MU's average trade lost money. Its full 69-trade record gained 11.54 percent

MU's average trade under donchian 20 breakout lost 0.51 percent. Not a typo, not a rounding artifact: negative 0.5057 percent, the mean of 69 completed trades on Micron over the strategy's ten-year window. The same 69 trades, taken together as one continuous record, returned positive 11.54 percent. A number that says the typical trade lost money sits inside a record that made money.

Why the two numbers do not have to agree

An average trade return is a simple mean: add up 69 individual percentage returns and divide by 69. A full record return is compounded: each trade's dollar result feeds into the account balance the next trade sizes off of, in whatever order the 69 trades actually happened. Those are two different pieces of arithmetic run over the same data, and nothing requires a negative mean to produce a negative compounded total, or a positive one to produce a positive total. MU's win rate under this strategy is 33.33 percent, meaning roughly two losses for every win, and its worst single trade lost 18.09 percent against a best trade of 29.95 percent. A record built mostly of small losses with a handful of larger wins can average out negative per trade while still compounding to a gain, if enough of the account's growth happened to concentrate in the winning minority at points where the position size was large enough to matter.

That is a plausible mechanism, not a proven one. Tenachine does not publish the order in which MU's 69 trades occurred or how large each position was in dollar terms, so which specific trades did the compounding and when is a real unknown here, not a detail being glossed over. What is published and verified is the pairing itself: a negative average trade and a positive full record, on the same completed backtest.

The mirror case: a winning average trade inside a losing record

APA under ADX trend breakout runs the pairing the other way. Its average trade won 1.0736 percent, a genuinely positive number across 26 completed trades. Its full record, the same 26 trades compounded together, lost 2.9459 percent. Tenachine's guide on APA under a different strategy already covers this ticker as the site's widest strategy-versus-buy-and-hold gap, under week 52 low reversion rather than ADX trend breakout. This is the same ticker under a separate rule, showing a separate kind of gap: not strategy against buy and hold, but a trade-level average against the account's own compounded outcome.

APA's win rate is 30.77 percent, close to MU's 33.33 percent, and its worst trade, negative 11.61 percent, is shallower than its best, 44.40 percent. On the trade-level numbers alone, APA looks like the stronger case: a higher average trade, a best trade nearly four times its own worst. The full record still lost money. A strategy where most trades are small losses and the average is rescued by a few large wins is exactly the shape that makes the compounded order of those wins and losses matter more than their simple average, and here that order worked against the account rather than for it.

What to check before repeating either number

An average trade return answers one question: across every completed trade, what did the typical one do. A full record return answers a different question: what did the account actually end up with. Both are real, both are published, and reading only one would have given an incomplete answer for either ticker here. MU's negative 0.51 percent average trade would have suggested a losing strategy on a stock that, compounded, gained 11.54 percent. APA's positive 1.07 percent average trade would have suggested a winning one on a stock that, compounded, lost 2.95 percent. Checking both figures, not just the one that happens to appear first on a results page, is the only way to catch a pairing like either of these before repeating the wrong one.

Neither pairing says anything about donchian 20 breakout or ADX trend breakout as strategies in general, and neither is a signal to trade MU or APA specifically. Ten years of one rule applied to one stock is a description of that decade, not a forecast, and a strategy that compounded a losing average into a gain on one ticker is not guaranteed to repeat that trick on the next one. Past performance does not predict future results for either ticker, or for the strategy that traded both.