THE BOOK--Playing The Percentages In Baseball

Has anyone made a stat that takes into account RBI opportunities? RBI+? For each of the 24 base-out states calculate the league average RBIs/AB, multiply by the number of ABs the player had with those states, and add these together. Not sure how one would deal with walks but that’s a relatively minor issue.

Then calculate the player’s RBIs divided by that number to get RBI+, so a player who drove in 10% more than league average would have an RBI+ of 110. It might not say all that much more than SLG does but it would be directly linked to real-world RBIs, just adjusted for opportunities.

Yes, that’s been done many times.  Google
Tom Ruane retrosheet Joe Carter

And you’ll get a good article.

@ waiting for spring… the Day by Day DB at baseballmusings.com has a place where you can search for RBI% leaders between two dates. You might be interested in that too.

I thought the author of the blog did a good job pointing out that he does not use RBI’s alone, and also considers saber stats.

RBI’s are simply one record of production, or what happened in games.  You can even take into account opportunities when looking at RBI’s.  Not predictive, but it helps you understand a players contribution to “actual” runs scored for his team.  Not that it says everything about a players contribution.  Teams teams with the most RBI’s usually win the game, and teams with the most runs scored always win the game, so ignoring them is not wise.

Associating RBI’s with pitchers W’s is an often used tactic to discredit them. The difference between RBI’s and pitchers W’s is many.  W’s are an indirect stat.  Pitchers w’s are assigned after a pitcher leaves a game most of the time, and depends on his teams runs scored, when they were scored, how many runs he allowed, and his bullpens run prevention. RBI’s are assigned whenever a run crosses the plate directly as a result of the batters actions.  The only contribution it needs is someone crossing the plate after his PA, and certainly good baserunners deserve some credit for the RBI, and for being on base.  In the case of a solo HR, no other contribution is needed.  It is a more direct stat than pitcher W’s.  Not perfect by any means, but not as bad a stat as pitchers W’s, and certainly now worse than giving a batter who walks credit for 0.3 runs when he does not even score.

I would actually like to see a synthesis of theoretical run calculations with actual run scoring to determine actual run creation by assigning proper credit to everyone.  Theoretical runs created are all well and good, but suffer from their detachment from actual runs created.  Kind of like the difference between ERA and FIP, but we don’t have anything like ERA for hitters (ERA of course suffers from UER so perhaps RA/9 is best).

RE24.

He points out that the AL leaders are a pretty fair set of hitters. I think that is a function of managers placing the best hitters in the middle of the lineup than rather than an indication that RBIs are a measure of hitting ability.

Yes, on the average, high RBI totals means better hitter.  So does high BA. High win total for a pitcher means better pitcher. Etc.

It is just that these are coarse ways of measuring a the skill of either producing or preventing runs/wins for various reasons. It is as simple as that.

Tall people are going to be more successful than short people. That does not mean that measuring someone’s height is a good way of evaluating their success.

Asians, as a whole are smarter than Caucasians, who, in turn, are smarter, as a whole, than African Americans. Again, that doesn’t mean that identifying race is a good way to measure intelligence…

I’ve been meaning to do this for a while, but has anyone produced a regression equation for wins on ERA or WAR (or some “real” measure of run prevention), BA on wOBA, or RBI on WOBA?

It is not that these traditional numbers tell us nothing. It is that they tell us nothing about the player himself that we already don’t know, and more, from better metrics.  The regression lines are going to be fairly flat and the average errors fairly large....