Guides / Losing Streak or Broken Strategy
GUIDELosing streak, or is the edge actually gone?
Every drawdown feels like proof the strategy stopped working while you're inside it. Here's a calmer, backtest-relative way to tell ordinary variance from an actual break.
Quick answer: compare the live losing streak's length and depth against what the strategy's own backtest has already produced. If it's inside the backtest's historical worst case, it's variance the strategy was always going to hit eventually. If it's clearly beyond anything the backtest ever showed, that's the signal worth investigating.
Why this is so hard to judge in the moment
During a drawdown, everything feels like evidence the edge is gone — recent losses are vivid, recent wins are forgotten, and the temptation to "do something" (cut size, pause the strategy, tweak a parameter) is strongest exactly when you have the least objective information to act on. That's recency bias, and it's the reason this question needs an answer set in advance, not one decided mid-drawdown.
The check that actually answers it
Every backtest already contains a worst historical losing streak and a worst historical drawdown — the strategy has already "survived" those, by definition, since they're in the sample it was built from. If live results stay inside that historical range, the strategy is behaving exactly as its own history said it eventually would. The question isn't "is this losing streak normal in general" — it's "is this losing streak normal for this specific strategy," and the backtest already has that answer.
A worked example
Say a strategy's full backtest contains a worst streak of 9 consecutive losses and a 12% max drawdown.
- A live streak of 7 losses and a 9% drawdown is inside that range — uncomfortable, but exactly the kind of stretch the backtest said would happen sooner or later.
- A live streak of 14 losses and a 19% drawdown is outside anything the backtest ever produced — that's the point where "is this normal" genuinely deserves a "no."
What else is worth checking
- Has something structural changed — a broker switch, a spread widening, a shift in the instrument's typical volatility — rather than just an unlucky run?
- Has the live win rate over a meaningful recent sample diverged clearly from the backtested win rate, rather than just a short rough patch?
Where traders get this wrong
- Stopping a strategy on the first losing streak that's still well inside its own historical range.
- Refusing to stop a strategy even after live results have clearly exceeded the backtest's documented worst case.
- Changing several things at once — parameters, timeframe, and pairs together — so there's no way to tell afterward what actually mattered.
What to actually do about it
Set the "is this normal" threshold before you're in a drawdown, directly from the backtest's own worst-case numbers — not from a gut feeling formed mid-drawdown. Then check live performance against that threshold automatically, rather than reopening a backtest report and trying to read it calmly during the exact stretch when you're least able to.
Watchdog compares your live account against your own backtest continuously, so you don't have to make that call while it's happening.
Open Watchdog