Models & methods
Ensemble
An ensemble combines more than one model — an average, a median, a weighted stack — so no single architecture has to be right on every week. Tournaments that pick a winner are a hard ensemble; blending is a soft one.
Simple averages of decent, uncorrelated models are embarrassingly strong. Complicated stacks overfit the bake-off.
In production, an ensemble is a serving contract: you must run every member, or distill. Latency and cost are part of the accuracy number, whether the paper said so or not.
A tournament that picks one winner is a hard ensemble. A median of three is a soft one. Use whichever you can actually operate at 6am.
Same folds, five candidates
Seasonal naive
16.4%
Smoothing
14.1%
Gradient boost
11.2%
Foundation
9.7%
Fine-tuned
10.1%