Calculator
Drawdown-Recovery Markov Simulator
Time to recover from an N% drawdown given monthly Sharpe + skew + kurtosis. Cornish-Fisher Monte Carlo, percentile distribution of recovery months.
- Inputs
- Form inputs / CSV
- Runtime
- Instant
- Privacy
- Client-side · no upload
- API key
- Not required
- Methodology
- Open →
Inputs
Median recovery (months)
10
From a 20% drawdown · monthly Sharpe 0.60 · skew -0.30 · 2,000 paths.
Recovery distribution
25th pct
7
best 25%
Median
10
75th pct
13
slow paths
95th pct
21
tail risk
Path outcomes
Recovered within 12mo
69.3%
1,386 of 2,000
Never recovered (240mo cap)
0.00%
Reading the result
Time-to-recover scales roughly as ln(1 / (1 − DD)) / monthly_excess_return. Negative skew + fat tails extend the right tail materially — set the kurtosis slider to 5+ and watch the 95th percentile move. See methodology.
How to use
Step-by-step
- 1
Upload your equity curve (daily values or trade-by-trade P&L).
- 2
Set the drawdown tier boundaries (default: 0-5%, 5-10%, 10-20%, 20%+). Tighter tiers give more granular transition probabilities but need more data.
- 3
Run the model. Read expected recovery time per drawdown tier, plus 95th-percentile worst-case recovery.
- 4
Compare against your strategy's max acceptable drawdown duration. If 95th-percentile recovery exceeds your patience, the strategy is mis-sized for your psychology.
- 5
Re-run on subset windows to test stability. Recovery times that change dramatically across regimes signal regime sensitivity in the underlying strategy.
Glossary references
Terms used by this tool
Questions people ask next
FAQ
Why use a Markov chain to model drawdowns?
Drawdowns aren't memoryless — the deeper the drawdown, the longer the typical recovery, with a non-linear relationship. Markov chains capture the state-conditional transition probabilities cleanly. The tool's chain has states for {flat, drawdown depth tier 1-N}, with transition probabilities estimated from your equity curve.
What's the recovery time output?
Given current drawdown state, the expected time (in trade days) until you return to a new high. The tool reports both mean recovery time and the 95th-percentile worst case. For shallow drawdowns (1-5%), expected recovery is short; for 20%+ drawdowns, the tail extends a long way.
How many observations do I need?
At least 250 trade days for stable transition probabilities. Below that, the drawdown-tier transitions are too noisy. The tool flags low-N estimates with a warning.
Does it work for trend-following systems?
Trend-followers have characteristic long, deep drawdowns followed by sharp recoveries. The Markov model captures this if your sample includes both regimes. If your equity curve is all bull market, the model will under-predict drawdown depth — sample selection matters.
What's the difference between this and Monte Carlo on returns?
Monte Carlo treats returns as i.i.d. from a fixed distribution. The Markov approach lets transition probabilities depend on current drawdown depth, which is empirically how strategies actually behave (drawdowns cluster). The Markov model is more conservative about recovery times in deep drawdowns.
Related deep dive
All articles →Read further
Long-form context behind the tool output.
- Tutorial · Runnable·9 min
How to Read a Backtest Report: 2026 Cheat Sheet
Five questions a backtest report must answer — edge real, persistent, cheap to trade, bearable, explainable — with the statistics that verify each.
Read - Pillar · Guide·10 min
Retail PnL vs Backtest
The eleven-point gap between a 14% backtest and 3.2% live PnL, decomposed across eight mechanisms with dollar examples — slippage, latency, fees, fills.
Read
Complementary tools
Users of this tool often explore
Fractional Kelly Sizer
Map conviction tiers to fractional Kelly bet sizes with a drawdown Monte Carlo simulator. Client-side. Private by default.
Position Sizing under Edge Variance
Bayesian-Kelly bet sizing when your edge is itself uncertain. Compare deterministic Kelly, Bayesian-adjusted, and conservative lower-bound versions.
Returns Distribution Analyzer
Paste a returns CSV. Histogram, normal-overlay, QQ plot, skewness, excess kurtosis, Jarque-Bera test, tail-weight index. See why Sharpe alone misleads.