"Edge" is the most overused word in retail trading. Every signal-seller has it. Every course has it. Every influencer has it. None of them mean the same thing by it, and almost none mean what a statistician means by it.
Here is what a statistical edge actually is, in three sentences :
An edge is a setup that, across a large enough sample, produces a positive expected outcome that you can demonstrate did not happen by chance. The size of the sample matters. The conditions under which it was tested matter. The way you test for chance matters most.
If your "edge" satisfies none of those three conditions, what you have is a hypothesis. There is nothing wrong with hypotheses, but they are not the thing you bet your account on.
Expectancy, the unit of edge
Expectancy is a single number. It is the average outcome of a trade, in whatever unit you measure (R, pips, dollars). Formula :
expectancy = (win_rate × avg_win) − (loss_rate × avg_loss)
A setup with a 40% win rate and a 3R average winner against a 1R average loser has positive expectancy : 0.4 × 3 - 0.6 × 1 = +0.6R per trade. A setup with a 70% win rate and a 0.5R average winner against a 2R average loser has negative expectancy : 0.7 × 0.5 - 0.3 × 2 = -0.25R per trade.
The 70% win-rate setup feels better. It loses money. The 40% win-rate setup feels frustrating. It makes money. Expectancy is the only measurement that ranks setups correctly.
If your journal shows you win rate without expectancy, your journal is lying to you. Not maliciously. Just incompletely.
Sample size, the unit of trust
Now the hard part. A positive expectancy across ten trades means nothing. Across fifty trades, very little. Across one hundred and fifty trades on the same setup in similar conditions, you have something worth thinking about. Across five hundred, you have something worth betting on.
Michael Mauboussin put the test for it about as plainly as it can be put :
When luck plays a part in determining the consequences of your actions, you don't want to study success to learn what strategy was used but rather study strategy to see whether it consistently led to success.Michael Mauboussin, The Success Equation
This is why most trader self-assessments are wrong : they are based on samples too small to mean anything. Your last twenty Friday afternoon scalps that all worked do not constitute an edge. They constitute a Friday afternoon.
Significance, the unit of "did this happen by chance"
The third condition is the one that retail trading almost never talks about. Even if you have a positive expectancy across one hundred and fifty trades, the question remains : how likely is it that you would have seen this result if your setup had no edge at all and just got lucky ?
This is what a p-value answers. A p-value of 0.05 means : if your setup had zero edge, the probability of seeing a result this good or better by chance is 5%. A p-value of 0.01 means 1%. A p-value of 0.20 means 20%, which is too high to act on.
An edge that has not been tested for chance is a story you are telling yourself.
You do not need to do this math by hand. You need a system that does it for you and refuses to call something an "edge" until it has passed the test. Otherwise you are betting on what feels like edge, which is the same thing every losing trader is doing.
The conditional edge
Here is the final layer. Most edges are not absolute ; they are conditional. The setup that wins 60% in trending markets might win 40% in chop. The entry that is positive expectancy in London session might be negative expectancy in late New York. The instrument you trade well in low volatility might bleed you in high volatility.
An honest edge analysis surfaces the conditions under which the edge holds, and the conditions under which it breaks. EdgeFound does this automatically : it segments your trades by every variable you have tagged (setup, session, day, instrument, risk size, mistake type, conviction level) and reports the conditional expectancy for each combination that has a sample large enough to matter.
The output is a list of sentences. Sentences like :
- "Your MMXM setup has +0.9R expectancy on Tuesday, Wednesday, Thursday. On Friday it is -0.3R. Sample : 84 trades. p < 0.01."
- "Your XAUUSD trades during US session have +0.5R expectancy. During Asia session, -0.4R. Sample : 67 trades. p = 0.03."
- "Your ‘3R+ planned’ trades have +1.1R realised expectancy. Your ‘1R planned’ trades have -0.2R realised. You are oversizing your bad ideas."
None of these are visible in a win-rate display. All of them change how you trade tomorrow.
What this means for you today
Pick your three most-traded setups. Ask yourself the following questions, honestly :
- Do you know the expectancy of each, in R ?
- Do you know the sample size you computed it on ?
- Do you know whether the result is statistically significant ?
- Do you know the conditions under which the edge holds vs breaks ?
If you cannot answer all four for all three setups, you do not yet have an edge. You have a hypothesis you are funding with real money.
This is solvable. It is what EdgeFound exists to solve. The math has been settled for a century ; it just needs to be run on your trades, automatically, and reported to you in plain English. That is the entire product.