Learn / Risk & Position Sizing
Risk per trade vs risk per day
Risk per trade bounds one decision. Risk per day bounds a sequence of them. Having only the first is the most common gap in an otherwise disciplined process, because it feels like a complete answer.
What a per-trade limit leaves open
Risking 1% per trade sounds like a bounded exposure. It bounds nothing at the day level, because the number of trades is unconstrained.
Take ten trades, lose all ten:
(1 − 0.01)^10 = 0.904 → −9.6%Every trade obeyed the rule. The day cost nearly 10%, which needs 10.6% to recover.
Twenty trades:
(1 − 0.01)^20 = 0.818 → −18.2%, needing +22.3%Still 1% per trade. The per-trade limit was never violated and the account had a very bad day — see the drawdown recovery calculator.
The trade count is the hidden variable
Nothing in per-trade sizing constrains how many trades you take, and the number is not stable. It rises exactly when it should fall:
- After losses, from the impulse to recover.
- In volatile conditions, because more setups appear.
- When bored, which is a real and undiscussed input.
- When a model is generating candidates, because supply is no longer limited by your attention.
That last one deserves emphasis in an AI-assisted setup. A human scanning charts produces a few ideas an hour. A model produces as many as you ask for, and per-trade limits do not scale down as candidate volume scales up.
What each limit does
| Limit | Bounds | Fails when |
|---|---|---|
| Per trade | One decision | Many decisions |
| Per day | A sequence | Multi-day drift |
| Per week / month | Drift across days | Structural problems |
| Position count | Concurrent exposure | Sequential trades |
| Correlation group | Hidden concentration | Correlation changes |
They are not alternatives. Each closes a gap the others leave open, and the per-day limit is the one most often missing.
Setting the daily limit
Work backwards from recoverability rather than forwards from tolerance:
- Take your risk per trade.
- Take your expected longest losing streak for your win rate.
- Compute what that costs:
(1 − f)^k. - Ask whether the recovery from that number is one you would actually attempt.
A limit above the cost of a normal losing streak never fires and is decoration. A limit below the cost of two or three ordinary losses fires constantly and trains you to disable it.
Between those is a number that binds only on genuinely bad days. Design details — what counts, when the day starts, what happens when it trips — are in designing a daily loss limit.
Why a trade-count limit is the crude version
“No more than five trades a day” is unsophisticated and effective, because it attacks the hidden variable directly.
It is worse than a loss limit in one way — five winning trades are stopped along with five losing ones. It is better in another: it binds before the losses, not after. A loss limit stops you having already lost the limit; a count limit stops you having taken the trades.
For revenge trading specifically, the count limit is the one that works, because the problem is the trades rather than the losses.
The interaction people miss
A per-trade limit and a per-day limit do not compose the way intuition suggests.
With 1% per trade and a 3% daily limit, you might expect three trades’ worth of room. But fixed-fractional sizing shrinks each subsequent loss, so it takes slightly more than three losses to reach 3% — and meanwhile winning trades increase the balance, raising the currency value of the next 1%.
The practical consequence: on a mixed day, the daily limit is reached by more trades than you would guess, which gives more room for a bad sequence than intended. If you want a hard stop after three losses, count them rather than relying on the percentage to get there.
For automated and AI-assisted setups
Both limits must be evaluated before submission, as functions that can return false. A limit the model knows about is an input it weighs; a limit that rejects the order is a limit — see why MCP needs an approval layer.
Add one more, specific to model-generated candidates: a cap on proposals per session, not just on executed trades. Otherwise the approval step becomes a volume problem, and volume is how approval steps stop carrying information — designing an approval workflow.
FAQ
Is 1% per trade a safe risk level?
It bounds one trade and says nothing about a day. Ten consecutive losses at 1% costs about 9.6% and needs 10.6% to recover; twenty costs 18.2%. Whether 1% is safe depends entirely on how many trades you take, which per-trade sizing does not constrain.
Do I need both a per-trade and a per-day limit?
Yes, because they bound different things. Per-trade limits control the size of one decision; per-day limits control a sequence of them. With only the first, a disciplined trader can lose 15% of an account without ever breaking a rule.
What should my daily loss limit be?
Low enough that reaching it leaves a recovery you would actually attempt, and high enough that an ordinary losing streak does not trigger it. Compute what your expected longest streak costs at your risk per trade, check the implied recovery, and set the limit below where recovery stops being realistic.
Should I limit the number of trades per day?
It is crude and it works, because trade count is the variable per-trade sizing leaves free. Its advantage over a loss limit is timing: a count limit binds before the losses happen, while a loss limit only stops you once you have already lost the limit.