I graded every signal I forwarded. Under one accounting method they lose money.

Posted by the operator of RebateMax

Corrected within hours of publishing, 2026-08-16. A duplicate audit completed today voided 481 rows where a source channel had reposted the same signal — each repost had been counted as a separate trade. The first version of this post said 917 trades averaging +0.52R scaled out. De-duplicated it is 654 trades averaging +0.46R: fewer trades, a worse average, same conclusion. The full void ledger is public, and this note exists because correcting the record beats pretending it was never wrong.

I forward trading signals from third-party channels to a Telegram community. I don't produce them, I don't claim an edge in them, and I've never published a win rate. This post is what happened when I graded every single one against exchange candle data and worked out whether following them actually made money.

The answer depends on one decision, and the gap is wide enough that publishing only one side of it would be a lie.

Take profit at each target as you reach it, and the book averages +0.46R per trade. Never book, ride every position to its stop or its final target, and the same book averages −0.08R.

Same signals. Same entries. Same stops. One is a business; the other loses money slowly. That difference isn't a rounding error or a methodology quibble — it's the whole result.

What I measured, and how

Every signal that reaches the community is logged when it arrives: pair, direction, entry range, stop loss, and its ladder of take-profit targets. There's no editing after the fact, because the row is written before the outcome is known.

Grading is done against exchange candle data. A trade counts as filled when price trades into the entry range. From there it's followed forward until it either reaches its stop or its final target, and every target it touches along the way is recorded.

As of this writing, that's 674 graded outcomes across 654 trades from 10 source channels, measured over seven weeks. Some signals can't be graded at all — the entry never filled, or the message was too malformed to parse a stop from. Those are counted separately and excluded rather than quietly dropped into the win column. You can see the live figures, re-measured daily, on the scoreboard.

I publish two averages because a signal is not a trade. A signal tells you where to get in and where to get out if you're wrong. What it does not tell you is what to do when price is up and hasn't reached the final target yet — and that decision, which the signal never makes for you, turns out to dominate the result.

The two conventions

Scale out means taking profit at each target as price reaches it. A trade that touches its second target and then reverses into its stop still banks what it took off at targets one and two.

Hold to exit means never booking. You ride every position to its stop or its final target, nothing in between.

Under scale-out, 24.6% of trades close at a loss. Under hold-to-exit, that rises to 58.7%.

Read that again, because it's the finding: the majority of these trades touch profit and then give it all back. They are not losers in the sense of going straight down from entry. They are trades that were winning, and then weren't.

The 223 trades that decide everything

223 trades touched a profit target and subsequently reversed into their stop.

Under scale-out accounting, those are wins — partial ones, but real money banked. Under hold-to-exit, every one of them is a full loss, because nothing was taken off the table and price came all the way back through the entry to the stop.

223 out of 654 is more than a third of the book, and it sits exactly on the fence between the two numbers. That single population is why the average swings from positive to negative. Nothing else in the data comes close to mattering as much.

I find this the most useful thing in the entire dataset, and it has nothing to do with signal quality. It's a statement about what happens to a retail trader who is right about direction and has no exit discipline. They were right. They just never took anything.

Why I publish the losses

The signal industry runs on unaudited win-rate claims. 80%, 92%, 95%, plastered on landing pages with no methodology, no sample, no acknowledgement that a "win" under one convention is a full loss under another. The numbers are marketing, and everyone selling them knows it.

I said in the founding post that if RebateMax ever published a scoreboard it would show every outcome including losses, and that it wouldn't be marketed until it had enough history to be worth publishing. This is that scoreboard, and this is me keeping the harder half of that promise.

There's a version of this post that quotes only +0.46R and calls it a 74% win rate. It would be technically defensible under one convention and it would be dishonest, because I'd be picking the accounting that flatters the number and hiding the one that doesn't. Both are published at the same size on the scoreboard for exactly that reason.

I'm not asking anyone to be impressed by −0.08R. I'm asking you to notice that I published it.

What this actually means if you follow signals

Three things the data supports, and nothing more:

Your exit rule matters more than the signal. The same signals produce a positive or a negative book depending on whether you bank profit along the way. If you take one thing from this, take that.

A signal is not a plan. It gives you an entry and an invalidation. Everything between those two points is yours, and that's the part the numbers say is decisive.

Fees are the one certainty in any of this. Whether these signals make you money is genuinely uncertain and depends on decisions I can't make for you. What isn't uncertain is that every one of those 654 trades pays a fee at both ends. That's the part of this business I actually control, and it's why the whole operation is a fee rebate rather than a signal service.

What I'm not claiming

I'm not claiming these signals are good. Seven weeks and 654 trades is a real sample but not a long one, and it covers one market regime.

I'm not claiming the scale-out figure is what you'd have made — it assumes you took every signal, sized every one identically, and executed every partial. Nobody does that.

I'm not claiming this is an edge. It's a measurement of what came through the door, published because measuring it and hiding it would have been worse than not measuring it at all.

The scoreboard re-measures daily and these figures will move. If they go negative under both conventions, that will be published too, at the same size, on the same page.