.24	80.9%	−493
Fixed TP 10 dollars	1,667	53.1%	0.99	0.88	53.3%	−95
Far target Fibonacci 161.8	1,242	38.9%	1.31	2.06	32.7%	+3,185

Profit factor is total profit divided by total loss. “Actual RR” is worked back from the Profit factor using RR = Profit factor × (1 − win rate) ÷ win rate, so it is an average already net of spread.

Work through the last row: RR = 1.31 × 0.611 ÷ 0.389 = 2.06. Break-even = 1 ÷ 3.06 = 32.7%. Check: per 100 trades, 38.9 wins × 2.06 = 80.1 units, and 61.1 losses × 1 = 61.1 units. 80.1 ÷ 61.1 = 1.31, matching the Profit factor.

The exit that won most often lost the most, because it gave up on trades that would have run far as soon as they reached +2 dollars. Gold pays out through a few big trades, and if you cut those, only the small ones are left. Another example from the same test set: an arm/lock exit on Fibonacci-level entries on gold M1 won 90% but lost 1,644 dollars (Jan–Jul 2026, real ticks, 0.02 lot, 3,000-dollar balance). For more on choosing a near or far target, read Near or far Take profit.

A caution: the far target with the same settings lost 1,052 dollars in 2022 and 666 dollars in 2023 (the older years use ticks simulated from 1-minute data). A low win rate is not automatically good either. Every version has to be compared with its break-even in the same way. All of this is backtest results. Past results do not predict future results, and they are not returns you should expect.

ICT sits right on the line too

The red and green points on the right of the chart come from our test of the ICT previous-day High/Low sweep setup. It won 31.2% at RR 1:2 and 24.6% at RR 1:3. That sounds low, but that is not the problem. The problem is that it sits right at break-even, so there is no edge. The full details, including test conditions and yearly results, are in ICT tested on gold. Backtest results do not predict future results, and they are not returns you should expect.

Read win rate correctly in 3 steps

1. Find the actual RR from average profit per winning trade ÷ average loss per losing trade, not the RR you planned when placing the trade
2. Calculate the break-even with 1 ÷ (1 + RR). If you use the planned RR, use the after-spread formula instead
3. Look at the margin: actual win rate minus break-even. A margin of 1–2 points is like a coin toss. Only 5 points or more, over many trades, starts to get interesting

And always ask where the win rate comes from. If it comes from a wide SL or from moving the Stop loss away from losing trades, read Wide or tight Stop loss and Grid and Martingale, the systems that win most often and crash hardest.

Common mistakes

• Comparing win rate with 50%. The right benchmark is the break-even for that RR. A 60% win rate can lose, and a 35% win rate can profit
• Forgetting spread. A small-target system can look like it breaks even by the formula but lose in reality because the spread eats almost everything
• Buying win rate by cutting winners early. A near TP or an early profit lock raises the win rate, but the profit from trades that run far disappears, like the arm/lock row in the table above
• Trusting a system after a long winning streak. A low-RR system winning a dozen trades in a row is normal, and it doesn’t mean there isn’t a big loss waiting. High-RR systems can also have long losing streaks. Read Losing streaks are normal