Risk management for crypto futures: what a systematic book does

The arithmetic that sets the rules

Losses and gains are not symmetric. A drawdown must be recovered from a smaller base:

Bitcoin held outright fell 77% from its 2021 peak; it needed a 335% rally to get back, and the round trip from the November 2021 peak to a new high took 28 months. A strategy that caps its drawdowns near 20% needs a 25% recovery, which a trend book can produce in one good quarter. Every rule below exists to keep the account in the top half of that table.

1. Size positions by volatility, not conviction

The single most effective control. Each position is sized so that its expected daily swing is the same fraction of equity: a coin that moves 5% a day gets half the size of one that moves 2.5%. The book as a whole targets about 20% annualised volatility. When markets get violent, positions shrink automatically; when they calm, positions grow. Over the 2021–26 backtest this kept gross exposure at 0.30× equity on average and never above 1.02×, although the venue allowed 5×.

The alternative, sizing by how confident you feel, is how retail accounts end up with their largest position in their most volatile asset at the worst moment.

2. Cap concentration

No asset above 25% of equity, whatever the signal says. This bounds the damage of a single stop-loss to roughly 6% of the account. The cap was not free: adding it to the backtest cost about five points of annual return. It stays because the alternative, at one point in testing, was 101% of equity in one coin.

3. Stops on the exchange, not in the software

Every position has a stop order resting on the venue from the moment it opens. If the bot, the server or the connection fails, the stop still fires. A stop that lives only in your own code is a stop that disappears with your process.

Stops are placed at a distance the strategy chooses, 25% from entry for this book, far from any liquidation level and wide enough not to be hit by normal noise. Tighter trailing stops were tested and lost money: they fire on noise and re-enter at worse prices.

4. Limits that remove risk without judgement

• Daily loss limit. After a set loss within one day, no new positions until the next day. Existing exits still run.
• Drawdown circuit breaker. If equity falls a set distance below its peak, all position targets are halved until a human reviews. This is the one rule that requires a person, and it requires them only to look, not to be brave.
• Cooldown after a stop. An asset that just hit its stop cannot be re-entered for a day. Trend signals often persist through a stop; re-entering immediately is how one loss becomes three.
• Funding and liquidity filters. No new exposure on the side paying extreme funding, and no position larger than a small fraction of daily volume.

All four share a property: they only ever block new risk. Exits and stops always execute. A risk rule that can prevent an exit is a liquidation waiting to happen.

5. Measure the leaks

Fees, slippage and funding are risks too, of the slow kind. Every fill in our book is logged with its slippage against the price seen at decision time, and the weekly report shows the gap between what the strategy intended and what the account made. That logging caught 9–22 basis points leaking on every exit within two days of going live (Note 07).

What the rules produced

Backtest, January 2021 to September 2026, net of costs and funding: 41.8% a year, worst drawdown −18.8%, twelve drawdowns deeper than 8%, none deeper than 19%. Note 03 lists every one. None of the rules predicts anything. They decide what happens when the prediction is wrong, which is often.