@vince.quant: Why quants penalize their own models? Ridge regression (Hoerl & Kennard, 1970) adds a penalty on coefficient size. You trade a bit of training accuracy for much more stable out-of-sample predictions. It works best when you have many correlated features, few observations, and low signal-to-noise -> basically most quant finance setups. Limitations: ridge never drops variables, assumes linearity and standardized features, and λ must be chosen carefully (usually cross-validation). It reduces variance but it doesn’t magically fix overfitting. #quant #finance #trading #algotrading #machinelearning

vince.quant

Open In TikTok:

Region: FR

Friday 13 March 2026 22:55:42 GMT