How Do You Spot a Fake Win Rate?

Why is a win rate meaningless on its own?

Because it says nothing about size. A strategy is profitable when the average win times the win rate exceeds the average loss times the loss rate, so the same win rate can describe an excellent system or a ruinous one.

The breakeven win rate for a payoff ratio R is 1/(1+R):

Payoff ratio (avg win : avg loss)	Breakeven win rate
3.0 : 1	25.0%
2.0 : 1	33.3%
1.0 : 1	50.0%
0.5 : 1	66.7%
0.1 : 1	90.9%

A 90% win rate at 0.1:1 is below its own breakeven, so it loses money, and a 35% win rate at 3:1 is comfortably profitable. Quoting the first number without the second is the whole trick, and it is usually not even deliberate.

How do the usual inflations work?

Scratches as wins. A trade closed for $3 on a $5,000 position is a win in the count and a rounding error in the money. A count-based win rate rewards taking a lot of them.

Unresolved trades excluded. A position that went wrong and is still open is not a loss yet. A denominator of closed trades only, with the bad ones left open indefinitely, produces an extremely high win rate and no lie is told.

Averaging down. Adding to a losing position lowers the average price until the trade closes green. Every individual trade is a win; the account is taking larger and larger risk to manufacture them, and the eventual loss is enormous.

Partial exits counted separately. One entry scaled out in four pieces becomes four wins. The same trade closed at a loss is one loss.

What should you ask for?

1. The payoff ratio. Average win divided by average loss. Without it the win rate is not a performance number.
2. The definition of a win. Is a $3 scratch a win? Is a partial exit a separate trade?
3. The count of open trades. If there are twelve positions open and eleven are underwater, the win rate describes the eleven you were not shown.
4. The worst drawdown. Averaging down is invisible in a win rate and obvious in a drawdown.

Then check the claim against the tier of trading being described. Sustained win rates above roughly 80% do exist — in high-frequency market making and in premium selling, both of which pay for it with a payoff ratio well under 1 and an occasional very large loss. A discretionary directional trader posting 90% is describing one of the constructions above.

What settles it permanently?

Every trade, including the losers, timestamped before the fact, on a record the person being evaluated cannot edit, with sizes and exits, so the win rate can be recomputed rather than accepted.

There is an existing professional answer to this, and it is worth knowing it exists. The Global Investment Performance Standards — the voluntary standard the CFA Institute maintains for anyone presenting investment performance — are built on two principles: fair representation and full disclosure, with performance reported across composites rather than selected accounts.^[1] A win rate quoted without its payoff ratio and its sample fails both, which is why the number on its own is not a performance figure.

The "unresolved trades excluded" construction above is not a quirk of dishonest sellers either. Odean found across 10,000 brokerage accounts that investors realise gains far more readily than losses,^[2] so at any moment the closed trades skew toward winners and the open ones toward losers. And on what a number has to clear to mean anything, Harvey, Liu and Zhu argued that after so much testing a new claim in finance needs a t-statistic above 3.0 rather than 2.0.^[3]

That is what a kappi profile carries: the trade is committed before it resolves, and PnL, RME, correlation to SPX, mean R:R and trade count are published over 30, 100 and 200-day windows, at $15/month. A win rate computed from a record that includes the losses cannot be constructed, only measured.