We mechanized Wyckoff and Al Brooks. Most of it was a coin flip.
Richard Wyckoff and Al Brooks are two of the most revered names in reading price action — springs and upthrusts, high-2 pullbacks, trend bars, three-push wedges. We turned their signature patterns into hard rules and ran them through the same out-of-sample gauntlet as everything else. The result is humbling, and it explains something important about why great traders are great.
The scoreboard
Six price-action methods, tested across 29 assets and fifteen years, scored only on data the rules never saw (walk-forward out-of-sample Sharpe):
| Method | OOS Sharpe | % positive | verdict |
|---|---|---|---|
| Wyckoff spring / upthrust | +0.07 | 56% | barely alive |
| Brooks wedge (3-push) | +0.06 | 59% | barely alive |
| Brooks high/low-2 pullback | −0.11 | 40% | failed |
| Brooks breakout-pullback | −0.20 | 34% | failed |
| Brooks trend bar | −0.33 | 18% | failed |
| Wyckoff sign-of-strength | −0.33 | 21% | failed |
Four of six failed outright. The two survivors are barely above zero. The single best individual result didn't reach what a plain mean-reversion rule already achieves. Mechanized, the legends are a coin flip.
Why — and this is the point
This is not "Wyckoff and Brooks were wrong." They are demonstrably excellent traders. The finding is subtler and more useful: their edge lives in the judgment, not the pattern. When a skilled reader identifies a Wyckoff spring, they are not just checking "did price dip below support and close back above." They are reading context — where the move sits in the larger campaign, the character of the volume, what failed before, the tone of the tape. Compress that into an if-statement and you strip out exactly the discernment that made it work in a human's hands.
The one that survived, and why it makes sense
Notice which method held up best: the Wyckoff spring — a false break of support that reverses. That is, at its core, mean reversion at a level: the market pokes past a boundary, fails, and snaps back. It aligns with a broader finding from the same gauntlet — that buying the failed move beat chasing the confirmed one. Meanwhile the method that failed hardest was the trend bar: following a big momentum candle, i.e. chasing. The masters' most reversion-flavored idea survived; their most momentum-flavored one died. Even in their failure, the data drew a coherent line.
What we take from it
We read tape the way Wyckoff taught — and we do not trade a mechanized version of it, because now we have measured that the mechanized version has no edge. Instead the discretionary read proposes ideas, and a system like this one disposes of them with out-of-sample evidence before any money is risked. The eye proposes; the data disposes. The legends belong on the first side of that sentence. Their genius is real; it just doesn't survive being frozen into code — which is the most respectful thing the data could possibly say about them.
Every figure traces to a walk-forward backtest on file. We tested the patterns as faithfully as hard rules allow; a discretionary master would object that the rules miss the context, and they would be right — that objection is the entire conclusion. Not advice; a lesson in the limits of automation.