Risk-reward and win rate: why being right a lot is not the same as making money

Two traders sit down at the end of the year. The first won 80% of his trades. The second lost more of hers than she won. The first one lost money. The second finished the year comfortably ahead.

Nothing about that is a paradox or a trick. It falls out of two multiplications anybody can do on an envelope, and the fact that most people selling systems lead with win rate, the number that means the least on its own, tells you what it is really for.

Win rate on its own is not a result

A win rate is the share of your trades that end in profit. That is all it measures. It says nothing about how big the wins are, nothing about how big the losses are, and therefore nothing about whether the account went up.

You can manufacture any win rate you like. Want 95%? Set a target one tick away and a stop fifty ticks away. You will win almost every trade and go broke on the ones you do not. Want 20%? Tiny stop, enormous target. Both fit in a headline. Only one of the two numbers you need to judge them is in the headline.

The formula that settles it

Let R be the ratio of your average win to your average loss. The win rate at which you exactly break even is:

Breakeven win rate = 1 divided by (1 + R)

That is the whole of it. Three cases people actually trade:

• R = 1 (win as much as you lose): 1 ÷ 2 = 50%
• R = 2 (target twice the risk): 1 ÷ 3 = 33.3%
• R = 3: 1 ÷ 4 = 25%

Anything above that line makes money before costs. Anything below it loses, no matter how good it feels.

The three cases, worked out in full

Take 100 trades, risking $100 on each, and do the arithmetic rather than trusting the formula.

At 1:1 and a 50% win rate. Fifty winners at $100 is $5,000. Fifty losers at $100 is $5,000. Net: zero. A hundred trades, a hundred commissions, and you finish behind where you started. This is the shape of most discretionary trading.

At 2:1 and a 33.3% win rate. Thirty-three winners at $200 is $6,600. Sixty-seven losers at $100 is $6,700. Net: minus $100, breakeven plus rounding. Note what happened: you were wrong two times out of three and finished level.

At 3:1 and a 25% win rate. Twenty-five winners at $300 is $7,500. Seventy-five losers at $100 is $7,500. Net: zero. Wrong three times out of four, still not losing.

Now push each a few points past breakeven, which is what a real edge looks like. At 2:1 and 40%: forty winners at $200 is $8,000, sixty losers at $100 is $6,000, net $2,000. Modest, unglamorous and entirely realistic. The same logic one notch further out is why a system that places its target at three times its risk on every entry does not need to be right most of the time: at 3:1 the line to clear is 25%.

You do not get paid for being right. You get paid for the size of being right, times how often.

Same win rate, six different outcomes

Hold the win rate fixed at 40% and vary only the ratio. Same hundred trades, same $100 of risk each, so forty winners and sixty losers every time.

One column changed. The same 40% hit rate is a losing system at 1:1, exactly level at 1.5:1, and a solidly profitable one at 3:1. If you wondered why serious systems obsess over where the target sits relative to the stop, this table is the reason.

Expectancy: the number that actually pays you

Combine the two and you get expectancy, the average dollars a system makes per trade:

Expectancy = (win rate × average win) - (loss rate × average loss)

The other half of the picture is the geometry rather than a hit rate. The target sits at three times the stop, so the breakeven line is 25%: one winner in four and the system is level. Everything above that line is the edge, and the profit factor is what tells you how far above it the run actually finished.

Do that check on any system you are considering. If the headline profit cannot be rebuilt from the component figures to within a percent or two, a number is wrong or missing, and it is worth knowing which before you pay.

Where the commissions go

Two dollars and change sounds like nothing because the expectancy is seventy-five. Put the same $2.40 against a system whose expectancy is $8 a trade and it takes nearly a third of the edge. Against a 1:1 scalping system running twenty trades a day it takes all of it. Costs scale with your trade count, not your profit, which is why high-frequency retail systems die quietly rather than dramatically.

Profit factor, and what a suspiciously good one means

Profit factor is gross profit divided by gross loss: dollars collected for every dollar paid out. Below 1.0 it loses money. Above 1.0 it makes money.

How a 90% win rate is manufactured

Now the trick, worked out in numbers so you never have to wonder again.

Set your target at $100 and your stop at $1,000, a ratio of 0.1 to 1. Price wanders a hundred dollars in your favor far more often than a thousand against you, so you win the large majority of trades. Your screenshots are a wall of green.

The breakeven win rate at that ratio is 1 ÷ 1.1 = 90.9%. So run a hundred trades at a 90% win rate, which any marketer would call outstanding:

• 90 winners at $100 = $9,000
• 10 losers at $1,000 = $10,000
• Net: -$1,000

Ninety percent right and losing money. The wins are the marketing; the tenth trade is the product. Every variation exists in the wild: grid bots with no stop, martingale sizing that doubles after a loss, "we never take a loss, we just hold." One shape: many small wins financed by a rare enormous loss that has not shown up in the sample you were shown.

The tell is easy to check. When a seller leads with win rate, ask for the average win and average loss in dollars and the maximum drawdown. If those are unavailable, proprietary, or answered with a different number, you have learned what you needed. More of these patterns are in how to spot a trading bot scam in five minutes.

Losing streaks are arithmetic, not malfunctions

A system whose target sits three times beyond its stop is built to be wrong more often than it is right and to be paid for the difference when it is right. Streaks follow directly, and they are longer than intuition expects. Take a system that loses 60% of the time:

• Six losses in a row: 0.6 to the sixth power = 4.7%. Across a thousand trades, that turns up several dozen times.
• Ten losses in a row: 0.6%. Across the same sample, roughly half a dozen times.

So somewhere in a long simulated history there are separate occasions where a system like this loses ten trades consecutively. Not bugs: the expected behavior of a coin weighted against you on frequency and three to one in your favor on size, and the reason the edge only shows up over hundreds of trades. What those streaks do to an account, and to the person watching it, is the subject of drawdown.

What to ask before you pay for anything

1. Win rate, average win and average loss, all three. Any two of them without the third is not an answer.
2. Trade count. A 65% win rate over 40 trades is noise. Over 4,000 it is a measurement.
3. Profit factor, and whether it is net of commissions. Gross-only numbers flatter high-frequency systems most.
4. Maximum drawdown and longest losing streak. These tell you what you have to sit through to collect the expectancy.
5. Whether the published total reconciles. Multiply expectancy by trade count. If it does not land within a couple of percent, ask why.

All five are on the performance page, and the reasoning behind the risk side is on the risk page. For the plain-language version of what a system like this is, start at what a trading bot actually is; for the honest version of the money question, how much a trading bot actually makes.

Frequently asked questions

What is a good win rate for an automated system?

There is no good win rate in isolation, because the only thing that matters is where it sits relative to the breakeven rate implied by the average win-to-loss ratio. A system winning 30% with a 4:1 ratio is far healthier than one winning 70% with a 0.4:1 ratio. Work out 1 divided by (1 plus the ratio) and compare it against the actual hit rate; the gap between those two numbers is the edge.

Is a high profit factor always better?

Up to a point. A profit factor comfortably above 1.0 across a large sample indicates a real edge, but figures of 4 or more in a retail backtest usually mean the system was fitted to its own history or its worst losses have not occurred yet. Treat an unusually high profit factor as a prompt to check the trade count, the out-of-sample period and whether costs were included.

Educational content, not investment advice. Trading futures involves substantial risk of loss and is not suitable for every investor. Any reference to Algomyth results is hypothetical, simulated over historical data, and shown in full on the performance page; past performance does not guarantee future results.