TenachineAI-Powered BacktestingSign in
Guide · 6 min read

Triple ma pullback's average return is positive, its median ticker lost money

Triple ma pullback's headline average return is positive. Its median ticker lost money. Neither number is wrong, and the gap between them is wider than the usual mean-pulled-up-by-winners story, because here the sign itself flips.

Triple ma pullback's average return across the 214 tickers that completed a run is positive 0.88 percent. Its median ticker lost 0.30 percent. Both numbers come from the same study, the same run, the same ten years. Neither one is a mistake.

Sort the 214 tickers by total return and the split is close to a coin flip: 45.3 percent lost money, 44.4 percent gained, and 10.3 percent landed within 1 percent of flat. What is not symmetric is the size of the outcomes on either side. The worst ticker in the study lost 20.93 percent. The best gained 32.88 percent. A wider, longer right tail is enough to pull the average above zero even while the ticker in the exact middle of the pack sits just below it.

Triple ma pullback, 214 completed tickers, by return bucket
Triple ma pullback, 214 completed tickers, by return bucketShare of tickers by total return bucket: 8.9 percent lost more than 10 percent, 15.0 percent lost 5 to 10 percent, 21.5 percent lost 1 to 5 percent, 10.3 percent finished within 1 percent of flat, 15.9 percent gained 1 to 5 percent, 14.0 percent gained 5 to 10 percent, 14.5 percent gained more than 10 percent.Lost more than 10%8.9%Lost 5% to 10%15.0%Lost 1% to 5%21.5%Flat, within 1%10.3%Gained 1% to 5%15.9%Gained 5% to 10%14.0%Gained more than 10%14.5%
Add the four loss-or-flat buckets and 55.6 percent of tickers did not clear 1 percent. The 0.88 percent average is carried by the 14.5 percent of tickers that gained more than 10 percent, including one at 32.88 percent.

A different 34 percent than it looks like

The study's published win rate is 34.19 percent, and it is worth being precise about what that number is counting before it gets mixed up with the bucket chart above. Win rate is computed per trade, across all 12,256 trades the strategy generated over the 214 tickers, about 57.3 trades per ticker. The bucket chart is computed per ticker, on each ticker's single total return over the whole ten years. A ticker can lose on most of its individual trades and still finish the decade in positive territory if its few winning trades were large enough, and 44.4 percent of tickers did exactly that against a 34.19 percent trade-level win rate. The two percentages measuring two different things landing close together is a coincidence worth not reading into, and treating them as the same number would understate how often an individual ticker actually finished ahead.

The aggregate risk to reward ratio, gross profit divided by gross loss across every trade, is 1.0279, just above the 1.0 breakeven line. Tenachine's guide on reading a win rate makes the general case for checking this number next to win rate before judging either one. Here it explains the gap between a 34 percent trade-level win rate and a 44 percent ticker-level gain rate: winning trades pay just enough more than losing trades cost that a minority of winning trades, concentrated in the right ticker, can carry a whole decade into the green.

This is not investment advice, and a positive average is not a promise that a randomly chosen position would have made money. Tenachine's guide on reading a return distribution makes the underlying case with a different study, where the mean also sits above the median but both stay positive. Triple ma pullback is a sharper version of the same warning: the mean and the median did not just disagree here, they landed on opposite sides of zero, and a reader who stopped at the 0.88 percent headline would not have known that the study's own typical outcome was a loss. Checking the median before the average costs one extra number and would have caught it here.