The QuantEngines scanner screens markets for seasonal patterns and reports verdicts that are corrected for multiple testing, including patterns that fail. A seasonal tendency is a historical observation, not a forecast, and transaction costs can erase small edges. Educational tool, not financial advice.
Sources
Quant Scanner
Screen monthly seasonal patterns across a fixed universe. Every cell reports its sample size, its effect size with a confidence interval, and a verdict corrected for the number of tests run — including the patterns that fail.
| Symbol | Month | n | Difference | 95% interval | p | q | Years against | Verdict |
|---|---|---|---|---|---|---|---|---|
| AAPL | October | 36 | +4.30 pp | [-0.03, 8.59] | 0.0354 | 0.346 | 10/36 | Weak |
| MSFT | October | 36 | +2.95 pp | [-0.02, 5.97] | 0.0497 | 0.351 | 12/36 | FolkloreMC |
| KO | October | 36 | +1.78 pp | [-0.47, 4.10] | 0.0668 | 0.401 | 11/36 | Folklore |
| PEP | October | 36 | +1.57 pp | [-0.75, 3.75] | 0.0970 | 0.485 | 10/36 | Folklore |
| JNJ | October | 36 | +1.21 pp | [-0.56, 2.96] | 0.1951 | 0.594 | 11/36 | Folklore |
| WMT | October | 36 | +1.39 pp | [-1.00, 3.90] | 0.2060 | 0.594 | 13/36 | Folklore |
| SPY | October | 33 | +0.80 pp | [-1.08, 2.53] | 0.3102 | 0.702 | 12/33 | Folklore |
| JPM | October | 36 | +1.51 pp | [-2.24, 4.97] | 0.3382 | 0.725 | 13/36 | Folklore |
| XOM | October | 36 | +0.61 pp | [-1.51, 3.02] | 0.5413 | 0.869 | 15/36 | Folklore |
| CVX | October | 36 | -0.40 pp | [-2.85, 2.25] | 0.7172 | 0.873 | 17/36 | Folklore |
How to read this
Read the q column, not the p column. A p-value below 0.05 looks convincing on its own, but 120 tests were run together here. The q-value corrects for that. At a false-discovery threshold of 10%, roughly one in ten entries that cleared it would still be expected to be a false positive — which is why we state the threshold rather than just the winners. See our guide to multiple-testing correction and the deflated Sharpe ratio for the statistics behind why this matters.
Years against counts the years in which that month moved opposite to the overall effect. A pattern with a decent average and a third of its years against it is a weaker claim than the average alone suggests, so both are shown.
A cell needs 20 observations to be graded at all and 25 to be eligible for Robust. Cells below the first threshold are reported, never quietly dropped.
Universe: 10 symbols (AAPL, CVX, JNJ, JPM, KO, MSFT, PEP, SPY, WMT, XOM). Data: fmp adjusted end-of-day, 1990-01-02 to 2026-09-28. Methodology version 0.4, generated 2026-09-28T23:32:19+00:00. Survivorship is not corrected for in this universe; that limitation applies to every figure above.
We compute and display. We never recommend. Nothing here is a forecast, and no verdict is advice to buy or sell anything — see the disclaimer.