Strategy Graveyard
← All strategies

short-term mean reversion / overbought

Overbought run-up short

RedundantArchived · short · 2014-2026

Pre-registered 2026-10-23 (commit 6910cb5) · 1 run · data snapshot 2026-09-19T090800Z

Prior: Defined in May 2026 as one of the "tier 1" shorts picked out of a wider search over the same 2014-2026 data, so the October full-history re-measurement is a confirmation of a selected configuration, not an independent test.

Specification

Signal, universe, exits and sizing are withheld because they are close to a strategy that is still in use.

Gates

Gate 7 — Jensen alphapass
After regressing out market exposure, is there alpha left?
t ≥ 2.13 (Bonferroni, 3 shorts measured together), net of 20%/yr borrow
t (HAC)+4.02alpha / trade %+0.46beta+0.81trades12,796
Gate 8 — Capture-preserved alphapass
Does the subset of trades a real portfolio can actually take keep the alpha?
t_cap ≥ 0.78 (scaled) and ratio ≥ 0.50
t captured+1.97capture ratio+1.02
Gate F — Frictionpass
Is the alpha per trade large enough to pay for its own execution?
alpha/trade ≥ 0.30% net of borrow
alpha / trade %+0.46
Stabilitypass
Is the alpha present in both halves of history?
t > 0 in both halves
t 2014 2019+0.98t 2020 2026+3.96
Own alphafail
Is the alpha new, or already captured by a strategy that passed?
t ≥ 2.13 on the trades that do not overlap the surviving short, n ≥ 100
t own+0.07n own7,550overlap pct+41.00
41.0% of its trades overlap the surviving strategy's trades; the other 7,550 carry no alpha at all.

Why it died

The alpha is real over full history (t +4.02, 0.464% per trade net of borrow, positive in both halves, captured by a small portfolio). But all of its alpha sits inside the trades of the lab's surviving short strategy, whose details are not published: on the trades it does not share with that strategy, t is +0.07. It is a partial window onto an edge the lab already has, not a second one.

What it taught

This strategy was archived twice for the wrong reason before it was archived for the right one. In May it failed Gate 8 under a capture ratio later shown to measure sample size; the archive note then quoted "t 0.13 on full history", which was actually a 3-6 month regime-window number. The mislabel was caught only because the strategy was used as a negative control elsewhere and came out at t +5.67. Asking "is the alpha new?" is what finally decided it.

Verdict history

  1. 2026-05-12Yearly and VIX-regime breakdown: 11 of 12 years with positive Sharpe; promoted to the demo at full size
  2. 2026-05-27Archived at Gate 8: t_captured 0.14 (engine alpha t +10.09), under the capture formula and simulator later found broken Not capturable
  3. 2026-09-14Re-examined on unseen data (2026-05-27 to 2026-09-11) with the corrected Gate 8: n 314, t +0.54 vs 2.128 needed, capture ratio 1.67 No alpha
  4. 2026-10-23Used as a negative control in another study, it failed as a control: t +5.67 over full history. The archived 't 0.13' had been a regime-window figure
  5. 2026-10-23Full-history re-measurement, pre-registered: passes Gates 7, 8, F and stability; own-alpha t +0.07 Redundant