Another dog nightmare for the model, with 9 dogs covering, including 6 straight upsets. FWIW I tweaked algorithm to run a test with full season stats, but it came out the same; so, if past performance numbers suddenly don’t matter, the model’s up a tree without a paddle. I routinely monitor the results of other stat-based services – the kind you pay for – and it’s a small comfort to know they’re doing no better (in some cases worse). Personally, I’ve largely switched to eye-testing team and player matchups for my own plays, but I’m barely breaking even. Hopefully, things change. GLTA
