Restructured from monolithic to modular briefings. Delta persists through a much deeper BTC drawdown (-39%)
| Arm | Return |
|---|---|
| Treatment (Briefings) | -10.35% |
| Control (Price Only) | -15.69% |
| Placebo (Stale Briefings) | -22.11% |
Moved from single full-context dump to named modular briefings. Extended window to 181 ticks.
This extended window covers Sep 2025 through Feb 2026, a brutal period for BTC. The price fell from $108K to roughly $66K, a -39.2% drawdown. The November 2025 selloff was followed by a partial recovery in December and January, then a second, sharper leg down in February 2026 driven by global risk-off flows and exchange liquidations.
The move from monolithic to modular briefings (btc.context, cross.regime, cross.breadth) was the first architectural change. Each briefing now had a clear purpose and the model could reference them by name. The delta held at +5.34pp despite nearly doubling the window length and testing through a much deeper drawdown.
The control arm lost -15.69% over 181 ticks. It showed a consistent pattern: buying dips that turned into further declines. During the January recovery attempt, control added long exposure, then held through the entire February crash. Without regime context, the model could not distinguish a genuine reversal from a dead-cat bounce.
The treatment arm lost -10.35%, still negative but substantially better. It recognized the February crash earlier than control, going short on Feb 5 when the cross.regime briefing flagged a shift to "risk-off" with deteriorating breadth across equities and crypto. Control did not go short until Feb 12, by which point much of the damage was done.
(3 more observations in the full report)
Previous: Run 2 | All Runs | Next: Run 4 | View Tick Data
Read the full research findings | How we test