Tail Drawdown

TL;DR. Tail Drawdown is the 95th-percentile peak-to-trough loss across bootstrap-resampled paths. Half of all alternate histories had a worst drawdown below the median; 5% had one bigger than the tail. Most traders quit during the gap between the median and the tail because they prepared only for the median.

What you'll be able to do

• Read a Tail Drawdown number and translate it into a "worst stretch I should be prepared for" figure.
• Understand why historical max drawdown chronically underestimates real risk.
• Use the metric to set position size before the live drawdown teaches you.

Why the 95th percentile

A drawdown distribution from ten thousand resampled paths produces a wide range of outcomes. Some paths barely have a drawdown — they trend cleanly upward. Other paths have terrible drawdowns — losing streaks compound, recovery takes forever. The right number to plan around depends on the question.

• Median drawdown is your average experience. About half your alternate-history selves had a drawdown smaller; about half had a drawdown larger.
• 95th-percentile drawdown is the lower bound of the worst-5% experience. Only 1 in 20 alternate histories had a worse drawdown than this.
• Maximum drawdown in the resampled distribution is the worst single observed, but it's a noisy estimate at fixed sample size — a sample of 10,000 paths produces a different max than a sample of 20,000.

The 95th percentile is the right tradeoff between informative and stable. It bounds 95% of plausible experiences without trying to estimate the outright worst case (which doesn't exist — distributions have no maximum).

Why the gap matters

The median and the tail are dramatically different in most strategies. A trend-follower might have a median drawdown of 6% and a tail of 18%. A mean-reverter might have a median of 3% and a tail of 22%. The gap between those numbers is what makes most live trading psychologically difficult.

Here's the dynamic: you start trading. The first six months produce a drawdown around the median, which you were prepared for. Live for a few more months and eventually one stretch hits the tail. You weren't prepared for the tail because the median was already close to your worst-case mental model. You quit, often within days of when the strategy was about to recover.

Tail Drawdown is the number that should set your mental model. Most traders set it from the median; the platform's job is to make the tail visible enough that you don't.

How it's computed

For each resampled path, the maximum peak-to-trough loss is recorded. Across ten thousand paths, the distribution of those maxima is summarized at the 95th percentile.

TailDrawdown = quantile(per_path_max_drawdown, 0.95)

Reported in both absolute dollars (drawdown from peak) and percent of starting balance, since both views matter for sizing decisions.

In AlgoLift

Tail Drawdown drives several visualizations:

• The Tail Risk Drawdown Envelope chart — the red shaded band is bounded by the 95th-percentile drawdown across paths at each trade index.
• The Tail Risk KPI strip — shows median and tail drawdowns side by side so the gap is visible immediately.
• The Tail Discipline component of the Robustness Score — feeds the score formula with TailDrawdown as one of three inputs.
• The mc-tail-risk insight chip — fires when TailDrawdown exceeds 25% of starting capital.

Reading the score

Tail Drawdown is best read as a percentage of starting balance rather than in dollars, because that's the right unit for psychological preparation.

The number to live by emotionally is the tail, not the median. If you can't sit through the tail, you can't deploy the strategy at this size — and the right move is to scale down until the tail is one you can.

Worked example

What this shows. At 0.5× the tail is comfortably below 10%. At 1× the tail is in the "typical" range — most traders can endure this if they've planned for it. At 2× the tail crosses 30%, which is where most traders give up. The strategy isn't different across the three sizes; the tail is. If you size for the median, you'll be unprepared for the tail. If you size for the tail, the median feels easy.

Common misunderstandings

• "Tail DD is the same as max DD across the resamples." No. The max across 10,000 paths is a single observation at the right edge of the distribution and has high variance — it would be different in the next 10,000 paths. The 95th percentile is much more stable because it averages across the top-5% region rather than picking the single outlier.
• "I should target 0% Tail DD." Not possible. Tail DD reflects the natural variance of the strategy's returns combined with autocorrelation in losses. Even strategies with massive positive expectancy have measurable tail drawdowns — that's what variance means.

When Tail Drawdown is the binding constraint

• Psychologically committed positions. When you're sizing to a level you can mentally endure rather than a level the math supports, the tail is the number to use.
• Strategies deployed at meaningful capital. A 25% tail drawdown on a $200k account is $50k. The dollar number is what makes the psychology real.
• Prop firm sizing. Most firms fail accounts at proportions of the starting balance; Tail DD relative to that proportion is the right sizing signal.

When it matters less

• Strategies with hard stop-losses below the tail. A strategy with a hard 8% stop-loss has a tail bounded by that stop — Tail DD won't exceed it. Read the stop-loss as the effective tail in that case.
• Long-only buy-and-hold equity at small size. When the position is a small fraction of capital and there's no leverage, tail drawdown is informational but rarely the binding sizing constraint.