ZAPAS Walk-Forward Backtest

ZAPAS Walk-Forward Backtest – –

Universe

Period start

Period end

Min mcap $M 2000

Min PM vol (shares) 20k

Min PM bars 5

Min |gap| % 0.5

Grade filter

Min grade

Direction

Risk model

Base risk $ 400

Stack (max positions/day) 20

Stop % (hard kill) 10

Auto-flatten $ 10000

Cost bps/leg 25

Grade sizing exp

Stop = max(risk / pos_size, stop_pct). Stop triggers on individual trade loss > stop_pct. Flatten = daily losses cumulative exceed $flatten → remaining trades skipped.

Liquidity (per-trade cap)

Mode

Mega >$50B cap $k 500

Large $10-50B cap $k 200

Mid $2-10B cap $k 50

Small <$2B cap $k 10

Cap = максимум $ на одну позицию. Если sizing × base_risk > cap → режется до cap. Это предотвращает нереалистичные $1M позиции в mid-caps.

Compounding (exp. growth)

Mode

Initial equity $ 25000

Kelly fraction % 0.4

Level: base_risk=$100 → $400 → $1200 → $3000 автоматически когда cum P&L пробивает $20K→$60K→$150K. Caps тоже растут.
Continuous: base_risk = equity × (kelly%/10). Ежедневно.

Trades

WR %

Avg Ret %

Total P&L $

Neg days

Trades/day

Sharpe (d)

Max DD $

Worst day $

Best day $

Neg day %

Profit Factor

Equity curve (cumulative $)

Daily P&L distribution

Trades per day distribution

By grade / direction

Grade	Dir	N	WR%	Avg%	Sum%

By year (walk-forward)

Year	Days	Neg	N trades	WR%	Avg$	Sum$	Sharpe

By month

Forecast — 1 Year Forward (Monte Carlo bootstrap, 1000 sims)

Expected P&L (median)

–

P95 (upside)

–

P5 (downside)

–

Expected neg days

–

P(profit) %

–

Max DD (median)

–

Метод: bootstrap resampling из дневных P&L текущего фильтра (grade/stack/risk). 1000 путей × 252 торговых дня. Показано: медиана + 5/95 перцентили полосой. Neg days distribution — сколько красных дней ожидать за год.