Managing a live algo strategy
how to read the signal. not the noise.
Running a live algo strategy means making real decisions: what to monitor, how to read a drawdown, and when to act. Done well, it's less about reacting to what happens and more about knowing in advance what to expect — and what to do when you see it.
What should you monitor when running a live algo strategy?
The most common mistake when managing a live strategy: checking it too often. For longer-timeframe strategies, daily P&L is noise. A strategy that rebalances monthly will have dozens of days where it looks wrong before it looks right. The right cadence for review depends on the strategy's timeframe — an end-of-day strategy might warrant a weekly check-in; an intraday strategy warrants more frequent attention.
Compare live performance against the track record — not just the return, but the drawdown profile and trade frequency. Is the strategy behaving the way the track record said it would? Small differences are normal. Large, persistent gaps warrant investigation.
Watch for changes in market regime. A momentum strategy will underperform in a choppy, mean-reverting market. That's not strategy failure — it's expected behavior. Understanding why a strategy is underperforming matters more than reacting to the fact that it is.
Market regimes are the context in which a strategy operates. A regime is a period of broadly similar market behavior — trending, mean-reverting, high-volatility, low-volatility. Most strategies are designed for specific regime conditions, even if that's not explicit in how they're described. A momentum strategy is implicitly designed for trending regimes. A mean reversion strategy is implicitly designed for range-bound ones.
Regime shifts are normal and expected. What changes when a regime shifts is which strategies are likely to perform and which are likely to struggle — not whether any strategy is fundamentally broken. The key question is: is this strategy underperforming because conditions are temporarily unfavorable, or because its underlying edge has decayed? See types of algo trading strategies for a breakdown of which types perform best in which conditions. Comparing a strategy's behavior against the prevailing market regime — whether the period has been trending, choppy, or high-volatility — can help distinguish regime-driven underperformance from strategy decay.
How do you tell a normal drawdown from strategy decay?
Most traders react to drawdowns. Knowing what a normal drawdown looks like — and having the framework to read it — is what separates managing a strategy from just watching it.
Every algorithmic trading strategy draws down at some point. The track record tells you the historical max drawdown — but in live markets, that number can be exceeded. The question is whether the drawdown is within the expected character of the strategy, or whether something has fundamentally changed.
Time is the best diagnostic. A strategy drawing down in conditions that have historically caused drawdowns is not necessarily in decay. A strategy that underperforms persistently across varied market conditions — including periods where it historically performed well — warrants a harder look. The relevant timeframe depends on the strategy's own cadence: a two-week drawdown means something very different for a monthly-rebalancing strategy than for an intraday one.
When should you stop an algo trading strategy?
There's no formula for knowing when to stop an algorithmic trading strategy — but there are clear signals. Stop when the strategy's live behavior diverges materially from its track record: different trade frequency, drawdowns that exceed the historical profile, or persistent underperformance across conditions where it historically performed well. The principles below give the framework.
Knowing when to stop is the hardest call. Knowing when to hold — because the strategy is behaving exactly as its track record said it would — is the skill that compounds over time.
When and how should you add a second algo strategy?
Running multiple uncorrelated strategies is one of the most effective ways to manage portfolio-level risk. When two strategies are uncorrelated — meaning they don't tend to lose at the same time — a drawdown in one may be offset by stable or positive performance in the other. The result is a smoother overall equity curve without necessarily sacrificing return.
Check correlation before adding anything. Two momentum strategies targeting the same market will tend to lose at the same time — that's concentration, not diversification. Momentum strategies across different asset classes or timeframes may be less correlated. When evaluating a second strategy, apply the same framework you used for your first — see how to evaluate an algo trading strategy for the metrics that matter.
The second strategy is easier than the first. You have a baseline, you understand your own risk tolerance, and you've seen what a real drawdown feels like. That's experience most traders don't have — and it makes every subsequent decision sharper.