Not supported · Instruments & watchlist · 2026-09-26 (15m) / 2026-10-10 (all TFs)
Are some stocks reliably better harmonic stocks than others?
What this means
A stock's harmonic results barely agree with themselves: the average R from its odd-numbered setups has a correlation of about 0 with the average from its even-numbered setups on 1h, 4h and daily. Almost none of the spread between stocks is more than sampling noise (about 8% on 1h, 0% elsewhere). The 15m study estimated that about 540 setups per stock would be needed to rank stocks reliably, far more than any stock has. Whether some stocks really are better cannot be measured at these sample sizes.
Explanation checked against the data.
By timeframe
| Timeframe | Verdict | n | Result |
|---|---|---|---|
| 15-minute | Not supported | 0% real variance | |
| 1-hour | Not supported | ~8% real variance | |
| 4-hour | Not supported | 0% | |
| Daily | Not supported | 0% |
- Hypothesis
- Per-instrument performance reflects a stable instrument quality rather than sampling noise.
- Method
- Variance decomposition (observed variance of instrument means vs that expected from sampling) and split-half reliability (odd/even setups; first/second half).
- Result
- Share of variance that is 'real': 15m 0%, 1h ~8%, 4h 0%, 1d 0%. Odd/even correlation 0.013 / 0.000 / -0.09 / -0.08. 15m study: split-half reliability 0.029 at 33 setups per instrument; ~540 setups would be needed for reliability 0.5.
- How much to trust it
- High: large samples, standard psychometric method
- Did it change the live rule?
- No.
- Notes
- Correct reading: instrument quality is unmeasurable at these sample sizes, not proven absent.
Scripts and data for this study will be linked here when the research code is published on GitHub.
Education and research only. Not financial advice. Backtests use today's index members (survivorship bias).