Pre-registered 2026-10-09 · 1 run · data snapshot 2026-09-19T090800Z
The idea
The 52-week high is a number every trader can see, and many anchor on it. When good news pushes a stock toward that level, investors hesitate to bid it past the old high, so the news is absorbed slowly and the price keeps drifting up afterwards instead of reversing. George and Hwang reported in 2004 that closeness to the 52-week high explains classic momentum rather than the other way round.
In this lab the signal had only been tested combined with two other conditions, under an older raw-Sharpe method known to have produced at least one false negative on a sibling signal. A corrected information-coefficient scan then ranked the factor strongly, with t +8.4 to +11.7 out of sample. That was reason enough to test it on its own.
How it was tested
Pre-registered on 2026-10-09 as a clean factor test: price within 5% of its 52-week high, nothing else. US stocks at $10 or more with at least $5M of 20-day dollar volume. Buy at the next open, hold 20 sessions, no stop and no target, taking the names closest to the high first. Neutral sizing: $5,000, 1% risk per trade, 10 slots. The bar was Gate 7, Jensen alpha t of at least 1.65.
Where it died
The single run, on 2026-10-10, gave a Jensen alpha of 0.072% per trade over 23,834 trades, t +0.53, with beta 0.38. Gate 8 was not evaluated. The alpha was also far below the 0.30% per trade that Gate F asks for, so even a passing t would not have paid for execution. The strong scan and the flat backtest measured different things: the scan ranks the whole universe, the strategy buys only the names nearest the high.
The record
Thesis
Being close to the 52-week high predicts positive returns over the next 20-30 sessions on its own. The 52-week high is a cognitive anchor: good news near it is absorbed slowly (underreaction), producing continuation rather than reversal. George & Hwang report that this measure subsumes classic momentum.
Prior: The factor had only been tested before combined with two other conditions, under the lab's older raw-Sharpe methodology, which had produced at least one documented false negative on a sibling signal. The corrected information-coefficient scan gave it t +8.4 to +11.7 out of sample.
George, T. & Hwang, C. (2004). The 52-Week High and Momentum Investing. Journal of Finance 59(5).
Specification
signal
price within 5% of its 52-week high — nothing else
universe
US stocks, price ≥ $10, 20d dollar volume ≥ $5M
entry
next open
exit
20 sessions, no stop or target (clean factor test)
ranking
closest to the high first
sizing
neutral default — $5,000, 1% risk per trade, 10 slots
Gates
Pre-registrationpass
Was the hypothesis, its parameters and its pass/fail rule written down before any data was looked at?
Gate 7 — Jensen alphafail
After regressing out market exposure, is there alpha left?
t ≥ 1.65
t (HAC)+0.53alpha / trade %+0.07beta+0.38trades23,834
Gate 8 — Capture-preserved alphanot evaluated
Does the subset of trades a real portfolio can actually take keep the alpha?
Not evaluated once Gate 7 fails.
Why it died
No alpha on its own: t +0.53 over 23,834 trades, beta 0.38. The result was stable to the data-cleaning choices tried (+0.79 without removing interleaved symbols). Part of what the information coefficient saw was market exposure, which the coefficient does not remove and Jensen alpha does.
What it taught
Information coefficient +11.7 → alpha t +0.5. The coefficient measures whether a factor orders the whole universe; a strategy buys one tail of it. The suspicion that the May rejection was a false negative of the old method did not hold up. One of five long momentum / 52-week-high configurations in this lab, none with Jensen alpha.