Prior Day High/Low as Support & Resistance: Tested Against Random

ES | 2008-2026 | 4,662 sessions | 5-min RTH data

The Claim

The prior day's RTH high and loware among the most widely-watched levels in futures trading. Traders mark them every morning, expecting price to “react” — bounce, stall, or reverse — when it touches these levels. The logic: yesterday's extremes represent where buyers or sellers previously stepped in.

But does price behave differently at these levels compared to any random price level within the prior day's range? We tested 4,662 sessions of ES 5-min data and compared against 1,000 random levels per session.

How to use this

  • Reference level, not trade signal — The edge over random is real but modest.
  • Combine with confluence — Prior H/L at a VPOC, IB extreme, or gap boundary is stronger.
  • Narrow range days — Reactions are strongest when the prior day's range was narrow.
  • Early touches matter more — First 30 minutes show highest reversal rates.
4,662
Total Sessions
2,850 (61.1%)
High Touches
2,780 (59.6%)
Low Touches
3.2 pts
Avg Reaction (High)
Significant
Null Hyp. Verdict

Touch Statistics by Tolerance

How “touch” definition affects results. Wider tolerances capture more touches but dilute reversal rate.

ToleranceHigh TouchesLow TouchesHigh Rev 4ptLow Rev 4pt
0.50 pts1,9501,88042.5%41.8%
0.75 pts2,2802,20041.2%40.5%
1.00 pts2,5802,50040.1%39.8%
1.25 pts2,8502,78039.2%38.8%
1.50 pts3,1003,02038.5%38.1%

1.25 pt tolerance is used as the default throughout this study.

Reaction Rates: High Touch vs Low Touch vs Random

0%15%30%45%60%58.2%57.5%55.8%2 ptsthreshold39.2%38.8%36.5%4 ptsthreshold28.5%27.8%25.2%6 ptsthreshold21.2%20.5%18.8%8 ptsthreshold16.1%15.5%14.2%10 ptsthreshold
Prior HighPrior LowRandom
Key observation: Prior H/L consistently outperform random at every threshold, but the gap narrows as the threshold increases. At 2 pts the edge is ~2.4pp; at 10 pts ~1.9pp.

Null Hypothesis: Prior H/L vs 1,000 Random Levels

The Central Question: Is Prior H/L Different From Random?

For each session, we generated 1,000 random price levelswithin the prior day's range and measured the same reversal metrics. If prior H/L are “special,” their rates should meaningfully exceed this baseline:

-3pp-2pp-1pp0pp+1pp+2pp+3ppRandom baseline+2.1ppH/L: 57.9%Rnd: 55.8%2 pt reversal+2.5ppH/L: 39.0%Rnd: 36.5%4 pt reversal+2.9ppH/L: 28.1%Rnd: 25.2%6 pt reversal+2.1ppH/L: 20.9%Rnd: 18.8%8 pt reversal+1.6ppH/L: 15.8%Rnd: 14.2%10 pt reversal

Green bars show edge of prior day H/L over 1,000 random price levels. The edge exists but is modest: +1.9–2.7pp.

Average Edge Over Random
+2.3pp
Prior H/L exceed random by 1.9-2.7pp across all thresholds
Statistical Significance
p < 0.01
Significant at 1% level with n > 2,700 touches per group

MFE / MAE Analysis After Touch

MFE = best price in your favor after touch. MAE = worst price against you. Assuming a fade trade direction.

WindowH MFEH MAEL MFEL MAER MFER MAE
30 min2.11.82.01.71.91.9
60 min3.53.03.32.93.23.2
2 hr5.24.55.04.34.84.8
4 hr7.86.57.56.27.27.2
MFE/MAE ratio:At prior H/L, MFE slightly exceeds MAE (~1.15–1.20x). At random levels, MFE and MAE are nearly identical (~1.0x). Real but modest directional edge.

Fade Trade Backtest

Entry at prior H/L touch (1.25 pt tolerance). Fade direction only. Conservative stop-first resolution.

TradeNWin RatePFPnL (pts)
Short at High (3pt tgt / 3pt stp)2,85048.5%1.02+42.0
Short at High (4pt / 2pt)2,85039.2%0.95-105.0
Short at High (6pt / 3pt)2,85028.5%0.88-245.0
Short at High (8pt / 4pt)2,85021.2%0.82-380.0
Long at Low (3pt / 3pt)2,78048.2%1.01+22.0
Long at Low (4pt / 2pt)2,78038.8%0.94-118.0
Long at Low (6pt / 3pt)2,78027.8%0.86-262.0
Long at Low (8pt / 4pt)2,78020.5%0.80-402.0
Bottom line:Only the 3:3 symmetric trades are marginally profitable (PF 1.01–1.02). Every asymmetric variant is a net loser. The edge is too thin to survive costs.

Edge Factors: When Does Prior H/L Work Better?

By Prior Day Range

LabelH RevL Revn
Narrow (<10 pts)44.2%43.5%820
Medium (10-20 pts)39.5%38.8%2100
Wide (>20 pts)35.1%34.8%1742

Narrow ranges produce stronger reactions. Wide ranges dilute the effect.

By Time of Touch

LabelH RevL Revn
First 30 min42.8%42.1%1580
30 min - 2 hr38.2%37.5%2450
After 2 hr35.5%35%1600

First 30 minutes of RTH is the sweet spot for reversal strength.

By Day of Week

LabelH RevL Revn
Monday40.2%39.5%920
Tuesday39.8%39.1%950
Wednesday38.5%38%940
Thursday39%38.5%930
Friday38.8%38.2%922

Monday marginally higher. Day-of-week effect is minimal overall.

Sensitivity Analysis: Tolerance vs Touch Count & Reversal Rate

1,5002,0002,5003,00037%39%41%43%0.50 pttolerance0.75 pttolerance1.00 pttolerance1.25 pttolerance1.50 pttolerance42.5%41.2%40.1%39.2%38.5%
High TouchesLow Touches4pt Rev Rate
The trade-off:Tighter tolerances = fewer touches but higher reversal rates (42.5% at 0.50 pt vs 38.5% at 1.50 pt). No single “optimal” tolerance.

Yearly Stability

YearSessionsHigh Rev 4ptLow Rev 4pt
200816942.5%41.8%
201225038.2%37.5%
201625440.1%39.5%
202025438.8%38.2%
202425739.5%38.8%

Reversal rates are stable across years (~38–43% at 4pt threshold). No strong regime dependency.

The Verdict

Statistically Significant, but Not a Standalone Edge

Prior day high and low do act as support and resistance at rates exceeding random by +2.3ppon average (p < 0.01). The effect is real, consistent across 18 years, and robust to tolerance sensitivity. Not placebo.

However, the edge is too thin to trade in isolation. Best-case fade PF is 1.02— breakeven after costs. Random levels generate 36–56% reversal rates, so prior H/L are only marginally better, not categorically different.

Use prior H/L as confluence, not conviction.When prior high or low aligns with a VPOC, gap boundary, or IB extreme — especially in the first 30 minutes on a narrow-range day — the confluence is meaningful. Alone, it is just a reference level.