Validation

OOS

Out-of-sample testing

Why

The parameters were chosen on the discovery sample. The sealed sample is a separate book. The same cash is calculated again, to see how much remains on prices that did not take part in the choice.

Formula

This book

The sealed sample is booked on its own, also starting at 10000. Costs match the discovery sample: taker 0.05, maker 0.02 (percentage points), 1× slippage of 10 basis points, and funding of 0.0001 per 8 hours. The next fill is not sized from the cash that remains.

Fills that count: both entry and exit are inside the sealed sample
A fill that crosses the wall (entry before the wall, exit after it) counts in neither book
An entry in the sealed sample must clear the embargo. An entry within the embargo bars after the wall does not enter this book

Cash of one fill

Long cash = (exit price − entry price) × quantity × contract multiplier − fee cash − funding cash
Short cash = (entry price − exit price) × quantity × contract multiplier − fee cash − funding cash

Slippage fraction s = 10 / 10000
Fee cash = notional × rate / 100
Funding cash = notional × 0.0001 × bars held / bars per 8 hours

A long buys higher and sells lower. A short does the opposite. Funding is paid by a long and received by a short.

Profit factor

PF = (sum of cash on winning fills) / (sum of the absolute cash on losing fills)

When there are no losing fills and there are winning fills, PF is recorded as 99. The denominator is the sum of the absolute losing cash, not the average loss.

Pass

Sealed-sample fills ≥ 20
PF ≥ 1
No liquidation

All three must hold. Compounding does not enter this gate. Passing the discovery sample does not make up for the sealed sample. Fewer than 20 fills, a profit factor below 1, a liquidation, or a book that cannot be read: not passed.