Survival Probability

TL;DR. Survival Probability is the conditional probability that a strategy ends profitable given its longest losing streak was at least N trades. The curve tells you the empirical bail-out point: the streak length past which even sims that lived through it weren't more likely than not to recover.

What you'll be able to do

• Read the survival probability curve on the Streaks panel and identify the coin-flip threshold.
• Use the curve to set a pre-committed bail-out rule for live trading.
• Explain to a teammate why "if you're in a 12-trade losing streak, sims with that streak ended profitable 60% of the time" is a stronger frame than "the strategy is good."

The inversion

Most risk metrics ask: how likely is a bad outcome? Survival probability asks the inverse: given a bad outcome, how likely is recovery?

That inversion matters because the bad outcome — a long losing streak — happens during live trading whether you planned for it or not. The question isn't whether the streak will happen. It's what you should do when it does. Quit, or hold on?

The survival curve answers that empirically. Look at every resampled sim that experienced a losing streak of N trades. What fraction of them ended the simulation profitable? That number, plotted against N, is the survival curve.

When you're 12 trades into a losing run and considering quitting, the curve at N=12 tells you the empirical recovery rate for sims that lived through exactly your situation. If it's 75%, your strategy is much more likely to recover than not. If it's 30%, your situation is worse than the average sim that quit at this point.

How it's computed

The Robustness Lambda records the longest losing streak for every bootstrap-resampled sim and the sim's total return. For each bin of streak length, the platform computes the fraction of sims in that bin that ended with positive total return.

SurvivalProbability(N) = P(total_return > 0 | longest_losing_streak ≥ N)

Bins with fewer than 5 sims are dropped to keep the right tail of the curve from being dominated by noise. The bin width is chosen automatically based on the spread of the streak distribution.

In AlgoLift

Survival Probability is the headline chart on the Streaks sub-tab:

• The Survival Probability curve — cyan line plotting P(profitable | longest streak ≥ N). Filled area below the line reinforces the probability reading.
• The 50% reference line — amber dashed line marking the coin-flip threshold.
• The original-streak marker — vertical white dashed line at the user's actual backtest's longest losing streak, so the user can see where they sit on the curve.
• The coin-flip annotation — small text callout showing the streak length where survival crosses 50%, if it does within the observed range.

Reading the curve

The curve produces three distinct reads.

The slope

A flat curve means the strategy is robust across streak lengths — losing streaks don't carry much signal about the strategy's eventual outcome. A steep curve means streak length is a strong predictor of failure.

The crossover

The streak length at which the curve crosses 50% is the empirical bail-out point. For trend-following strategies this is typically 20-30 trades; for mean-reverters it's often 12-18. Below the crossover, the strategy is more likely to recover than not. Above it, your expected outcome is worse than random.

Your position on the curve

Your original backtest's longest losing streak is marked on the curve. If you're at the 85% portion (most of the curve is to your right), the strategy is structurally healthy — even much-longer streaks usually recover. If you're at the 55% portion, you're already close to the coin-flip; a slightly longer streak is psychologically possible and statistically dangerous.

Worked example

What this shows. Strategy A has a healthy survival profile — the coin-flip threshold is at 25 trades, well past the original's 12-trade streak. Live trading might hit longer streaks, but the curve is shallow enough that even substantial extensions stay on the "more likely to recover than fail" side. Strategy B's coin-flip is at 11 trades. The original backtest had an 8-trade streak. Live trading is highly likely to experience streaks longer than 8 — and once those streaks cross 11, the strategy is empirically more likely to fail than recover. The two strategies might have similar headline Sharpes, but their survival profiles tell very different deployment stories.

Common misunderstandings

• "The curve should be monotonic — strictly decreasing." Usually but not always. Very small sample sizes in some bins can produce non-monotonic jitter at the right tail; the platform drops bins below 5 sims to mitigate this, but small jitter is normal.
• "My coin-flip is at 30 trades — that means I should never quit before 30." The curve is a statistical aggregate. It doesn't account for the specific reason you're in a losing streak. If your strategy is in a streak because of a regime change or a structural break, the survival probability for that specific cause is lower than the aggregate. Use the curve as a starting point, not the only signal.

When the survival curve is most informative

• Strategies with infrequent but deep losing streaks. The curve gives a concrete bail-out threshold instead of vague "stop trading when uncomfortable" advice.
• Pre-deployment commitment exercises. Read the curve before you go live and write down your bail-out streak length — it's much easier to commit cold than during the streak.

When it's less informative

• Strategies with very short backtests. Below ~500 trades, the longest-streak distribution is sparse enough that the curve has wide confidence intervals.
• Strategies with hard stop-losses that prevent long streaks structurally. If your stop-loss caps streak lengths at 5 trades, the curve past 5 is undefined because no sims reached that point.