Not supported · Entries · 2026-10-11

Where the harmonic backtest sits against random buying and SPY

What this means

The backtest's headline gain over SPY is leverage plus a survivor universe, not the pattern. Per trade it matches random buying of the same names. Taken while the stock was a member, it earns +0.196R, below point-in-time random (+0.222R). At ~1× exposure it gives a SPY-like return with shallower drawdowns. That is weak at best and indistinguishable from luck against same-universe randoms. Trust SPY over the backtest.

Explanation checked against the data.

Capped at 10 open positions (realistic): harmonic vs random buying vs SPY.

Uncapped (every signal at 1% risk, ~4–5× exposure).

Mean R per trade: harmonic vs random on today's members vs random on the point-in-time index.

Hypothesis: The harmonic daily system's outperformance over SPY is not explained by survivorship and leverage alone.
Method: The 2,006 actual harmonic daily trades run through the same account as 1,000 random systems on today's members and on the point-in-time index. Uncapped, and capped at 10 (harmonic: 1,000 orderings of same-day signals). Pre-registered decomposition: harmonic lead over exposure-matched SPY = pattern (harmonic − random today's) + survivorship (today's − PIT) + residual.
Result: Mean R: harmonic +0.256 (CI +0.19 to +0.32), random today's +0.278 (diff −0.02, CI −0.09 to +0.05). Uncapped: harmonic 26.2% CAGR at 4.1× vs exposure-matched SPY 35.4%; 10th percentile of today's random; MTM Sharpe 0.72 vs SPY 0.79. Capped: 12.8% (orderings 10.7–14.8%) vs SPY 13.2% / exposure-matched 12.9%; MTM MaxDD −16% vs −34%, MTM Sharpe 0.96. That is the 87th/90th percentile of today's random, not significant. Decomposition, CAGR pp, uncapped / capped: lead −9.2 / −0.1; pattern −8.6 / +2.8; survivorship +10.5 / +1.7. Trades on stocks already in the index at entry: +0.196R; not yet in the index: +0.611R (pit_check).
How much to trust it: Medium: Large random reference, exact trade reproduction. The capped result depends on which ~400 signals fit; one bull decade; no costs.
Did it change the live rule?: no live change; harmonics stay frozen for the forward test (review ~2027-04)
Notes: spy/RESULTS.md; the prereg rule says survivorship explains it uncapped and per trade, and not capped (pattern +2.8pp > survivorship +1.7pp, but n.s.).

Scripts and data for this study will be linked here when the research code is published on GitHub.

← All experiments

Education and research only. Not financial advice. Backtests use today's index members (survivorship bias).