Why Win Rate Alone Tells You Almost Nothing

Win rate is the most quoted statistic in trading and the least informative. Not uninformative — it is a real measurement of a real thing — but incapable of supporting a conclusion on its own, in a way that is easy to demonstrate and hard to unsee once you have.

This lesson demonstrates it. Two systems, the same win rate, and opposite outcomes, with the arithmetic laid out so you can check every line. Then the general rule that connects the two numbers, and a short list of what to ask when someone quotes a win rate at you.

!Important

Every figure on this page is illustrative. The two systems are constructed to make an arithmetical point, and both use a constant position size of 10,000 per trade so that no compounding effect can be blamed for the result. They are not backtests, and no claim is being made about any real strategy.

The counterexample

Both systems take exactly 100 trades. Both win 60 of them. Both report a win rate — which is to say, a "WinnersPercent" — of 60.00.

	System A	System B
Trades	100	100
Winners	60	60
Losers	40	40
Win rate	60.00	60.00
Average win	+400	+700
Average loss	−700	−400

Everything up to the last two rows is identical. Now finish the arithmetic.

System A

Profit of winners  =  60 × 400   =   24,000
Loss of losers     =  40 × −700  =  −28,000
Net Profit         =  24,000 + (−28,000)  =  −4,000

Avg. Profit/Loss   =  −4,000 / 100  =  −40.00
Payoff Ratio       =  400 / 700     =    0.571
Profit Factor      =  24,000 / 28,000  =  0.857

System B

Profit of winners  =  60 × 700   =   42,000
Loss of losers     =  40 × −400  =  −16,000
Net Profit         =  42,000 + (−16,000)  =  26,000

Avg. Profit/Loss   =  26,000 / 100  =  260.00
Payoff Ratio       =  700 / 400     =    1.750
Profit Factor      =  42,000 / 16,000  =  2.625

System A loses 4,000. System B makes 26,000. They have the same win rate. Someone who saw only the win rate saw the same thing in both cases, and it was the same thing: a number that had not yet been combined with anything.

?Reality check

“A 60 per cent win rate” is a claim that sounds like evidence of an edge and is not. It becomes evidence only when paired with the size of the wins relative to the losses. Any time you meet a win rate quoted alone — in an advertisement, a forum post, a book, or your own notes from three months ago — the correct response is that the sentence is incomplete, not that it is impressive or unimpressive.

The rule behind the example

The two numbers are locked together by one line of algebra. A system breaks even when the money won equals the money lost:

p × AvgWin  =  (1 − p) × AvgLoss

Rearranged, the payoff ratio a system needs in order to break even at win rate p is:

break-even Payoff Ratio  =  (1 − p) / p

Win rate	Break-even Payoff Ratio
30 per cent	2.333
40 per cent	1.500
50 per cent	1.000
60 per cent	0.667
70 per cent	0.429
80 per cent	0.250
90 per cent	0.111

At a 60 per cent win rate the break-even payoff ratio is 0.667. System A’s 0.571 is below it, so System A loses. System B’s 1.750 is far above it, so System B wins. The table predicts the counterexample without needing the counterexample.

Read the table the other way and it is more useful still. A rule advertising a 90 per cent win rate needs a payoff ratio above 0.111 to break even — which sounds easy, until you notice it means the average loss may be up to nine times the average win, and that is exactly the shape of rule that produces a 90 per cent win rate in the first place. A tight target and a wide stop manufacture win rate directly. The question is never whether the win rate is high; it is whether it is high enough for the payoff it was bought with.

The same argument from the other direction

Here are two more illustrative systems, again 100 trades each at constant size, with completely different win rates and identical expectancy:

	Trend-following	Mean-reverting
Winners	35 at +900	75 at +350
Losers	65 at −350	25 at −700
Win rate	35.00	75.00
Profit of winners	31,500	26,250
Loss of losers	−22,750	−17,500
Net Profit	8,750	8,750
Avg. Profit/Loss	87.50	87.50
Payoff Ratio	2.571	0.500
Profit Factor	1.385	1.500

Same expectancy per trade, to the cent. Win rates of 35 and 75. If win rate measured quality, one of these would be more than twice the system the other is; it measures something else entirely — the shape of the return distribution, not its centre.

The shapes do matter, but not in the direction the win rate implies. The trend-following system’s entire result rests on 35 trades, so its outcome is more dependent on a few events and its equity curve will spend longer going nowhere. The mean-reverting system loses twice as much per losing trade as it wins per winner, so a cluster of losses does more damage per event. Those are genuinely different risks, and neither is visible in “35 per cent” or “75 per cent”.

What the win rate does tell you

It is not a useless number. It is a useful number about something other than profitability.

It sets your expectation of losing runs. Pair "WinnersPercent" with "LosersMaxConsecutive", the longest run of consecutive losing trades in the test. At a 35 per cent win rate, if trades were independent, the probability of ten losers in a row is 0.65 raised to the tenth power, which is 1.35 per cent — and in 100 trades there are 91 places such a run could begin, so encountering one is unremarkable rather than a sign that the rule has stopped working. Trades from one strategy are not independent, so treat that as a lower bound on how bad a run to expect.

It tells you whether you can live with the rule. A person who abandons a system after six losses cannot trade a 35 per cent win rate rule, whatever its expectancy. That is a real constraint and it belongs in the design, not in a lecture about discipline. Knowing the win rate and the consecutive-loser figure before you commit is how you find out.

It describes the exit design. A high win rate usually means a fixed target that is close, or a stop that is far. A low one usually means a rule that lets winners run and cuts losers quickly. Reading it as a description of the exit is legitimate; reading it as a score is not.

Why the number gets quoted anyway

Some observations about how the statistic is used, offered as observations rather than accusations about anyone in particular:

• It needs no context to sound good. Bounded between 0 and 100, it reads like a school mark. “Net Profit % of 62 over five years on 214 trades in a filtered ASX 200 universe, after 0.10 per cent commission per side” is a more informative sentence and does not fit on a banner.
• It matches what people want. The intuition of trading is “being right”, and win rate is the only reported statistic that measures being right.
• It is the easiest statistic to engineer. Widen the stop and tighten the target and the win rate climbs, on the same signals and the same data. Nothing about the rule improved; the distribution was reshaped, and the half that got worse is the half nobody is quoting.
• It is technically true. That is what makes it durable. There is nothing to refute.

✕Common mistake

The version of this error that costs you money is not believing an advertisement. It is optimising your own system on win rate, or abandoning a rule after a losing streak that its own reported "LosersMaxConsecutive" said to expect. Both are cases of reading one number without the one it is inseparable from.

Questions to ask instead

When a win rate is quoted — including by you, to yourself — these are the questions that turn it back into information:

1. What is the Payoff Ratio? Without it the win rate is half a sentence.
2. What is Avg. Profit/Loss? This is the expectancy per trade, and it is the number the win rate was standing in for.
3. How many trades? A 60 per cent win rate over 20 trades is twelve wins. The previous lesson’s standard-error table applies.
4. What is the largest single win as a share of Net Profit? If one trade supplied a third of the profit, the averages are describing an accident.
5. What was the exit rule? A fixed profit target manufactures win rate by construction. Knowing the exit tells you whether the win rate is a finding or a design choice.
6. What is Max. system % drawdown, and how long did it last? The win rate says nothing about either.
7. What costs were assumed, over what universe and period? Every figure above changes when these change, and a win rate quoted without them is not reproducible.

If the answer to the first two is unavailable, you have not been told anything about the system yet.

Recap

Two systems with a 60 per cent win rate ended 30,000 apart because their average win and average loss were swapped. The break-even payoff ratio at any win rate is (1 − p) / p, which turns the counterexample into a general rule you can apply to any pair of numbers. Two more systems with win rates of 35 and 75 had identical expectancy, which shows the same point from the other side: win rate describes the shape of the distribution, not its centre. Avg. Profit/Loss — AmiBroker’s expectancy — is the combination that actually answers the question, and the win rate’s real jobs are predicting losing runs and describing the exit design.

The next lesson leaves the summary table entirely and looks at the shape of the equity curve, where a different kind of self-deception lives.

Check your understanding

Question 1. A system wins 45 per cent of its trades. What Payoff Ratio does it need simply to break even, before costs?

• 0.45
• 1.00
• 1.222
• 2.222

Show the answer and why

Answer: 1.222

The break-even payoff ratio is (1 − p) / p, which is 0.55 / 0.45 = 1.222. The average win must exceed the average loss by about 22 per cent for the system to reach zero. Anything less and a 45 per cent win rate loses money, however respectable the percentage sounds.

Question 2. System P: 60 winners at +700, 40 losers at −400. System Q: 60 winners at +400, 40 losers at −700. Both take 100 trades at constant size. What separates them?

• Their win rates, which differ
• Nothing measurable; the two are equivalent
• Their Payoff Ratio, and therefore their expectancy: P makes 26,000 and Q loses 4,000
• Their exposure

Show the answer and why

Answer: Their Payoff Ratio, and therefore their expectancy: P makes 26,000 and Q loses 4,000

Both win 60 per cent of their trades. P has a Payoff Ratio of 1.75 and an Avg. Profit/Loss of +260; Q has a Payoff Ratio of 0.571 and an Avg. Profit/Loss of −40. The identical win rate conceals a 30,000 difference in Net Profit, which is the whole reason a win rate must never be read alone.

Question 3. Which of these are legitimate uses of the win rate? Select all that apply.

• Estimating how long a losing run to expect, alongside the consecutive-losers figure
• Deciding whether a rule is one you could actually keep trading
• Ranking two systems by quality
• Inferring something about how the exit rule is designed

Show the answer and why

Answer: Estimating how long a losing run to expect, alongside the consecutive-losers figure, Deciding whether a rule is one you could actually keep trading, Inferring something about how the exit rule is designed

Win rate is informative about the shape of the return distribution: expected losing streaks, psychological tolerability, and the kind of exit that produced it. It cannot rank systems by quality, because two systems with the same win rate can end on opposite sides of zero and two with very different win rates can have identical expectancy.

Question 4. You tighten a system’s profit target from 8 per cent to 2 per cent and the win rate rises from 41 to 68 per cent. What have you learned?

• That the system has been improved
• That the entry signal is more accurate than it appeared
• That you moved a design dial: more trades close positive and every winner is now capped at 2 per cent, so the payoff ratio has fallen
• That the original target was wrong

Show the answer and why

Answer: That you moved a design dial: more trades close positive and every winner is now capped at 2 per cent, so the payoff ratio has fallen

Nothing about the entry changed. A closer target converts trades that would have retraced into small winners and truncates the large winners at 2 per cent, which raises the win rate and lowers the payoff ratio by construction. Whether the change helps or hurts is decided by Avg. Profit/Loss after costs, not by the win rate — and shorter, more frequent trades pay costs more often.

Check my answersClear