Overfit

32 trend variants: the timing was real, picking the winner was not

· 5 min read

The idea

Hold each ETF while its trend is up (price above its 200-day average, or positive 12-month momentum), otherwise cash. Variants add a market-wide regime gate, a per-fund volatility target, dual momentum (top-k by past return, with cash if it doesn't beat T-bills), and two rebalance speeds. 32 variants.

What happened

Training years 2006–2016 Sharpe Yearly return Worst drop
Best variant (200-day, vol target 10%, monthly) 0.70 4.35% 5.3%
SPY buy and hold 0.41 7.50% 55.2%
SPY at the strategy's own day-by-day invested level 0.63 4.90% 10.2%

The best variant beat every benchmark at default and doubled costs. Costs barely mattered (0.03–0.24% a year).

The full results

All 32 variants, at default costs, on the 2006–2016 training window (2005 is warm-up for the 200-day and 252-day signals). Sharpe, yearly return and worst drop credit idle cash at the 3-month T-bill rate, the same basis as the headline numbers above. Nothing here was run on later years.

Histogram of the Sharpe ratios of all 32 trend variants, with SPY buy and hold, the median variant and the best variant marked.
The best variant (0.70) sits at the top of a pack: the median variant scored 0.51, and 28 of 32 variants scored above SPY buy and hold (0.41). The 32 versions are close substitutes, which is why picking the top one tells you little.
VariantSharpeYearly returnWorst dropTime investedCost drag a year
200-day average · vol target 10% · monthly Best0.704.35%5.3%43%0.04%
200-day average · SPY gate · vol target 10% · monthly0.694.25%5.3%41%0.04%
200-day average · vol target 10% · every 10 days0.674.01%5.0%42%0.05%
200-day average · SPY gate · vol target 10% · every 10 days0.663.86%5.6%40%0.05%
12-1 month momentum · SPY gate · vol target 10% · monthly0.654.16%5.9%39%0.03%
12-1 month momentum · vol target 10% · monthly0.624.31%6.2%43%0.03%
200-day average · monthly0.605.73%12.2%62%0.05%
12-1 month momentum · SPY gate · monthly0.605.43%11.3%55%0.04%
All 32 variants, best first
VariantSharpeYearly returnWorst dropTime investedCost drag a year
200-day average · vol target 10% · monthly Best0.704.35%5.3%43%0.04%
200-day average · SPY gate · vol target 10% · monthly0.694.25%5.3%41%0.04%
200-day average · vol target 10% · every 10 days0.674.01%5.0%42%0.05%
200-day average · SPY gate · vol target 10% · every 10 days0.663.86%5.6%40%0.05%
12-1 month momentum · SPY gate · vol target 10% · monthly0.654.16%5.9%39%0.03%
12-1 month momentum · vol target 10% · monthly0.624.31%6.2%43%0.03%
200-day average · monthly0.605.73%12.2%62%0.05%
12-1 month momentum · SPY gate · monthly0.605.43%11.3%55%0.04%
200-day average · SPY gate · monthly0.595.45%10.5%58%0.05%
200-day average · SPY gate · every 10 days0.575.15%8.6%57%0.08%
200-day average · every 10 days0.575.34%12.2%62%0.08%
12-1 month momentum · vol target 10% · every 10 days0.574.03%6.4%43%0.03%
12-1 month momentum · monthly0.565.75%13.7%62%0.04%
Dual momentum, top 5 by 252-day return · SPY gate · monthly0.537.21%18.1%83%0.13%
12-1 month momentum · every 10 days0.535.45%12.5%62%0.05%
Dual momentum, top 5 by 126-day return · SPY gate · monthly0.527.18%16.8%82%0.15%
Dual momentum, top 5 by 126-day return · SPY gate · every 10 days0.516.76%16.8%79%0.24%
Dual momentum, top 5 by 126-day return · monthly0.517.26%26.4%89%0.15%
Dual momentum, top 3 by 126-day return · SPY gate · every 10 days0.516.69%15.8%76%0.23%
Dual momentum, top 3 by 252-day return · SPY gate · every 10 days0.496.22%16.4%78%0.21%
12-1 month momentum · SPY gate · vol target 10% · every 10 days0.493.18%5.8%37%0.04%
Dual momentum, top 5 by 252-day return · monthly0.466.75%20.3%91%0.12%
Dual momentum, top 3 by 252-day return · SPY gate · monthly0.466.17%17.8%79%0.14%
Dual momentum, top 5 by 252-day return · SPY gate · every 10 days0.465.91%16.3%80%0.19%
Dual momentum, top 5 by 252-day return · every 10 days0.446.38%19.0%89%0.17%
Dual momentum, top 5 by 126-day return · every 10 days0.446.25%30.6%87%0.22%
Dual momentum, top 3 by 252-day return · every 10 days0.436.00%19.3%80%0.18%
12-1 month momentum · SPY gate · every 10 days0.413.77%10.4%52%0.06%
Dual momentum, top 3 by 126-day return · SPY gate · monthly0.395.44%19.5%77%0.16%
Dual momentum, top 3 by 126-day return · monthly0.395.54%22.9%80%0.15%
Dual momentum, top 3 by 252-day return · monthly0.334.60%19.3%80%0.12%
Dual momentum, top 3 by 126-day return · every 10 days0.304.27%32.4%79%0.22%
Growth of one dollar and drawdown from peak for the best trend variant, SPY held on the same days, and fully invested SPY, 2006–2016.
The best variant over the training years. The gap in worst drop is mostly cash: the variant was invested about 43% of the time. SPY held on the same days comes close (Sharpe 0.63, worst drop 10.2%), which is the page's point: timing was real in the search years, while the choice among 32 versions is mostly noise.

This search did not include a scrambled-data run. Its overfitting checks were the deflated Sharpe (0.945 against a 0.95 bar) and the probability of backtest overfitting (0.619 against the rule's 0.50 ceiling). At default costs, 18 of the 32 variants had a higher Sharpe than both matched-SPY comparisons (same invested days, same volatility).

Why it failed

The rule we wrote beforehand said the best variant must also keep the family's probability of overfitting (how often a search winner is only average later) at or below 0.50. It came in at 0.619: in 62% of re-splits, the in-sample winner landed below the later median. The deflated Sharpe was 0.945 against a 0.95 bar.

The honest reading: trend timing is real in the search years (SPY held on the same days reaches a Sharpe of 0.63 with a 10% worst drop), but the 32 versions are close substitutes, and choosing the “best” one among them is mostly noise. Per the rule, nothing went on to later years. The plain, untuned version of this idea has its own page.

Receipts: the strategy family, the sweep, and the credibility rule were written down before the run. 32 trials plus 64 cost-sensitivity runs.