Trading Windows — The ADA Oracle

63.8%

Overall win rate

77.1%

Best window (Sat 2 PM)

51.2%

Worst window (Wed 6 AM)

25.9pp

Edge spread

Volume insight. Low-volume trades (below-median volume ratio) win at 67.7% vs high-volume at 59.9% — a 7.8pp edge that holds across all 5 years, all regimes, and both directions. Use the volume filter below to see which windows are robust vs fragile. The Oracle's vr (volume ratio) field already captures this — a future engine gate could use it.

Statistical disclaimer. Individual cells contain 56–130 trades. Wilson 95% confidence intervals are ±8–11 percentage points wide. The "77.1% Saturday evening" could realistically be anywhere from 66% to 85%. Use these patterns as supplementary context, not as standalone signals. The Oracle's score, tier, and regime remain the primary decision drivers.

Hour × Day Win Rate Heatmap

Click any cell for detailed breakdown. 4-hour CST windows × 7 days. Color intensity = win rate relative to 63.8% baseline.

vr split at median (1.03×)

<52%

52–58%

58–64%

64–70%

70%+

By day of week

Aggregated across all hours

By 4-hour window

Aggregated across all days

Methodology

Data source

All statistics are computed from the Daily Engine's 5-year backtest: 3,453 simulated trades from March 12, 2021 to March 12, 2026. Each trade is assigned to a 4-hour CST window and day of week corresponding to its entry timestamp. A trade is counted as a "win" if its realized PnL is positive (TP hit or profitable exit), and a "loss" otherwise (SL hit or negative exit). The two rightmost columns (6 PM and 10 PM) represent the previous calendar day in CST — e.g., the "Monday 6 PM" cell contains trades that entered Monday at 6 PM CST (which is Tuesday 00:00 UTC).

Confidence intervals

Win rates use the Wilson score interval at the 95% confidence level (z = 1.96), which outperforms the naive normal approximation for small sample sizes. The formula is:

center = (p + z²/2n) / (1 + z²/n)

spread = z × √((p(1-p) + z²/4n) / n) / (1 + z²/n)

Where p = observed proportion, n = sample size, z = 1.96.

Color mapping

Cell color intensity reflects win rate relative to the overall baseline of 63.8%. Cells significantly above baseline appear green; cells near or below baseline shift toward amber and red. The color scale is continuous, not discrete — the legend shows approximate ranges.

Minimum sample size

All 42 cells (7 days × 6 buckets) contain at least 56 trades, exceeding the 30-trade minimum threshold for reliable Wilson intervals. No cells are excluded or masked.

What this does NOT show

This heatmap does not account for market regime, score tier, signal type, or direction. A "high WR" cell may be driven entirely by favorable market conditions during the backtest period rather than a persistent structural edge. Cross-reference with the Oracle's regime detection and tier system before using timing data for trade decisions.