The high-win-rate pitch, version two
Buy a stock in a long-term uptrend when it closes below its lower Bollinger Band, and sell the snap-back. It is usually sold the same way as the 2-period RSI trade in Experiment 02: on how often it wins. This page is the rule, frozen, before the trades existed.
Signals are logged automatically after each close and graded five sessions later. No verdict language of any kind is computed before 2026-11-02, and none is computed then unless both sample floors below are met. The running count is in the live experiments report, regenerated by the scheduled job — and the verdict will be published here either way.
Why this one
Mean-reversion strategies like this one are win-rate-flattering by construction: many small gains as prices revert, punctuated by occasional large losses when they don’t. A 70% win rate and a losing strategy can be the same thing.
That shape is exactly what our reporting exists to expose. Every experiment here publishes the mean, the median and a company-clustered confidence interval together, and the win rate is never allowed to stand in for the result. This experiment runs the same pitch as Experiment 02 through a different mechanism — a band that adapts to volatility — so the two can be read side by side.
It is one of four techniques registered together on 2026-08-06, chosen from a survey of what is currently being taught and sold to retail traders. All four run on the same frozen universe and the same machinery as Experiment 03.
The rule, frozen
- Universe
- The same 40 liquid US large caps as Experiments 02 and 03, fixed at registration. Adding or removing a name voids the test.
- Entry condition (evaluated after the close)
- The close is below the lower Bollinger Band (20-day, 2.0 standard deviations) and above the 200-day simple moving average — the taught, long-only version. If more than five names qualify, take the five with the most negative z-score first.
- Entry price
- The opening price of the next session. Signals are written to an append-only log before that open.
- Exits graded
- The same-day close (the scored horizon) and the fifth session’s close (reported).
- Costs
- 0.10% round trip, applied to every trade, declared up front so it cannot be tuned afterwards.
- Benchmark
- A day-matched control: the equal-weight return of the entire 40-name universe over the identical window, from the same data fetch. The scored quantity is the excess over that control, not the raw return — otherwise the test mostly measures whether the market went up.
- Window
- Only signals dated strictly after 2026-08-06 count. Nothing before the registration date is evidence.
What it has to clear
Frozen at registration on 2026-08-06 and tightened once, on 2026-08-07, before a single outcome had been graded. The bar can only have got harder:
1. At least 30 post-registration graded signals, spread over at least 20 different companies. Thirty trades on four names is not thirty pieces of evidence.
2. The mean and the median day-matched excess positive, net of costs. Both, not either.
3. A 95% confidence interval on that excess that excludes zero, treating each company as the unit of evidence rather than each trade.
4. One look. The verdict is computed once, on the pre-declared date of 2026-11-02 (or at the first run after it where both floors are met), written to an append-only file, and never recomputed. Re-running the numbers later cannot turn a null into a pass.
And one correction for running many tests at once. This is one of eight experiments (02–09) running at the same time, each with roughly a one-in-six prior. About one chance pass is expected across the family. So a single pass is reported as consistent with chance until it survives a Benjamini–Hochberg correction across all eight, and then a fresh forward window. Experiments 06–09 share one universe, so they are correlated tests, not independent replications — how often they fire on the same day is reported alongside the verdicts.
The win rate is reported but is explicitly not a pass criterion.
What we expect to happen
Stating the prediction before the data is part of the method.
Our registered expectation is that the day-matched excess is indistinguishable from zero — and that the win rate will look better than the expectancy does. We put the chance it clears the full bar at roughly one in six.
How often a strategy wins tells you almost nothing. How much it makes when it wins, and loses when it doesn’t, tells you everything.
And if it does clear the bar, we will say so — and then say how many other things we tested at the same time, because that is the number that tells you how surprised to be.
How you’ll be able to check it
Every signal is written to experiments/EXP08-BOLLREVERT-signals.csv before its entry session, and the graded outcomes alongside it — both public and append-only. The frozen rule lives in HYPOTHESES.md as H-EXP08 (registration batch #10), with the constants in experiment_harness.py. The verdict is written once to an append-only verdicts file, recorded in the audit log, and written up on this page.
If we change any constant in that rule, the test is void and we’ll say so. That’s the deal on every experiment here.
What this experiment is not
It is not a recommendation to trade this or anything else, and nothing here is advice. It is not a claim about any author, platform or educator — we test a published technique, never a person. And whichever way it lands, the verdict describes this rule, on these 40 names, over this window. It will not be evidence about Bollinger Band mean reversion in general.
Get every verdict by email
A verdict lands every few weeks. We'll email you each one, free, the day it publishes — and nothing else. No signals, no picks, no offers. Verdicts stay free and public for everyone either way; this just means you don't have to remember to check back.
One confirmation email first — you're not on the list until you click it. Privacy
Educational and informational only — not investment advice, not a recommendation, and not a broker-dealer. ThePickLog is operated by AMD Ventures, LLC (Florida).
← Experiment 07: the moving-average pullback · Experiment 09: volatility contraction (NR7) →
Evidence Validation · Score & metrics · Trust & Data
Tools Calculator · Vetting Guide · 7-Check Scorecard
Legal Disclaimer · Privacy · Terms