Why almost every trading signal fails
Anyone can find a rule that would have won in the past. The past always lets itself be explained: you try a thousand combinations and one fits by pure chance. That is not a signal, it is memorizing the exam. The problem is that when that rule meets the future -data it has never seen- it stops working. That is called overfitting, and it is why almost every signal fails.
Overfitting, in one sentence
If you test enough strategies, you will find one that looks brilliant purely by luck. With 1,000 attempts, a handful beat the market without having any real edge. The seller shows you that one, hides the other 990, and you see a “winner” that is really noise in disguise.
The test that actually matters: walk-forward
The only honest way to know whether a strategy has an edge is to test it on data it did not use to build itself. It is called walk-forward: you train on one stretch of time, test on the next, move forward and repeat. If it only wins on the stretch where it was designed (in-sample) but falls apart on the ones that follow (out-of-sample), it is overfitting. If it holds up stretch after stretch, it starts to smell like a real edge.
Without costs, it is all fiction
A signal that does not account for fees and slippage does not exist in the real world. The price you actually get in and out at is not the one on the chart. Any result you see has to include costs; otherwise, they are showing you a fantasy backtest.
Winning big is not winning often
Winrate -the % of winning trades- is the favorite metric of the ads and tells you almost nothing. You can be right 43% of the time and have an edge, if your winners are bigger than your losers. And you can be right 80% and go broke, if the occasional loser eats everything. What matters is expectancy (how much you make on average per trade), not how often you are right.
How to tell if a signal works
Before paying for anything, demand three things: that they show walk-forward, out-of-sample results with costs, that they also publish the losing trades (hiding them is the oldest lie), and that they do not promise you returns. Anyone promising you a guaranteed % either does not understand risk or is lying to you.
That is exactly what we do: the track record is published in full, walk-forward and real forward, with the losers included and the reason why. No promises.