Brilliant breakdown on how in-system digs actually create offensive leverage. The 70-point gap in win probability between low and high in-system dig percentage is wild, but it tracks with what I've seen coaching at lower levels where teams obsess over kill percentages but ignor the quality of transition. Curious if the model tested for interaction effects between freeball generation and opponent attackquality, since forcing a freeball off an elite hitter seems way more valuable than off a middling one.
This is a great question - and your intuition is right.
In this version of the model, effects are estimated at the set level and treated as additive, so a freeball is valued based on how often it creates access to an in-system offense on average, not who it comes from.
You’re absolutely right that context matters. Forcing a freeball off an elite attacker almost certainly carries more downstream value than forcing one off a middling one - but capturing that requires either interaction terms or much finer-grained attack quality inputs than we currently have.
That tradeoff is intentional here. The goal of this piece was to understand what tends to create leverage across the sport, not to optimize for specific matchups. The interaction question you’re pointing to is exactly where this framework becomes team - and opponent -specific.
Brilliant breakdown on how in-system digs actually create offensive leverage. The 70-point gap in win probability between low and high in-system dig percentage is wild, but it tracks with what I've seen coaching at lower levels where teams obsess over kill percentages but ignor the quality of transition. Curious if the model tested for interaction effects between freeball generation and opponent attackquality, since forcing a freeball off an elite hitter seems way more valuable than off a middling one.
This is a great question - and your intuition is right.
In this version of the model, effects are estimated at the set level and treated as additive, so a freeball is valued based on how often it creates access to an in-system offense on average, not who it comes from.
You’re absolutely right that context matters. Forcing a freeball off an elite attacker almost certainly carries more downstream value than forcing one off a middling one - but capturing that requires either interaction terms or much finer-grained attack quality inputs than we currently have.
That tradeoff is intentional here. The goal of this piece was to understand what tends to create leverage across the sport, not to optimize for specific matchups. The interaction question you’re pointing to is exactly where this framework becomes team - and opponent -specific.