← backtests · 09-17 02:03Z · MOCK data
phase 11 hosted verification
isolated full period (exploratory) fx.momentum EURUSD
MOCK world mockworld-seed20240101 mockworld-1.0.0+65d52d17 · hash 05b2a3be84c3 · synthetic
Return+0.43%CAGR —%
Max drawdown0.58%4 days
Sharpe / Sortino2.16 / —Calmar —
Trades2win 50% · PF 2.23
Expectancy+0.604R$215 per trade
Costs$13financing $6 · turnover 28.5
| regime usd | Trades | Exp R | P&L |
|---|---|---|---|
| usd strengthening | 2 | +0.60 | +431 |
| session | Trades | Exp R | P&L |
|---|---|---|---|
| london | 2 | +0.60 | +431 |
| review class | Trades | Exp R | P&L |
|---|---|---|---|
| wrong regime | 1 | −1.02 | −349 |
| thesis confirmed | 1 | +2.23 | +780 |
- event policy
- 1
- confidence minimum
- 1
- no conflicting position
- 4
MOCK data: results demonstrate the machinery, not edge in real markets
- Source
- MOCK world
- Dataset
- mockworld-seed20240101
- Version
- mockworld-1.0.0+65d52d17
- Hash (sha256)
- 05b2a3be84c34aa315380de1f5f22d371285150d408ca36879516444d306bba9
- Window
- 2026-08-01 00:00Z → 2026-09-01 00:00Z
- Warm-up loaded from
- 2026-04-14 16:00Z
- Series hashed
- 15 · 1h
| Instrument | First bar | Last bar | Bars | Warm-up | Largest gap |
|---|---|---|---|---|---|
| EURUSD | 2026-08-02 21:00Z | 2026-08-31 23:00Z | 507 | 1877 | 49h |
no split: an in-sample exploration, not evidence of edge.
- Trades outside the segment
- 0
- History in confidence (D3)
- 0 of 8 decision(s) with history
| Examined before? | Period | Earlier runs | Earlier segments |
|---|---|---|---|
| full fx.momentum EURUSD | 2026-08-01 → 2026-09-01 | 0 | none |
Controls in force and the history rule
- point-in-time contexts: a feature or regime row available after the decision time raises LookAheadError
- a decision at a bar close fills no earlier than the next bar's open
- poisoned-future test in the suite: decisions before a cutoff are unchanged when later bars are replaced
- embargo between train, validation, out-of-sample and holdout equal to the strategy's maximum holding period
- final holdout reserved by walk-forward and evaluated separately
- data-reuse ledger: runs that examined the same bars of the same dataset are counted
- dataset hash over every bar and event the run could see
- historical-performance confidence is point-in-time (D3)
History rule: point-in-time: the latest walk-forward record created at or before the decision, same data mode, pooling only its out-of-sample trades closed by then; otherwise neutral.
Modelled
- spread
- half the bar spread is paid on every market and stop fill; synthetic spreads from the MOCK world (session multipliers applied)
- slippage
- base bps by asset class + volatility coefficient x realised-vol percentile + size coefficient x notional / hourly liquidity, times a deterministic per-order factor in [0.5, 1.5]; market and stop fills only (limits fill at the limit or better)
- fees
- instrument commission (per million notional or bps of notional)
- latency
- a decision at a bar close can fill no earlier than the next bar's open; latency_ms is recorded on each fill but does not move the price
- sessions
- market orders are refused while the venue is closed and do not fill in closed or maintenance bars; the execution policy places no orders in closed, maintenance, rollover, weekend_reopen
- gaps
- stops that gap fill at the open (worse than the stop); limits that gap through fill at the better open
- intrabar order
- exits and stops are processed before take-profits, so a bar that touches both the stop and the target is counted as stopped (conservative)
- financing
- carry charged at each 17:00 New York rollover from the instrument's annual long and short rates, FX Wednesdays x3, 360-day count Rates for EURUSD are illustrative.
- partial fills
- entries are capped per bar at hourly liquidity x a liquidity-state multiplier; the remainder is cancelled after 3 strategy bars
- kill switch
- in a backtest, breaching the daily-loss or drawdown limit halts new entries for the rest of that trading day and cancels working orders of the configured roles
- rejections
- entries are refused without risk approval, for invalid size or insufficient margin (instrument max leverage), plus a deterministic 0.2% simulated requote
Not modelled
- order-book depth and queue position: fills are simulated from OHLC bars, not ticks
- the price path inside a bar beyond its open, high, low and close
- exit size limits: exits and stops fill in full whatever their size
- lasting market impact: slippage applies to the order only
- broker margin calls and forced liquidation: positions are never closed for lack of margin
- time-varying financing rates and crypto perpetual funding: rates are fixed in config
- commission tiers, minimum tickets and exchange fees
- futures contract rolls (commodities trade as continuous instruments)
- broker swap tables: financing rates for EURUSD are illustrative placeholders
- real market spreads, liquidity and prices: the data is synthetic
- risk
- risk-1.0.0+e561f3f2
- execution
- exec-1.0.0+821dd562
- confidence
- conf-1.0.0+18bcb659
- data_source
- mock
- feature_set
- fs-1.0.0+bbfe6af5
- regime_model
- regime-rules-1.0.0+ba09c5ab
- strategy:fx.momentum
- fx.momentum@1.0.0