Run 4: Architecture Breakthrough

New 4-briefing set triples the delta from ~5pp to 14.9pp. Treatment stays flat while BTC drops -37.5%

Key Statistics

Arm Returns

ArmReturn
Treatment (Briefings)+0.21%
Control (Price Only)-14.69%
Placebo (Stale Briefings)-16.81%

Deltas

What Changed

Added btc.energy (miner economics) and btc.momentum (trend indicators). Replaced btc.context with specialized briefings.

Market Context

BTC fell from $108K to roughly $67.5K over Sep 2025 through early Mar 2026, a -37.5% drawdown. The January 2026 recovery attempt pushed BTC back above $80K briefly before the February crash took it to new lows. By early March, prices had stabilized near $67K with low volume and subdued volatility.

Observations

The new 4-briefing architecture (btc.energy, cross.regime, cross.breadth, btc.momentum) tripled the treatment-vs-control delta from roughly 5pp to 14.9pp. The treatment arm finished at +0.21%, essentially flat, while BTC itself fell -37.5%. Staying flat during a crash of this magnitude is a meaningful result.

The btc.energy briefing, covering miner economics and hash ribbon signals, proved critical. During late January, hash ribbon flipped to a capitulation signal as smaller miners shut down rigs. The treatment arm went short on Jan 28, three days before the February crash accelerated. Control, seeing only price, interpreted the same consolidation as a buying opportunity.

The btc.momentum briefing provided trend confirmation that the previous btc.context briefing lacked. Instead of raw price data, the model now received pre-computed momentum indicators (7d, 30d, 90d trends with percentile rankings). This reduced the model's tendency to overfit on short-term noise.

(3 more observations in the full report)

Briefings Used

Navigate

Previous: Run 3 | All Runs | Next: Run 5 | View Tick Data

Read the full research findings | How we test