Win rate and risk-reward ratio: why you can make money while winning fewer than half your trades

Win rate is the percentage of trades that close in profit, and on its own it doesn’t tell you whether a system makes money. What decides that is how it combines with the risk-reward ratio: if your winners are, on average, twice the size of your losers, being right more than 33.3% of the time is enough to come out ahead, before costs. The price of being right less than half the time is losing streaks that run longer than intuition expects, and here we work them out.

What are win rate and the risk-reward ratio?

Win rate (also called hit rate) is the percentage of closed trades that ended in profit, out of the total. If 40 out of 100 trades end positive, your win rate is 40%. In the MetaTrader 5 Strategy Tester report it appears as “Profit Trades (% of total)”. It’s the number people show off the most, and the one that misleads the most when it stands alone.

The risk-reward ratio compares what you lose when you’re wrong with what you make when you’re right. There’s the one you plan when you open the trade, from the distance to your stop loss and take profit, and the one you actually realize, which is the one that counts. The realized ratio is calculated by dividing the average gain of your winning trades by the average loss of your losing trades. If you lose €100 on average and make €200 on average, your ratio is 1:2: for every euro that goes out when you’re wrong, two come in when you’re right.

The two numbers go together. A win rate without a ratio is like knowing how many bets you won without knowing how much you staked on each one: it says nothing about the final result. That’s why two systems with the same win rate can end up one making money and the other wiped out.

How the ratio is written (and why it’s confusing)

You’ll see the ratio written in several ways. “1:2” is usually read as “I risk 1 to make 2,” but some people write “2:1” to mean the same thing, or just a single number, R = 2. In this article we always use 1:R: you risk 1 to make R. When you read a ratio somewhere else, check which number comes first.

Trading expectancy: how win rate and risk-reward combine

Win rate and ratio combine into a single number: expectancy per trade (or mathematical expectation), which is what each trade leaves you on average if you repeat the system many times. If you call the win rate p, the average gain G and the average loss L:

An illustrative example: you’re right 40% of the time, you make €200 on average and you lose €100 on average. E = 0.40 × 200 − 0.60 × 100 = 80 − 60 = €20 per trade. You’re wrong six times out of ten and, even so, the expectancy is positive: if those figures held, the account would grow over the long run.

If you set E to zero, you get the breakeven win rate: the minimum you need to be right just to avoid losing. With a 1:R ratio it’s a simple division: p = 1 / (1 + R).

Now add costs

That formula is a lab version. In a live account every trade pays the spread, a commission and sometimes swap and slippage, whether you win or lose. With C as the average cost per trade, it becomes E = p · G − (1 − p) · L − C. In the previous example, a cost of €10 per trade cuts the expectancy from €20 to €10: half the margin goes on tolls.

The minimum win rate also rises, to (1 + c) / (1 + R), where c is the cost expressed as a fraction of the average loss. The right-hand column of the table assumes c = 10%: with a 1:2 ratio you go from needing a 33.3% win rate to 36.7%. Gold’s costs, one by one, are in spread, slippage and commissions.

Why a high win rate can be a trap

If being right often guarantees nothing, why does it sell so well? Because the mind likes being right: ten green trades feel like talent, and four losses in a row feel like failure, even when the account says the opposite. Anyone who designs a system to look good in a screenshot knows this.

Do the math on a system that makes €10 on nine trades out of ten and loses €100 on the tenth. It’s right 90% of the time, and its expectancy is 0.90 × 10 − 0.10 × 100 = 9 − 10 = −€1 per trade. With that ratio, 1:0.1, it would need to be right 90.9% of the time just to break even, and with costs at 10% of the average loss, 100%: not a single losing trade. It’s a machine that loses money slowly while looking like a machine that makes it.

The extreme version is systems that won’t accept a loss: they postpone it by averaging down against the move, or they double the lot size after every loss, which is the martingale. Almost all their trades end in the green, until the one that doesn’t, which wipes out months of gains or the whole account. Why you should be wary of them is covered in are trading robots profitable?; here the arithmetic is enough: the smaller the ratio, the more everything hinges on the rare trade.

The opposite is also true: a low win rate doesn’t make a system bad, as long as its winners make up for its losers. It just asks something else of you in return: the stomach to sit through losing streaks.

Win rate and profit factor in the AUREUM simulation

The homepage shows AUREUM’s backtest on gold (XAU/USD) from 2020 to 2026, run with the MetaTrader 5 Strategy Tester: 439 trades, 43.51% winning trades, a profit factor of 1.49, a maximum drawdown of 21.84% and a simulated net return of +410%. It’s a simulation on past data, not a live account, and it guarantees nothing about the future.

The simulation lost on more trades than it won. So how does that fit with a profit factor above 1? Profit factor (PF) is total gains divided by total losses. With n trades, total gains are n · p · G and total losses are n · (1 − p) · L, so:

Solving for the ratio, G/L = PF × (1 − p) / p = 1.49 × 0.5649 / 0.4351 ≈ 1.93. In the simulation, the average gain of the winning trades was about 1.93 times the average loss of the losing ones: a ratio close to 1:2. Careful: this figure isn’t published. It’s our own arithmetic deduction from two of the simulation’s figures, and it’s only as good as they are.

With that ratio, the breakeven win rate would have been 1 / (1 + 1.93) ≈ 34.1%. The simulation was right 43.51% of the time, about 9.4 points above that. That cushion explains why the simulation’s expectancy came out positive, and the homepage expresses it in dollars: an expected payoff of $46.75 per trade on a $5,000 initial deposit. In your account, costs and slippage will be your broker’s, not the simulation’s, and they can shrink that cushion.

Losing streaks in trading: how many in a row are normal?

With a 43.51% win rate, each trade loses with probability q = 1 − 0.4351 = 0.5649, or 56.5%. If trades are independent, the probability that k trades in a row, starting from a given one, are all losers is q^k:

• 5 losses in a row: 0.5649⁵ ≈ 5.75%, about one time in 17.
• 8 losses in a row: 0.5649⁸ ≈ 1.04%, about one time in 96.
• 10 losses in a row: 0.5649¹⁰ ≈ 0.33%, about one time in 302.

They look small, but every trade is a new starting point, and in 439 trades there are hundreds of them: what’s unlikely at any single one becomes likely across the whole set.

The longest losing streak in 439 trades

A classic approximation for the longest losing streak in n trades is log(n · p) / log(1/q). Here, log(439 × 0.4351) / log(1/0.5649) ≈ 9.2. To refine it, we ran a Monte Carlo simulation: 200,000 series of 439 independent trades, each with a 43.51% chance of winning, measuring the worst streak in each series.

So the typical worst streak is around nine losses in a row, and in one series out of ten it reaches 13 or more. An exact calculation of the same distribution confirms the median, the 90th percentile and that 47%.

What this math doesn’t see

The calculation treats each trade like a loaded coin with no memory. Markets aren’t always like that: there are periods when a strategy fits worse and losses cluster together. That’s why reality can bring longer streaks than the ones in the table, and the future win rate doesn’t have to match the backtest’s. Take these numbers as the minimum you should be prepared to see, not as a ceiling.

How to use win rate before you switch on a trading robot

• Don’t mistake a normal streak for a malfunction. With a win rate like the simulation’s, eight or ten losses in a row are within what you should expect. Switching the robot off at that point and back on after the recovery is one of the costliest mistakes with an Expert Advisor. Decide in advance, with a cool head, what drop would make you stop.
• Size your lot for a plausible streak, not for the average. An illustrative example: if each loss cost you 1% of your balance, 13 losses in a row (the simulation’s 90th percentile) would leave you 12.2% down; at 2%, 23.1% down. And the total drop, the drawdown, is usually bigger than the worst streak, because a bad phase mixes losses with gains that don’t make up for them.
• Look at profit factor and drawdown before win rate. Profit factor already sums up win rate and ratio in one number, and drawdown tells you how much it hurts along the way. How to read them together: the Strategy Tester report.

In AUREUM, the default lot mode calculates the size for you: 0.01 lots for every 1,000 of balance. The installation guide calls it a conservative rule and warns that, with a lot size that’s too aggressive, “a losing streak hurts.” How much capital you commit is still your decision.

The core idea fits in one line: you don’t make money by being right often, but by what you make when you’re right versus what you lose when you’re wrong, minus costs. And to stick with a system, it helps to know in advance how long its streaks can last.