Quantitative core

Position sizing & stops

Risk-first sizing bounded by capital, liquidity and exchange rules; ATR/structure/hybrid stops; risk-reward and trailing targets; precision-safe rounding.

Sizing answers one question: how much can this trade lose, and how many units does that allow once liquidity, capital and exchange rules are honored. The number is deterministic and conservative — when constraints conflict, the tightest one wins.

Risk-first sizing

Size is computed by the risk engine, never by the strategy or the LLM. It starts from the amount you are willing to lose, then the stop distance fixes the quantity.

position_sizing_v1

Risk amount & raw quantity

R i s k A m o u n t = E q u i t y \times R i s k P er ce n t Q u an t i t y = \frac{R i s k A m o u n t}{∣ E n t r y P r i ce - S t o pP r i ce ∣}

One unit is assumed to lose the stop distance if the stop fills exactly. Risk amount divided by that distance gives the quantity.

$R i s k P er ce n t$: fraction of equity at risk, e.g. 0.005 = 0.5%

position_sizing_v1

Slippage- and fee-adjusted size

A d j Qt y = \frac{R i s k A m o u n t}{∣ E n t r y P r i ce - S t o pP r i ce ∣ + A d v er se S l i pp a g e _{u ni t} + E x i tF e e _{u ni t}}

The real loss per unit is wider than the stop distance: adverse slippage and the exit fee are added before dividing.

python

stop_distance = abs(entry_price - stop_price)
slip_per_unit = entry_price * adverse_slippage_percent
exit_fee_per_unit = stop_price * exit_fee_rate
loss_per_unit = stop_distance + slip_per_unit + exit_fee_per_unit
raw_quantity = risk_amount / loss_per_unit

The final quantity is a minimum

position_sizing_v1

Bounded by every limit at once

F ina l Qt y = min (R i s k, C a p i t a l, L i q u i d i t y, E x p os u r e, E x c han g e)

The smallest of every binding constraint wins. A generous risk budget never overrides a thin pool or an exchange minimum.

The final quantity must respect, in code:

Bound	Source
Capital	Available quote balance, reserve, max allocation %
Liquidity	Order-book depth — see Liquidity & slippage
Exposure	Portfolio heat, per-asset / per-strategy caps
Exchange	Min/max amount, min cost, step size, tick size

Stops

A fixed-percentage stop is simple but not adaptive. ATR and structure stops adapt:

atr_stop_v1

ATR stop

S t o pP r i ce = E n t r y P r i ce - k \times A T R_{n}

$k$: ATR multiplier (backtested)
$A T R_{n}$: average true range over n periods

hybrid_stop_v1

Hybrid stop distance

S t o p D i s t = max (M inim u m, A T R, S t r u c t u r e, S p r e a d A d j u s t e d)

Take the widest of the candidate distances so the stop is never tighter than volatility, structure or the spread allow.

A trade is rejected when the stop is unsafe:

stop distance is zero, or the stop is at/above entry for a long;
distance is below the minimum market movement, spread or slippage buffer;
rounded quantity violates exchange rules or pushes risk over the limit;
no reliable invalidation level exists.

Take-profit & trailing

target_v1

Risk-reward target

T ak e P r o f i t = E n t r y P r i ce + R R \times (E n t r y P r i ce - S t o pP r i ce)

$R R$: target reward-to-risk ratio

trailing_v1

ATR trailing stop (long)

T r ai l in g S t o p_{t} = max (T r ai l in g S t o p_{t - 1}, H i g h S in ce E n t r y - k A T R_{t})

The trailing stop only ratchets up, locking in gains as the high-water mark rises.

Partial exits track remaining quantity after each fill, e.g. 25% at 1R / 2R / 3R, the rest trailed.

Exchange precision & rounding

position_sizing_v1

Round down to the step size, then re-validate risk

R o u n d e d Qt y = ⌊ \frac{R a w Qt y}{S t e pS i z e} ⌋ \times S t e pS i z e

Round down to a legal increment, then recompute the planned risk on the rounded quantity — rounding can change it.

Re-validate after rounding

The order is rejected if rounding produces a cost below the minimum, a zero quantity, a planned risk above the configured limit, a quantity above the available balance, or an invalid price increment.

A c t u a l R i s k = R o u n d e d Qt y \times L oss P er U ni t

must still clear the limit.

Prefer no trade over an unquantifiable one

If no reliable stop exists, or the post-rounding risk cannot be proven under the limit, the engine declines. A position that cannot be sized safely is not taken.