Market studies with the method, the sample size and the code behind them. When a study came from someone else's work we say so and rebuild the numbers ourselves.
Crowd-driven stocks trade 19.8% of their volume in surges, against 9.0% for quiet large caps, and the surges keep going. The loudest bars are followed by an hour 1.17× bigger than normal — and by a move in the same direction only 48% of the time. Stampede uses the reading for the first and not the second.
Four futures markets spend a session’s volatility on completely different schedules. Halfway through the clock, 59% to 68% of the day’s movement is already done — and at three-quarters through, a clock-based projection draws gold 38% too wide.
The same 1,200 tests asked twice on four markets. Predicting the direction of the next move replicates 0 times out of 300; predicting its size replicates 175. And seven famous price levels, tested against a placebo at the same distance, do slightly worse than the placebo.
Split the snap-back to a Bollinger, Keltner, VWAP or regression band into what price did and what the line did. Across 16 band-and-window combinations on four futures markets, 96 to 98 percent of it is the line refitting itself. The tradeable half is 0.84 ticks against a 2.28-tick round turn.
212 scheduled Fed decision days and 59 minute-by-minute sessions. The morning runs at 0.6× a normal day and the press-conference hour at 3.1×. The direction everyone quotes — hike 55%, hold 57%, cut 48% — does not survive a confidence interval.