← Back to Research
Infrastructure5 min readNovember 28, 2024

Why Every Trade Needs a Paper Trail

Automated trade logging capturing all 38 feature values, model confidence scores, and execution details for every trade.

LoggingAudit TrailDebugging

What Gets Logged

Every trade in V7 generates a complete log entry containing: all 38 L1 feature values at signal time, L1 confidence score, L2 gate decision and score, entry price and actual fill price, stop and target levels, exit price and exit reason, bars held, R-multiple result, and the regime classification (Hurst, ADX, volatility zone, K-Means cluster).

This produces approximately 200 bytes per trade, totaling about 900KB for the full 4,505-trade backtest. In live trading, logs are written as structured JSON to both file and database for redundancy.

Debugging with Complete Context

When a trade loses money, the journal tells you exactly why. Was L1 confidence borderline? Was ADX below 20 but above the threshold? Was spread elevated but below the filter level? Without complete logging, debugging a losing trade means guessing. With S27, every decision is transparent.

The journal enabled identification of the L2 over-filtering issue. By analyzing logged L2 scores, we discovered that L2 was skipping 60% of signals, far more than intended. Without trade-by-trade L2 score logging, this would have been invisible.

Compliance and Trust

For FTMO and eventual fund operations, audit trails are not optional. S27 provides regulatory-grade trade documentation that shows every input, every decision, and every outcome. When an investor asks "why did you take this trade," the answer is a structured data record, not a narrative. The logging overhead is negligible (sub-millisecond per trade) and the storage cost is trivial. The value in debugging, compliance, and continuous improvement makes S27 one of the highest-ROI modules in the system despite producing zero direct alpha.