Gold's December-to-February rally, tested on 40 years: 25 of 40Gold FuturesCOMEX_DL:GC1!MarketTrendsForFriendsEvery December the same story comes back: gold rallies from mid-December into February. With gold just under its June low (Friday 3,898.9 against 3,955.4), I expect to hear it early this year. Maybe. I wrote the test down at 17:18 on Saturday, before computing anything: buy the second Friday close of December, sell the last Friday close of February, every season since 1986, against any 11-week stretch as the control. To call it high probability I set the bar at 70 pct of seasons up, 15 points better than the control, and both halves of the record agreeing. It failed. Gold rose in 25 of 40 seasons (62 pct), against 54 pct for any 11-week stretch. The median season gained 2.0 pct, against 0.6. Better than nothing; not high probability. The record splits in two. From 1986 to 2005: 9 of 20 up, median -0.9 pct, a little worse than an ordinary stretch. From 2006 to 2025: 16 of 20 up, median +6.1. The famous seasonal is a 20-year pattern inside a 40-year record, and which half applies to 2026 is a judgment, not a test. The seasons that began like this one, with gold under a falling 40-week average: 6 of 13 up (1 of 7 before 2006, 5 of 6 since). Forced into a trade that pays 2.5 to 1 (stop 5 pct under the entry, target 12.5 pct over it), 4 of 40 seasons reached the target first and 9 hit the stop first. The last season was the best of the 40: +21.2 pct, from 4,328.3 on Dec 12, 2025 to 5,247.9 on Feb 27, 2026, the highest weekly close on record. One season is an anecdote, not a base rate. Silver's same window is steadier: 25 of 40 up against 50 pct for any stretch, and both halves agree (11 of 20, then 14 of 20). Still not 70. So there is no high-probability trade here, and I am not posting one. The calendar is a mild tailwind, not a trigger. The live window runs from Dec 11, two days after the Fed decides, to Feb 26. Inside it I still wait for a setup: a new weekly low that holds, then a Friday close above the falling 10-week average at a price that pays 2.5 to 1 to the first target. At flat prices that geometry arrives in early December. That setup is untested for gold; my model book's tested gold rule waits longer, for a Friday close above the 40-week (about 4,495). Nothing new while France's 10-year spread over Germany is above 1.75 with the euro under 1.1113, or while the S&P 500 closes a week under 7,508 with high yield above 3.50. What would change my mind: a Friday close back above 3,955.4 by Nov 13. On the record that reclaim moves the stop closer; it does not improve the odds six months on. Want to learn more? Contact me in chat, i teach and provide free content. Research and education only. Never financial advice.