Every trading education repeats the same shape: impulse, consolidation, impulse. The market coils, energy builds, and the breakout explodes out of the quiet. It's an appealing picture — it implies you can sit out the boring days and be positioned for the violent ones.
We tested it. Fourteen signals, all computable before a session begins, against six definitions of "big day," on 1,876 sessions of Nasdaq futures, 1,869 of gold and 1,311 of crude. We wrote the verdict rule down first: a signal counts only if it shifts the odds by at least 1.5×, on at least 40 days, with the same sign in both halves of the sample.
Big days turned out to be genuinely predictable. And the coil-then-explode intuition turned out to be exactly backwards.
Read the left side first, because that's the part that kills the myth. After a tightly coiled five-day stretch, a huge day is less than half as likely as normal — 0.43×. Not neutral. Less likely. The narrowest-range-of-seven-days setup, a staple of breakout trading, lands at 0.89× — slightly worse than picking a day at random. Three consecutive quiet days: 0.74×. An inside day: no signal at all.
Now the right side. Every "the market is already moving" signal predicts more movement. A gap of more than 1% at the opening bell makes a huge day 2.37× more likely. Expanding five-day range: 2.24×. Heavy volume yesterday: 2.04×. Yesterday itself moved more than 1%: 1.96×.
This isn't a quirk of the Nasdaq. Gold and crude, tested independently with the same pre-declared rules, say the same thing:
None of this is new to statisticians — volatility clustering is the observation that won a Nobel Prize (Engle's ARCH models, 1982). What's striking is how completely the popular trading version of the story inverts it.
So: you can predict the big days. The obvious next thought is "great — trade only those." We tested that too, and this is where it gets genuinely instructive.
The best big-day predictors are not the best money-making days. Filtering to days after a wide-range session cuts profit per day from $8.78 to $1.08 — a 90% destruction of the edge while predicting big ranges perfectly well. Taking every "expansion" signal together does worse than taking all days. And the quiet days everyone wants to skip? They earn $9.52 per day at a Sharpe of 1.00 — better than the baseline.
The reason is mechanical, and it's the same lesson as our finale: a strategy's risk is usually measured in the same units as the market's volatility. Our breakout's stop is the width of the opening range. On a big-range day, that stop is enormous. The move has to be proportionally larger just to break even — and it usually isn't. You get the drama without the payoff.
We didn't have to hunt for confirmation. Our own 2026 is the case study:
| year | median opening range | median day's move | move ÷ risk | champion P&L |
|---|---|---|---|---|
| 2024 | 91 pt | 100 pt | 1.09 | +$3,847 |
| 2025 | 124 pt | 125 pt | 1.01 | +$3,980 |
| 2026 YTD | 181 pt | 162 pt | 0.89 | −$594 |
2026 has the widest ranges in our seven-year record — the most "opportunity" by every volatility measure — and it's the only losing year. The win rate held up fine at 53%. The market's intraday drift was actually strong. But risk per trade inflated 46% while the moves it was harvesting grew 30%, and that gap is the entire loss. The biggest days delivered the worst results, exactly as this study predicts.
Two honest caveats before anyone acts on this page. First, we examined fourteen filters — with that many looks, one or two will always appear profitable by luck alone. Two did look good here (an expanding-range regime, and heavy prior-day volume, both around double the baseline profit rate). We are explicitly not treating those as findings. They go into the same machine as everything else: a pre-declared grid, out-of-sample cross-validation, an overfitting probability, and slippage stress. We expect them to fail, and we'll publish it either way.
Second, what is solid is the negative result, and negatives are the most portable thing in this business: if your strategy's premise is "wait for the coil, catch the expansion," the data across three markets says the coil is evidence against the expansion. That's worth knowing before you build on it.
scripts/big_mover_study.py. Nothing here is trading advice.