Reproduces Run 4's 14+ pp delta. Treatment now profitable (+8.46%) while control loses -6.31%
| Arm | Return |
|---|---|
| Treatment (Briefings) | +8.46% |
| Control (Price Only) | -6.31% |
| Placebo (Stale Briefings) | -18.85% |
No briefing changes. Pure replication of Run 4 to confirm the result isn't a fluke.
BTC fell from $108K to roughly $71K through Mar 6, 2026, a -34.3% drawdown. The Feb crash dominated this window, but by early March there were signs of stabilization. Volume picked up near $70K support and funding rates turned negative, hinting at crowded short positioning.
This was a pure replication run: same 4 briefings as Run 4, no changes. The treatment-vs-control delta came in at +14.77pp, within 0.13pp of Run 4's result (14.90pp). This level of reproducibility across independent runs is the strongest evidence that the delta is systematic, not lucky.
The treatment arm was now profitable at +8.46%, a significant improvement over Run 4's near-flat +0.21%. The additional 2 ticks (Mar 5 and Mar 6) captured a short-covering bounce near BTC's $70K support level. The treatment arm went long on Mar 5 when the btc.momentum briefing flagged a momentum divergence (price making lower lows while momentum made higher lows).
The control arm returned -6.31%, better than Run 4's control (-14.69%). This variance in control performance is expected: without context, the model's decisions are more random and path-dependent. Some runs the control gets lucky, others it doesn't. The treatment arm's consistency across runs (stable delta) is what matters.
(2 more observations in the full report)
Previous: Run 4 | All Runs | Next: Run 6 | View Tick Data
Read the full research findings | How we test