The Gauntlet.
Thirteen checks between a good-looking backtest and a real edge. Most strategies don't make it through. That's the point.
Looked great, counted wrong. Overlapping trades made 40 results look like 400.
Lost to random. Random entries at the same times did just as well.
Edge smaller than costs. Profitable before commissions and fills, not after.
Rare, and reviewed. Survivors get a closer look with the community in Discord.
Thirteen ways a strategy dies.
Each check is built to catch one specific way backtests lie. They run in order, and a strategy that fails one doesn't get to keep going.
Rules frozen first
Entry, exit, data window and the pass bar are locked before any results are seen.
Data that fails loudly
Missing bars, bad timestamps and gaps stop the run instead of being quietly smoothed over.
No peeking ahead
Only information available at that moment is used, and dead or expired symbols stay in the test.
Enough history to start
Indicators get their full warm-up before the first trade, so early signals aren't built on half-formed data.
Honest trade count
Overlapping or clustered trades are counted for what they really are, not as independent wins.
Beat a matched control
The strategy has to outperform random entries taken at comparable times with comparable exposure.
Costs are a gate
Commissions, spread and the venue's real fill rule are applied before any verdict, not after.
Drawdown marked to market
Losses are measured while trades are open, not only at the exits that make the curve look smooth.
Tails before averages
The worst outcomes are examined before the average. A great mean can hide one trade that ends the account.
Sample-size honesty
Too few trades is reported as "not enough evidence," never rounded up to a pass.
More than one market
It has to hold up across different market conditions, not just the one period it was found in.
Same skepticism both ways
Good results get exactly the same scrutiny as bad ones. Wanting it to work isn't evidence.
No slicing after a fail
A failed test isn't rescued by hunting for the one subset, symbol or time window where it worked.
Bring a strategy. Leave with a verdict.
1. Spec it
Write down the rules, the market, the timeframe and what would count as a pass, before you test.
2. Run the Gauntlet
Every check runs in order. You see exactly which stage a strategy fails, and why.
3. Learn from the kill
A dead strategy still teaches you something. Refuted ideas go into the Strategy Library so nobody repeats them.
4. Survivors go to Discord
A strategy that clears all 13 gets reviewed with the community before anyone trades it.