These are MATLAB plots.

LikeLike

]]>LikeLike

]]>Use a multilayer regression, or in fancier terms but all the same: a neural network. You can control the loss function using a lagrangian penalization term.

LikeLike

]]>What I know is that points follow CMP and threes also CMP.

How could someone simulate threes given that they have some results already from simulating points?

Example:

Player A: points~ CMP(l1,n1) and threes ~ CMP(l2,n2)

Points are a cmp mixture of threes and ones or twos combined.

If a simulate 25 points for a match, how can I simulate threes?

And overall respect points and threes distributions?

LikeLike

]]>LikeLike

]]>Also, are there any websites for advanced stats,.for the euro league or the cba?

LikeLike

]]>Thanks man, and also cool website. I’m a total novice with stats. In the most basic terms. Would you say ridge is better than linear forecasting a players performqce or teams?

LikeLike

]]>You can get RAPM-like estimates over on nbashtocharts.com. I say RAPM-like only because the partitioning function creates some influential inputs into the ridge regression. Aside from that, overall, you can get your fix of updated RAPM estimates.

LikeLike

]]>LikeLike

]]>LikeLike

]]>