Under Re-Verification (2026-04-22) — This study was built using a pre-audit version of the backtest pipeline. The current code (post methodology-audit, commit c92ef0f) produces 88 episodes with 4h PO peak > 80, not the 118 shown below. The event log, signal bar chart, and velocity buckets reflect the older analysis and are pending rebuild. Do not trade from the specific numbers on this page until it has been refreshed.
Home
Study — 4-Hour Phase Oscillator

When the 4h PO peaks
above 80 — what makes
it a real reversal?

The 4-hour Phase Oscillator in extended territory is not a sell signal. 65% of the time, price consolidates sideways and the 8 EMA catches up. But when a real reversal comes, it follows a specific escalation sequence — and the warning signs are early.

Over 25 years of SPY data, we found 118 episodes where the 4-hour Phase Oscillator peaked above 80. We tracked what happened next: did price consolidate and continue, or did it reverse into a 5%+ drawdown?

118
Episodes with
4h PO peak > 80
35%
Led to 5%+ drawdown
(41 episodes)
44%
Consolidated <3%
(52 episodes)

The 4h PO being extended is not the sell signal — it just loads the spring. Two thirds of the time, price barely dips and the uptrend resumes. The question isn't whether the PO is extended. It's what happens after it starts falling.

When the 4h PO peaks above 110 and then drops below 100, how often does price mean-revert to the 4h 8 EMA? 100% of the time. But the price action is almost always sideways consolidation, not a drop.

Mean Reversion Rate
PO > 110 → crosses below 100
100%
7 of 7 episodes touched the 4h EMA8
Median time1 bar (4h)
Max time8 bars (48h)
What Price Actually Did
Peak PO → cross below 100
-0.22%
Median price change — all 7 were sideways
Max drawdown-0.59%
Fwd 10 bars+0.74%

The "mean reversion" is misleading. Price barely moves — the EMA catches up to price during a shallow pause, not the other way around. In the PO > 110 group, forward 10-bar returns were positive 100% of the time.

What actually distinguishes the 35% that drop 5%+ from the 65% that consolidate? We tested every indicator in the system. Here are the results, ranked by predictive edge over the 35% base rate.

Signal Predictive Power
Drop rate when signal present vs. 35% base rate · 118 episodes

The killer combination: 4h fast cloud flips red while price was >2% above the 21 EMA at the PO peak. This has predicted a 5%+ drop 14 out of 14 times over 25 years. When the uptrend has stretched that far and the short-term trend structure breaks, it doesn't recover.

Notice what does not predict the drop: the daily PO zone. Whether the daily PO is also extended (above 80) or not makes no difference — 33% vs 35% drop rate. Both timeframes being overbought is not more dangerous than just the 4h.

Every real 5%+ reversal from an extended 4h PO followed the same sequence of breakdown events. The order is remarkably consistent across 41 episodes spanning 25 years.

1
Price closes below 4h EMA8
~5 bars after PO peak (20 hours)
The first crack. By itself, this happens in both benign and dangerous cases — 37% drop rate vs 24% when absent. This is the "heads up" signal.
2
Price closes below 4h EMA21
~11 bars after PO peak (44 hours)
Trend structure is breaking. The 21 EMA is the core pivot — when price falls through it, the Phase Oscillator is typically collapsing. Drop rate: 41% vs 21%.
3
4h fast cloud flips bearish
~15 bars after PO peak (60 hours)
EMA8 crosses below EMA21 — the fast cloud turns red. This is the confirmation signal. When this fires within 10 bars, the drop rate jumps to 67%. When combined with >2% EMA21 stretch at peak: 100% (14/14).
4
Conviction bear arrow
~25 bars after PO peak (100 hours)
EMA13 crosses below EMA48 — the slow cloud confirms the reversal. When this fires within 20 bars: 89% drop rate (8/9). But it's late — by this point, much of the damage is done.

The key insight: the sequence is the signal, not any single event. A close below EMA8 alone is a coin flip. But when it's followed by a close below EMA21 and then a cloud flip — within the expected timeframes — the reversal is underway.

How fast the 4h PO drops in the first 10 bars after peaking is one of the strongest discriminators. A slow, orderly decline means consolidation. A violent collapse means reversal.

PO Decline Speed → Drop Probability
4h PO change in first 10 bars after peak
Real Reversals
41 episodes with 5%+ drawdown
-68
Median PO drop in 10 bars
Mean PO drop-77 pts
Avg drawdown-9.7%
Benign Consolidation
52 episodes with <3% drawdown
-43
Median PO drop in 10 bars
Mean PO drop-47 pts
Avg drawdown-1.5%

Every signal tested, with sample sizes and edge. The base rate is 35% (41 of 118 episodes led to 5%+ drawdown). Signals are ranked by edge — the difference in drop rate when the signal is present vs absent.

Signal Present n Absent n Edge Avg DD
Cloud flip + EMA21 stretch >2% 100% 14 26% 104 +74pp -11.5%
Conviction bear within 20 bars 89% 9 30% 109 +59pp -11.9%
PO crashes >100 pts in 10 bars 71% 14 30% 104 +42pp -9.1%
Cloud flip + PO drop >80 67% 24 27% 94 +40pp -8.2%
EMA21 distance >2% at peak 61% 36 23% 82 +38pp -7.8%
EMA8 distance >1% at peak 61% 36 23% 82 +38pp -7.9%
PO drops >80 pts in 10 bars 64% 28 26% 90 +39pp -7.8%
Cloud flip within 10 bars 67% 12 31% 106 +36pp -9.1%
Cloud flip within 20 bars 52% 46 24% 72 +29pp -7.0%
PO drops >50 pts in 10 bars 45% 71 19% 47 +26pp -5.8%
Close below EMA21 within 20 bars 41% 80 21% 38 +20pp -5.9%
4h PO peak > 100 46% 24 32% 94 +14pp -6.1%
Close below EMA8 within 10 bars 37% 97 24% 21 +13pp -5.2%
Daily PO > 80 (both TFs extended) 33% 39 35% 79 -2pp -4.7%

When the 4h Phase Oscillator is extended above 80–110, here's what the data says to watch for.

Don't Sell the Extension
The PO being extended is not a short signal. 65% of the time it consolidates sideways, the 8 EMA catches up, and the trend resumes. When PO peaked above 110 and crossed below 100, forward 10-bar returns were positive 100% of the time.
Watch for the Sequence
Real reversals follow a consistent order: close below EMA8 → close below EMA21 → fast cloud flip → conviction bear. If the first step doesn't happen within ~5 bars, it's probably consolidation. If it does, watch for escalation.
The Kill Shot: Cloud Flip + Stretch
If the 4h fast cloud flips red and price was >2% above EMA21 when the PO peaked, every single occurrence in 25 years produced a 5%+ drawdown (14 for 14). Average drawdown: -11.5%. This is the highest-conviction reversal signal in the dataset.
Velocity Matters
A slow PO decline (<50 pts in 10 bars) means consolidation — only 19% drop rate. A violent PO collapse (>100 pts in 10 bars) means reversal — 71% drop rate. The speed of the unwind, not just the direction, tells you which one it is.

Complete log of all 41 episodes where the 4h PO peaked above 80 and a 5%+ drawdown followed. Each row shows the PO peak, the maximum drawdown, and the sequence of warning signals with their timing (bars after peak).

Date Peak PO Max DD Signal Sequence (bars after peak)