Joel 交易决策基准：Jev 1.13 跑输持有 QQQ

Joel 🌊 @joelsstafford原推发布于 2026.10.13

在 X 上打开原推

I built a new benchmark (Joel's Trading Benchmark v1) for AI decision models, based on my test Trading bot (Nero DX) data collected in 2026 Each model gets $1000 and makes 90 real world trading decisions: which asset to hold right now, or none. Same data, real fees and zero future leakage Perfect decisions would have turned $1000 into $3946 Results: Microsoft-Decision-1: $1064 Jev 1.13: $1031 (Just holding QQQ with no decisions at all: $1038) The best model captured 2.2% of what was possible Which decision model should I test next?