Supported · Instruments & watchlist · 2026-10-11
How much does survivorship flatter harmonic results and the edge over random?
What this means
Survivorship adds at least ~0.07R per trade to the level of R. It also inflates a dip-timing edge somewhat, by ~0.1R in leavers vs survivors, weighted by the leavers' small share. The best harmonic trades are on stocks that joined the index later; on stocks already in the index the daily edge is only +0.03. A true PIT harmonic test would need TradingView runs on delisted tickers, which is often impossible.
Explanation checked against the data.
- Hypothesis
- Testing on today's index members inflates the strategy's R, but the harmonic-minus-random edge is largely immune because both legs share the same stocks and periods.
- Method
- tvpet/pit point-in-time S&P 500 membership plus Yahoo daily prices (109 of 244 post-2016 leavers have prices). No harmonic setups exist for leavers, since the TradingView sweeps cover today's list only. Three checks: (1) harmonic trades split by index membership on the entry date; (2) random entries (same stop %/2R) on PIT members vs survivors; (3) a mechanical dip-buy proxy vs +-60-day randoms on survivors vs leavers.
- Result
- (1) Daily, member at entry: harmonic +0.20, edge (fixed control) +0.03 (-0.02,+0.08), n=1,718. Not yet a member: +0.61, edge +0.17 (+0.06,+0.29), n=288. 1h members: edge +0.05 (+0.01,+0.10). (2) Random R +0.274 on survivors vs +0.202 on PIT members. (3) Dip-proxy edge +0.38 on survivors vs +0.28 on leavers.
- How much to trust it
- Medium: PIT membership is solid, but the 135 leavers without prices (mostly takeovers/bankruptcies) are missing, so the bias is underestimated
- Did it change the live rule?
- none
- Notes
- Drift share (random R / harmonic R): 1d 107% (original control) / 81% (fixed); 1h 92% / 65%; 4h 272% / 85%.
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).