Golden cross captured 28.5 percent of NVDA's 28,357 percent decade
Golden cross missed almost all of NVDA's, TSLA's, and AMD's decade. It also avoided most of WBD's collapse and turned a losing OXY position into a winning one. The 9.02 percent average return is what both halves of that trade look like added together.
NVDA's buy and hold return across the ten years covered by Tenachine's golden cross study is 28,357.38 percent. The golden cross rule itself, trading NVDA over those same ten years, produced two completed trades and a 28.50 percent return. Buying and holding one stock beat trading it with the study's own rule by roughly a thousand times.
That gap is not a bug in the rule or an error in the data. Golden cross buys when the 50 day moving average crosses above the 200 day average and exits when it crosses back below, and a moving average crossover happens after a move has already started and exits sometime after it has already reversed. A stock that goes almost straight up for ten years, the way NVDA did through the AI buildout, gives a crossover system very little to add beyond staying long, and every stop and re-entry along the way gives back a piece of the total move. Two trades and a partial capture of an extraordinary run is closer to how this mechanism is supposed to behave than a full 28,357 percent would be.
The same pattern shows up elsewhere in the study. Buy and hold on TSLA returned 2,754.93 percent across the same ten years, golden cross's own nine trades on TSLA returned 119.41 percent. Buy and hold on AMD returned 4,485.77 percent, golden cross's four trades on AMD returned 0.40 percent. In each case the underlying stock was one of the decade's most extreme performers, and the rule captured a small, inconsistent fraction of what it did.
What the rule gets back for missing the biggest winners
Run the same comparison on the other end of the universe and the picture reverses. Buy and hold on WBD lost 69.52 percent over the same window. Golden cross's ten trades on WBD lost only 5.24 percent, a fraction of the buy and hold loss. Buy and hold on SLB lost 41.74 percent. Golden cross's six trades on SLB lost 8.72 percent. Buy and hold on DVN lost 25.07 percent, and golden cross's five trades on DVN gained 12.08 percent instead. Buy and hold on OXY lost 18.12 percent, and golden cross's six trades on OXY gained 27.86 percent, turning a losing position into a winning one.
The study's own headline number, a 9.02 percent average return across 193 completed tickers, blends both halves of this trade: a near total miss of the decade's most extreme winners, NVDA, TSLA, and AMD among them, and a partial rescue from some of its worst losers, WBD, SLB, and DVN among them. Tenachine's guide on reading a return distribution already shows that the 9.02 percent average sits well above the 5.51 percent median, evidence that a handful of tickers pull the average upward. This is a different cut of the same study: not just that a few tickers carried the average, but that the rule structurally could not have captured most of what those few tickers actually did, and was never trying to.
This does not make golden cross a bad rule or a good one. A trend follower that fully captured NVDA's decade would also have had to fully ride out every one of the decade's worst collapses, since the same lag that clips a winner also clips a loser. Past performance does not predict future results, and a rule that missed one AI boom is not guaranteed to miss, or catch, the next one.