Our signals win 64.9% of the time and lose money

August 13, 2026

Across 11,615 resolved futures signals we win 64.9% of the time and average -0.0208 R per trade. Here is the arithmetic that makes those two facts compatible, and why a published win rate tells you almost nothing on its own.

Our futures signal rail wins 64.9% of the time.

It also loses money: -0.0208 R per trade across 11,615 resolved outcomes.

Both numbers are correct. Understanding why they are compatible is the single most useful thing you can learn about reading anyone's track record, including ours.

The arithmetic

A win rate is a count. Expectancy is a weighted average. They only agree when the average win and the average loss are the same size, and they almost never are.

In our case the shape looks like this:

The wins are small and frequent. The losses are large and rare. A stop costs roughly 2.7 times what a typical win returns, so at 64.9% the maths comes out fractionally negative:

0.649 x (+0.365) + 0.351 x (-0.969) = -0.103

The realised figure is better than that back-of-envelope because target hits pay more than trail locks. It is still below zero.

Why so few targets

Only 748 of 11,615 outcomes reached the target. The rest resolved as a protective trail lock, a stop, or an expiry.

That is what a protective exit does: it converts a would-be large win into a reliable small one, and it converts some would-be losses into small ones too. Whether that trade is worth making is an empirical question, not a stylistic one — and for us the honest answer so far is approximately break-even, slightly negative.

The version of this that fools people

We studied a public signal channel advertising 90.6% wins and a profit factor of 6.77. Impressive, until you look at how a trade closes.

There were no stops. Winners hit target and closed, entering the statistic. Losers had no exit but liquidation, so they stayed open and never entered it. While that 90.6% was on display, the channel's own open book held two positions near -20%, one deteriorating from -19.51% to -21.73% in four hours against a -33% liquidation level.

This is not deception. It is what a no-stop design must produce:

A win rate computed over closed trades is only interpretable when losses are

forced to close on the same terms as wins.

Our rail is built the opposite way on purpose. A real stop plus target, stop and expired resolution means a loser enters the record on identical terms to a winner. Our number therefore looks worse and means more.

Also: state the baseline

A 90.6% win rate sounds extraordinary. With that channel's barriers — a 7% target against a 33% liquidation — random entry produces about 82.5% wins by chance.

So the edge on display was roughly eight points, not ninety. A win rate without its random-entry baseline is not a measurement; it is a number with a percent sign.

We wrote separately about how we measure that baseline instead of assuming it, because the standard shortcut is wrong by up to 24 points.

What to ask instead

When you see a track record — ours or anyone's:

• What is the expectancy per trade, in R? A win rate without it is decoration.
• Are losses forced to close? If a loser can stay open indefinitely, the win

rate is measuring patience, not skill.

• What is the random-entry baseline for those barriers? Measured, not assumed.
• How many trades, over how many distinct days? One good day can carry a

whole sample; we cap each day's weight for exactly that reason.

• Is expiry counted? A thesis that ran out of time is a real outcome. Dropping

it flatters the rate.

Why we publish this

We could report 64.9% and stop. It is true, it is flattering, and almost nobody would check.

We publish the -0.0208 R beside it because a terminal you use to make decisions is worth less than nothing if its numbers are chosen to impress you. The point of the rail is not that it prints winners. It is that it keeps score honestly enough to tell us when it does not.