ALGOINDICATEUR
Login Create an account
🤖 Algo AI Trader 🧠 AI Assistant 📡 Algo Signals VIP 💹 Trading

Platform

📈 Markets ⭐ Watchlist 🌍 Market movers

Tools

🧮 Trader calculators 🔄 Converter 📅 Economic calendar 📖 Trading glossary 🎓 Trader academy 📓 Trade journal 🧰 Advanced calculators 📊 Chart patterns ✨ App features

Resources

📰 News ❓ FAQs ℹ️ About 📨 Telegram signals
Create an account Login Become VIP
Trader academy

Searching for a strategy without fooling yourself

Algo Signals’ lab: fees, 70/30 validation, blind test, chance test, variants… and why a “good” strategy found among thousands is often just luck.

Lesson 96 / 96 · 10 min read · Advanced

If you test enough combinations on the same history, one of them will inevitably look excellent, purely by chance. That is the central trap of strategy searching. Algo Signals’ lab exists to protect you from it as far as possible, not to promise a profit.

What the lab does

  • It tries indicators alone, in pairs and, if you tick it, in triplets, with the TP / SL methods you choose.
  • It replays each combination on the loaded history (the market and timeframe displayed) and ranks the results by a criterion: expectancy in R per trade, profit factor, win rate or total gain.
  • “R” is the multiple of the risk taken: winning 1 R means winning as much as the initial risk.
  • A minimum number of trades (30 by default) rules out strategies that almost never traded: too few trades prove nothing.

The safeguards

  • Fees: a percentage (0.08% by default) is deducted from every trade. The spread is not included.
  • The 70 / 30 validation: the strategy is learned on the first 70% of the history, then recomputed on the next 30% that it has never seen. The ranking keeps the worse of the two results.
  • The blind test: the most recent 20% are set aside during the whole search, then used as a last check.
  • The chance test: the lab replays the whole sorting with randomly drawn entries and measures the share of trials that match your best result.

How to read the result

A strategy is marked “Validated” only if it stays positive on the unseen part and on the blind part, with enough trades. The chance test gives a percentage: below 5%, the result is unlikely to be due to chance; between 5 and 20%, it is uncertain; above that, it is very probably chance, because with so many strategies tested such a result appears on its own. “Validated” does not mean “winning tomorrow”.

Setting variants: more choice, more risk of fooling yourself

  • The “variants” option also tests wider and narrower stops and targets. It multiplies the combinations.
  • The more combinations, the more likely one looks good by pure luck: variants increase the risk of fitting the result to the history, hence the value of the chance test.
  • No indicator is magic: all are computed from the same past prices. What matters is a result that holds on never-seen data, on several markets, with fees.

📝 Lesson test

Log in to take this lesson’s test. Once you have passed all the tests, you will receive a certificate in your name.

Log in

These lessons are for understanding and are not investment advice. No method guarantees winning.