The initial goal was to predict XR values for pitchers prior to the cards coming out. So I used MLB 2019 stats and ABL 2020 XR values (all parks). Got ERA, xFIP, & SIERA from Fangraphs. Correlations were fair, around 0.7. Best was xFIP at 0.71 for starters.
Finally found some linear weights at Baseball Savant, namely wOBA. This provided a much better correlation. (Also tried xwOBA, but it was worse.) Used a list of exclusive starters and exclusive shorts. They were fairly different, with shorts having worse correlation. I’m guessing that’s a combination of smaller sample size and deviation from average platoon rates. Did not try longs, closers, or combos. The pitcher estimate equations:
STARTER: XR = 824 * wOBA - 103 R=0.88 SHORT: XR = 601 * wOBA - 33 R=0.80
If I ever had to choose keeps without cards, I would get data for all qualified pitchers, then sort them into categories, when appropriate.
Did batters too. Unsurprisingly, correlation to wOBA was even better. (xwOBA was again worse.) For batters:
XR = 736 * wOBA - 138 R=0.86
- TPB pitcher ratings correlate with linear weights, as opposed to, say, ERA.
- I could get reasonable XR estimates for both batters and pitchers if I didn’t have access to the cards.
The spreadsheet with the data is named ABL Predictions.