When a 4-hour Phase Oscillator rolls over from the distribution zone during the OpEx Friday → post-OpEx 1-5d window, and price is already extended on the weekly or monthly ATR, dealer gamma unwinds in the trading days that follow. 25 years of SPY: 62% of these signals produce a ≥1% intraday drop within 10 trading days, 46% produce a ≥1.5% drop, 38% a ≥2% drop. Median drawdown −1.16%. The tail prints big: May 2001 (−4.83%), Oct 2011 (−5.11%).
(signal_close − prev_week_close) / weekly_ATR_14 ≥ 0.618(signal_close − prev_month_close) / monthly_ATR_14 ≥ 0.618The edge is tail expansion, not hit rate. The 5-day window under extension only hits 1% 46% of the time — roughly baseline. What changes materially is the 10-day window: hit rates climb sharply and the 25th-percentile drawdown extends further. The pin suppresses action around OpEx; dealer unwinds play out over the following 1-2 weeks.
Max drawdown from signal close measured across the N-day window (the worst point reached at any time from signal fire through end of day N). A "≥1.0% 10d" hit means the drawdown touched −1% at some point in the 10 days, regardless of which day produced the low. Note how 1d and 3d windows underdeliver, while 10d captures the delayed release:
| Horizon | N | ≥0.5% | ≥1.0% | ≥1.5% | ≥2.0% | Median | 25th pct | Worst |
|---|---|---|---|---|---|---|---|---|
| 1 day | 13 | 54% | 31% | 23% | 8% | −0.76% | −1.01% | −2.11% |
| 3 days | 13 | 69% | 46% | 31% | 23% | −0.97% | −1.55% | −5.11% |
| 5 days | 13 | 69% | 46% | 31% | 31% | −0.97% | −2.04% | −5.11% |
| 10 days | 13 | 69% | 62% | 46% | 38% | −1.16% | −2.34% | −5.11% |
Deep extension (weekly or monthly ATR ≥ 1.0) actually underperforms moderate extension at 5 days — strong trends resist a quick break. But at 10 days, deep extension catches up on tail risk.
| Filter | N | ≥1.0% 5d | ≥1.0% 10d | ≥1.5% 10d | ≥2.0% 10d | Med 10d | Worst 10d |
|---|---|---|---|---|---|---|---|
| Unfiltered baseline (full sample) | 81 | 53% | 67% | 40% | 28% | −1.35% | −12.87% |
| OpEx window (no ext filter) | 15 | 53% | 67% | 53% | 40% | −1.54% | −5.11% |
| OpEx window + Extended (≥0.618) | 13 | 46% | 62% | 46% | 38% | −1.16% | −5.11% |
| OpEx window + Deep Ext (≥1.0) | 6 | 50% | 50% | 33% | 33% | −1.06% | −5.11% |
Two pockets of strength: OpEx Friday itself (67% hit rate, n=3) and the Monday immediately following OpEx (75%, n=4). Tuesday–Thursday of post-OpEx week weaken as the suppression fades. Post-OpEx Friday shows a second bump (67%).
| Bucket | N | ≥0.5% | ≥1.0% | ≥1.5% | Med 5d |
|---|---|---|---|---|---|
| OpEx Friday (day 0) | 3 | 67% | 67% | 67% | −2.34% |
| Post-OpEx day 1 (Mon) | 4 | 100% | 75% | 25% | −1.08% |
| Post-OpEx day 2 (Tue) | 3 | 67% | 33% | 33% | −0.76% |
| Post-OpEx day 3 (Wed) | 0 | (no events) | |||
| Post-OpEx day 4 (Thu) | 2 | 50% | 0% | 0% | −0.47% |
| Post-OpEx day 5 (Fri) | 3 | 67% | 67% | 67% | −2.04% |
| Non-OpEx baseline | 66 | 74% | 45% | 32% | −0.93% |
Ranked chronologically. Each drop column shows the max drawdown from the signal close over that horizon — the worst point reached at any time within the window, not a single-day move. Hit = drawdown reached ≥1% within 10 days. Signal close = the 4h bar that crossed below 80. "P+N" = N trading days after OpEx.
| Signal Date | Peak PO | OpEx | Wk ATR | Mo ATR | 1d | 3d | 5d | 10d | Hit? |
|---|---|---|---|---|---|---|---|---|---|
| 2000-03-20 | 100 | OpEx−1 | 0.67 | 0.49 | −1.01% | −1.01% | −1.01% | −1.01% | YES |
| 2001-04-20 | 100 | OpEx | 0.81 | −0.02 | −1.57% | −2.34% | −2.34% | −2.34% | YES |
| 2001-05-22 | 83 | P+2 | 1.05 | 1.11 | −1.56% | −2.54% | −4.83% | −4.83% | YES |
| 2005-05-20 | 89 | OpEx | 1.29 | 0.18 | +0.09% | −0.21% | −0.21% | −0.21% | no |
| 2005-11-22 | 88 | P+2 | 0.79 | 0.51 | +0.42% | +0.29% | −0.31% | −0.31% | no |
| 2007-03-23 | 93 | P+5 | 1.69 | −0.06 | −0.98% | −1.55% | −2.04% | −2.04% | YES |
| 2011-10-28 | 83 | P+5 | 0.65 | 0.63 | −2.11% | −5.11% | −5.11% | −5.11% | YES |
| 2015-10-23 | 83 | P+5 | 0.52 | 0.85 | +0.03% | −0.34% | −0.34% | −0.34% | no |
| 2017-10-23 | 81 | OpEx−1 | 0.61 | 1.02 | −0.32% | −1.16% | −1.16% | −1.16% | YES |
| 2019-04-24 | 81 | P+2 | 0.35 | 0.85 | −0.76% | −0.76% | −0.76% | −2.44% | YES |
| 2019-11-18 | 80 | OpEx−1 | 0.41 | 0.92 | −0.23% | −0.92% | −0.92% | −1.54% | YES |
| 2019-12-27 | 92 | P+4 | 0.36 | 1.20 | −0.84% | −0.97% | −0.97% | −0.97% | no |
| 2021-10-21 | 94 | P+4 | 1.26 | −0.03 | +0.02% | +0.02% | +0.02% | +0.02% | no |
The 4h rollover is a watch signal, not an immediate trigger. At 1 day, only 31% of extended events hit 1%. The dealer pin suppresses breakouts in both directions while the monthly chain is alive. Sizing into weeklies expiring within 3 days of the signal is fighting the same flows that caused the pin.
Plan for 10-day expiries minimum. The 25th-percentile 10d drawdown is −2.34% and the worst printed −5.11% (Oct 2011). Long-dated puts — not weeklies — capture the release.
OpEx Friday + Post-OpEx Monday are the clean triggers. Those two buckets hit 67% and 75% at just 5 days (unfiltered). If the 4h rollover fires on the OpEx close or the following Monday, the setup is at its cleanest. Post-OpEx Tue/Wed/Thu weaken rapidly — by then, the pin is already gone and the information is stale.
Deep extension ≠ better edge at 5 days. Signals at weekly or monthly ATR ≥1.0 hit 1% at 5 days at 50% vs moderate extension (46%) — roughly matched. Strong momentum can resist a fast break. At 10 days, deep extension converges with moderate. Size for the tail, not the median.
Nulls concentrate in 2019–2021. Several "no drop" events came from the 2019–2021 stimulus era. If regime is unusually pinned (zero-rates, dealer long gamma), discount the signal. Hit years in the current sample (2000, 2001, 2007, 2011, 2017, 2019) were tighter liquidity environments where dealer unwinds actually produced vacuum effects.