Newbie here and I'm confused how the normalized data is calculated. Does anyone know the log5 formulas for normalized 2B, 3B, and HR per 100? Based on the research I've done it more or less uses the rate at which a player does an action divided by the rate at which the league does that action that year, multiplied by the rate at which the league does that action in the normalization year. Take a player like 1901 Nap Lajoie, who hit 14 HR's in 544 AB, or ~2.5 HR per 100 AB. According to the data that WIS provides, the average rate of HR in 1901 was 0.6 HR per 100 AB, making Lajoie's HR rate 4 times above league average. Now take 2005 Alex Rodriguez, who hit 48 HR in 605 AB, or ~8 HR per 100AB compared to an average HR rate of 3 per 100AB that year. His HR rate is 2.6 times league average. If normalized HR's are based off of this, shouldn't Nap Lajoie's normalized HR total be greater than A Rod's? Nope, Nap's is 3 HR/100#, while ARod's is 6 HR/100#. Hoping someone can help me out! Thanks.

These are listed in more places, but I know they’re here, so:

https://www.whatifsports.com/forums/Posts.aspx?topicID=519598&threadID=11831758#l_11831758

Thank you, that's super helpful in figuring out how the algorithm works. I've updated my formulas to account for H/AB, HR/H, 3B/H, and 2B/H and they are much more accurate now. Is there any further info on how the regression uses ERA to determine doubles or triples? Does the engine treat it as likelihood of giving up an XBH, or does it subdivide it into 3B rate and 2B rate for pitcher, and is there a known coefficient?

I’m fairly confident it breaks it down to the pitcher level for the formula on a 2B/H and 3B/H level to complete the formula as shown in the slides.

I started to try to find the coefficient or at least an approximation of it at one point, but never finished the process.

if you wanted to try to map out your own regression, baseball prospectus has actual 2B and 3B allowed rates for all pitchers from 1954 on, so you could use that as a solid baseline for the relationship between 2B/H and 3B/H to ERA, for estimating their coefficient.