Portfolio drawdown is not the sum of individual max drawdowns. Strategies overlap in risk factors, liquidity stress, and regime exposure. If you size each bot to its solo backtest max DD, you will usually overshoot real portfolio pain.

This article proposes a simple drawdown budgeting model you can implement without fancy math.

Step 1: define the portfolio risk envelope

Pick hard numbers you can defend when things break:

maximum acceptable portfolio drawdown (peak to trough)
maximum acceptable time underwater
maximum acceptable daily loss

These are owner constraints, not optimizer outputs.

Step 2: assign strategy budgets (not capital budgets)

Convert each strategy into a drawdown budget in dollars or percent of portfolio.

Example policy:

no single strategy may consume more than 40% of the portfolio DD budget
correlated strategies share a bucket budget (trend bots vs mean reversion bots)

Correlation does not need perfect estimation. Even coarse bucketing beats ignoring overlap.

Step 3: tie budgets to kill-switch actions

Each strategy should have pre-agreed actions at budget milestones:

50% of budget: reduce size, widen safety checks, pause new entries
80% of budget: flatten, freeze research changes, post-mortem required
100% of budget: stop, no exceptions

If you wait for emotions, you will overshoot.

Step 4: rebalance budgets monthly

Live performance changes correlations and volatility. Revisit:

realized covariance of returns
worst joint stress weeks
whether one strategy is dominating tail risk

Liquidation and margin are portfolio events

If any bot can margin-call the account, your drawdown budget is not per-strategy anymore.

Treat leverage and cross-margin as a portfolio constraint that can override individual bot budgets.

Stress the joint tail, not only averages

Monthly correlation can look tame while joint crash weeks dominate losses.

Keep a small library of replay weeks (liquidity shocks, funding spikes) and ask whether multiple bots would hit budgets together.

Communication: one owner for the envelope

If two operators each believe they own "50% of risk," you will exceed the envelope in a crisis.

Write down who can pause strategies, who can resize, and who can restart after a breach.