Sep 1 - Nov 22, 2025, 83 ticks, Sonnet 4.5
Treatment
-2.03%
Control
-6.49%
Placebo
-6.79%
Treatment vs Control
+4.46pp
BTC Buy & Hold
-21.4%
Run 1's briefings on a second model, Sonnet 4.5, over a window 8 days shorter: the briefings arm finished 4.46 points above price only.
Switched from Opus 4.6 to Sonnet 4.5, with Run 1's briefings. The window ends 2025-11-22, 8 days before Run 1's.
BTC fell from $108K to about $85K by 22 November 2025, down 21.4%. The window ends 8 days before Run 1's, the day before the low of the November selloff.
Same briefings as Run 1, different model (Sonnet 4.5 instead of Opus 4.6)
full-context
Complete market context dump, all available data in a single briefing
regime-check
Binary regime classifier: bullish vs bearish market state
market-breadth
Cross-asset breadth indicators: how correlated are risk assets
See the exact preamble and briefing content for all arms
Run 2 used Sonnet 4.5 in place of Opus 4.6, with Run 1's briefings, over a window that ends 2025-11-22, 8 days before Run 1's. The briefings arm finished 4.46 points above the price only arm (Run 1: 5.28). One run on each model cannot show where the gap comes from.
The briefings arm returned -2.03%, the price only arm -6.49%.
How We Test
The simulated backtest design and its controlled variables
Research Deep Dive
Full research analysis and findings across all runs
Context Engineering for Crypto Agents
Why context format affects agent decisions