Stop guessing. Start building.

Here is what this big-range-day study cannot tell you. It's frictionless: no commission, no slippage. The forward drifts (down days +0.305% over 5 days, low closes +0.351%) are gross. The volatility-clustering signal doesn't care about costs, but any return you'd try to harvest from it does. It's an event study, not a system. Events overlap, and the most extreme buckets are thin. The 3+ ATR days are only 40 events from 1993 to 2026, the 2.5-3x ATR bucket only 49, flagged rather than dropped, and their exact numbers will move on new data. If you try to harvest those drifts without a cost model, you are reading gross numbers as net. What survives: next-day range rises with today's ATR x size, and the effect decays over weeks not hours. Long ranks above short (median CAR/MaxDD +0.010 to +0.020 long, negative short), but even long sits below a random-entry control's 0.025. No edge over luck. Turning any of it into a strategy is separate work. Full caveats: https://statoasis.com/overfit/research/sp500-big-range-days

2,736 backtested variants. 8,397 events. Three size lenses that flatly disagree. The design: Every SPY session since 1993 flagged as a big range day, measured for forward return AND next-day volatility at 0-20 day horizons. Then the same days bucketed three ways: raw points, price %, and ATR x, plus direction (up/down) and close location (low/mid/high). Then a flat-only backtest across 942 long and 942 short reliable variants, $35,000 each, frictionless. Points says big days are a 2020s phenomenon: 549 of 635 six-point days landed after 2019. Price % peaks in the 2000s. ATR x spreads evenly across every decade. If you pick the points ruler, you will think the 2020s invented volatility. Only one of those can be the honest yardstick, and it's the one that normalises each day against its own era. The trading sweep is a ranking of sides, not a finished system. Full methodology and the era-distortion table: https://statoasis.com/overfit/research/sp500-big-range-days

I expected the volatility spike to fade by the next session. It didn't. The intuition is that a big range day is a one-off, a shock that resolves overnight. The data says the elevated volatility is sticky. For the biggest days (1.5+ ATRs), the next-day range runs 1.297x the average, and it's still 1.202x five days later and 1.128x a full month out. Calm days do the mirror image: they start at 0.855x and drift back toward normal. The shock doesn't reset. Not a flicker. It decays, slowly, over weeks. If you size a position off yesterday's range, you are already late by a week of elevated risk. That changes how you'd size and stop around one of these days. The risk environment you just entered is the one you're living in for a while, not a one-session spike. Full decay curve and horizon table: https://statoasis.com/overfit/research/sp500-big-range-days