Valuation

Dip-buy exposure (retired)

The rule that turned the previous residual into a position multiplier — retired on 2026-09-07 after it was shown to subtract value at every scale.

Why it matters

It's documented here precisely because it was retired: it's the project's best example of a rule that passed its original test with a spectacular p-value and was wrong anyway. The number wasn't false — across 57 quarters the overlay beat buy-and-hold on quarterly Sharpe 64.9% of the time, at p=1.5e-06 — but Sharpe can't be averaged across windows and read as a property of the whole period: whole-period Sharpe was 0.721 against 0.840 for doing nothing. It won many quarters by a little and lost the high-variance ones. The metric is still computed because its history feeds charts and models, but it must not be read as a recommended position size.

Build it yourself

Drag the residual z and see what exposure the rule asked for — and, next to it, what the following 90 days historically returned on average in that zone.

Exposure = 1.0× if z ≥ 0 · 1 − 0.6 × z ÷ 2 if z < 0 (clipped to 1.0-1.8×)
Exposure the rule asked for

The zone figures are measured, not illustrative: they come from re-evaluating the rule over the full 14 years of history (analysis/scripts/phase8_overlay_leverage_2026-09-07.py in the project's repo).

Illustrative example numbers for practice — not real data.

How it's calculated, step by step

  1. It starts from the power-law residual z (the previous indicator).
  2. If price is at or above trend (z ≥ 0), exposure is 1.0× — it never goes below. The rule was asymmetric by design.
  3. If price is below (z < 0), exposure = 1 − 0.6 × z ÷ 2, clipped to [1.0×, 1.8×]: the further below trend, the more leverage.
  4. That multiplier would scale the position. That's where it breaks: at maximum exposure (≥1.75×, 104 days in all of history) BTC is on average 48.7% below its high and the next 90 days return −1.3%; at neutral exposure, the next 90 days return +48.7%.
Descriptive. Retired as a sizing rule. With exposure = 1 + k × (exposure − 1), the growth-optimal k is 0.00 (plain buy-and-hold) in all four windows tested, including the 2023+ holdout; and at matched average exposure the shape still loses (Sharpe 0.733 against 0.840). Honesty about the verdict itself: those 104 maximum-exposure days are 3-5 bear episodes, not 104 independent observations, so the defensible claim is "there is no evidence it helps and every estimator says it hurts", not "it significantly predicts worse returns".

Chart

This is real Bitcoin data, computed from the blockchain — not the illustrative example above.

Data last updated on 2026-09-09.

← All indicators