Risk Management in Crypto Trading: Position Sizing and Stops
Sound crypto risk management means deriving position size from stop distance so every loss costs the same fraction of equity — typically 1% — placing the stop where the trade idea is proven wrong rather than at a round number, capping exposure per correlation cluster rather than per symbol, and setting account-level limits that halt trading before judgement degrades.
Risk management is usually presented as a list of maxims: cut losses, use a stop, don't over-leverage. Maxims are unfalsifiable. This article covers the specific arithmetic — how size is derived, why round-number stops are systematically worse, why eight correlated longs are one position, and what a losing streak from a genuinely profitable strategy actually looks like.
- Size is derived from the stop, never chosen independently:
size = (equity × risk%) / |entry − stop|. - Fixed notional sizing makes your equity curve a lottery on which setups happened to have wide stops.
- Eight long positions across correlated majors is one large long position with eight sets of fees.
- Leverage does not increase risk directly — the stop does. Leverage sets the liquidation distance.
- A 40%-win-rate strategy with real edge will still produce eight-loss streaks. Plan for them before they arrive.
Position sizing: the calculation everything else depends on
Most retail traders choose size first and place a stop afterwards. That is backwards, and it is the single most consequential habit to reverse.
The correct order: decide what fraction of equity you are willing to lose if wrong, identify where the trade idea is invalidated, then let the arithmetic produce the size.
risk_amount = equity × risk_fraction position_size = risk_amount / |entry_price − stop_price|
Worked example. Equity $10,000, risking 1% ($100). Entry $50,000, stop $49,000 — a $1,000 move. Size = $100 / $1,000 = 0.1 BTC, which is $5,000 of notional. At 10× leverage that requires $500 of margin. The size was derived; nothing was chosen except the risk fraction and where the idea fails.
Why fixed notional is broken
Trading a constant $1,000 per position sounds disciplined and is not. Consider two trades at the same notional:
| Trade | Notional | Stop distance | Loss if stopped |
|---|---|---|---|
| A — tight setup | $1,000 | 0.4% | $4 |
| B — wide setup | $1,000 | 3.0% | $30 |
Trade B carries seven and a half times the risk of trade A at identical size. Over a hundred trades your results are dominated by which setups happened to need wide stops — noise, not edge. It also makes expectancy in R meaningless, because your R is not a constant unit. See building a strategy that survives costs.
How much per trade?
One percent of equity is the conventional answer and a good default. The reasoning is survival arithmetic rather than tradition: at 1% risk a ten-loss streak costs roughly 10% of equity, which is unpleasant and recoverable. At 5% risk the same streak costs about 40%, which requires a 67% gain to recover from and is where people stop following their own rules.
Scale down, not up, when uncertain: a new strategy, a new market regime, or a period where you are trading worse than usual all argue for a smaller fraction until the evidence returns.
Where to place a stop
A stop marks the price at which your reason for the trade is no longer true. That is its entire job. It is not a comfort level, and it is not a round number.
Structural stops sit beyond the level whose break invalidates the setup — under the swing low for a long, above the swing high for a short. This is the most defensible method because the level means something to the market rather than to you.
Volatility stops scale with recent range, usually a multiple of average true range. The virtue is that the same rule adapts across regimes: in a quiet market it tightens, in a violent one it widens, so you are not stopped out by ordinary noise that happens to be larger this week.
Percentage stops are the weakest, because a fixed 2% has no relationship to the instrument's volatility or to any level. They are acceptable as a backstop and poor as a primary method.
The round-number trap
Stops cluster at obvious levels — round prices, the exact swing low, the figure. Clustered stops are a pool of resting liquidity, and price reaching into that pool before continuing in the original direction is one of the most reliably observed patterns in liquid markets. Placing your stop a little beyond the obvious level, rather than exactly on it, costs slightly more when wrong and avoids a large fraction of the exits that were never about your thesis being wrong.
Stops and cost
Tight stops are more expensive than they look, and the relationship is not linear:
cost_in_R = round_trip_fraction × entry_price / |entry_price − stop_price|
At 0.15% round trip, a 0.5% stop costs about 0.30R per trade while a 2% stop costs about 0.075R. Same fees, four times the cost in the unit that matters. Before tightening a stop to improve reward-to-risk on paper, check whether the strategy still clears its cost at the new distance.
Leverage: what it does and does not do
A persistent misconception is that leverage determines risk. It does not, directly. Your loss when stopped is position_size × stop_distance — leverage does not appear. Two positions with the same notional and the same stop lose the same amount at 3× and at 20×.
What leverage determines is margin required and liquidation distance. Higher leverage means less margin posted and a liquidation price much closer to entry. The danger is not that leverage magnifies your planned loss; it is that it moves the liquidation price inside the range where your stop lives, so a wick reaches the liquidation before it reaches your stop — and a liquidation is worse than a stop in both price and fees.
The practical rule: choose leverage so the liquidation price sits comfortably outside your stop, with room for a spike. If a 2% stop and your leverage put liquidation at 2.5%, you have no margin for a bad wick. How liquidation price is calculated.
Portfolio risk: correlation is the hidden multiplier
Per-trade risk management is necessary and insufficient. Ten positions at 1% each are ten percent of equity at risk only if they are independent — and in crypto they are not.
Major alts correlate strongly with BTC, and correlations rise sharply during stress, exactly when diversification is supposed to help. Eight long positions across eight majors during a sharp move down is one large long position that pays eight sets of fees.
Three controls:
- Cluster caps. Group instruments by realised correlation and cap total risk per cluster — not per symbol. A limit of 3–4% aggregate risk across a correlated cluster is a reasonable starting point.
- Directional caps. Limit total long or short exposure regardless of how many symbols it is spread across.
- Concurrency limits. A hard ceiling on simultaneous open positions, because attention and margin are both finite.
Drawdown gradients
Rather than a single hard stop at some drawdown, reduce size progressively as drawdown deepens. Full size to −5%, three quarters to −10%, half beyond that. The reasoning is behavioural as much as statistical: a deep drawdown is both a period where the market may be hostile to your strategy and a period where your judgement is measurably worse, and cutting size addresses both without requiring you to decide anything in the moment.
The drawdown a real edge produces
A profitable strategy with a 40% win rate has a 60% chance of losing on any given trade. Runs happen. Approximate odds of a losing streak within a hundred trades at that win rate:
| Streak length | Roughly how likely within 100 trades |
|---|---|
| 5 losses | Near certain |
| 8 losses | Very likely |
| 10 losses | Perfectly plausible |
None of that is malfunction. It is what a 60% loss rate does. The failure mode is not the streak, it is what the trader does during it: abandoning the strategy, doubling size to recover, or overriding rules that were correct.
Which is why account-level circuit breakers matter more than any per-trade rule. A daily loss cap, a consecutive-loss pause, a limit on how quickly a new position can follow a loss — these do not improve the strategy. They prevent you from destroying it during a statistically ordinary bad run. Anti-tilt makes the fuller case.
Protecting the position mechanically
Every risk rule above is theoretical until an order exists on the exchange. Three mechanical requirements:
- The stop is placed in the same operation as the entry. Not afterwards, not when you get to it. If the entry fills and the stop is rejected, you hold leverage with no floor.
- Something re-checks that the stop is still live. Orders get rejected for precision, for notional minimums, for transient errors. The process that verifies must be independent of the one that placed it, because the most common reason a stop is missing is that the placing process died.
- Position and orders are reconciled against the exchange, not against memory. After any restart or disconnect, exchange state is the truth and your internal state is a hypothesis.
TradeFloor implements all three: brackets are placed atomically with entry, a healing process reconciles open positions against live protective orders continuously and re-places anything missing, and liquidation monitoring escalates through four levels of which only the last acts, and only on wallets you explicitly arm.
Frequently asked questions
How much should I risk per trade in crypto?
One percent of account equity per trade is the standard default and a good starting point. At 1% risk a ten-loss streak costs about 10% of equity — recoverable. At 5% risk the same streak costs roughly 40%, which needs a 67% gain to recover from and is where most traders abandon their rules. Scale down for new strategies or unfamiliar regimes, not up.
How do I calculate position size from a stop loss?
Divide the amount you are willing to lose by the distance to your stop: position_size = (equity × risk_fraction) / |entry − stop|. With $10,000 equity, 1% risk and a $1,000 stop distance on BTC, that is 0.1 BTC. Size is derived from the stop, never chosen independently of it.
Does higher leverage mean higher risk?
Not directly. Your loss when stopped is position size times stop distance, and leverage does not appear in that. What leverage changes is the margin required and how close the liquidation price sits to entry. The real danger is leverage high enough that liquidation falls inside your stop distance, so a wick liquidates you before your stop fires.
Where should I put my stop loss?
Where the trade idea is proven wrong — beyond the structural level whose break invalidates the setup, or at a volatility-scaled distance such as a multiple of ATR. Avoid exact round numbers and the precise swing level, since stops cluster there and price frequently reaches into that liquidity before continuing. Then check the strategy still clears its cost at that stop distance.
How many positions should I have open at once?
Fewer than the number that makes the aggregate risk unknowable. The binding constraint is correlation, not count: eight longs across correlated majors is effectively one large long. Cap total risk per correlation cluster at roughly 3–4% of equity rather than capping per symbol.
What is a normal drawdown for a profitable strategy?
Deeper and longer than most people expect. A genuinely profitable strategy with a 40% win rate will very likely produce an eight-loss streak within a hundred trades, and a ten-loss streak is plausible. Know your expected worst streak before starting, so an ordinary bad run does not read as a broken strategy.
Liquidation in Crypto: What It Is and How to Avoid It
How liquidation price is calculated, what actually triggers it, and the four-level escalation that prevents it.
// questions or corrections · [email protected] · more essays · /blog