DISCLAIMER: This change went into effect for the SBT Plays section on February 7, and now the All Model Plays section reflects this with the pick dots that use the machine learning methods starting February 9.
For those of you still checking to see what my college basketball picks are every day despite the avalanche of losses recently, you probably noticed that today’s plays(February 7) have mostly negative edges and therefore may be wondering if there is a mistake.
Mistake, there is not. I am experimenting with shifting toward a machine learning based method to select my plays of the day instead of the keys method I have been using. I spent yesterday testing out this machine learning based approach to select plays instead of keys and the results I got were good enough to allow me to shift over. What do I have to lose at this point? The key method hasn’t been working out this season so I figured I might as well give a machine learning based approach a shot instead. So do not be alarmed to see plays being selected despite negative edges.
The benefit of a machine learning based approach is that negative edges can be transformed into value, unlike a key approach where only > 0 edges are used.
I am not saying key based approaches are useless, but its not been working out in my college basketball model this year and I want to try something new.