How do I make RBI into a useful statistic?
June 10, 2011 9:11 AM Subscribe
Rather than just Runs Batted In, I would like to be able to find out how successful a player is in a given RBI opportunity. Is there a way to discover (for example) what percentage of the time a player drives in the run when there is a runner on third and none out (or one out...or two out). I'm aware of the existence of B-ref, but I wasn't able to figure out how to find this type of data.
Ultimately, I'd like to see a statistic that measured RBI/RBI opportunity which is weighted from easiest/least (man on third none out) to hardest/most (2 out grand slam). Is there anything like this out there in the Sabersphere?
Ultimately, I'd like to see a statistic that measured RBI/RBI opportunity which is weighted from easiest/least (man on third none out) to hardest/most (2 out grand slam). Is there anything like this out there in the Sabersphere?
The good folks over at Baseball Prospectus appear to have kept track of this in a RBI Opportunity Report for a few years (2005 was the latest I could find). The real answer, though, and the reason the thing you're talking about isn't more widely tracked in "the sabersphere," is that hitting with runners on base is simply not a separate skill from hitting. By far the best explanation for any difference between a player's batting stats with bases empty and batting stats with runners on is simply random variation. RBI and RBI/opportunity don't measure anything meaningful about a hitter that can't be better captured in other ways.
posted by RogerB at 9:47 AM on June 10, 2011
posted by RogerB at 9:47 AM on June 10, 2011
So, to answer the question in your title: How do you make RBI into a useful statistic? Ignore it. Like pitching W-L record, it's a hopeless muddle of team and random factors masquerading as a measurement of individual ability.
posted by RogerB at 9:49 AM on June 10, 2011
posted by RogerB at 9:49 AM on June 10, 2011
Response by poster: hitting with runners on base is simply not a separate skill from hitting
I get this. Here's what's bugging me. Sometimes I feel that the best "hitter" may not be the best guy to have at the plate in a given situation. If there's a man on third and one out, I don't think I want Jim Thome or Adam Dunn at bat if it's a critical run. Yes they are more likely to hit a two run homer every so often, but if that is a critical run (let's say to tie the game in a late inning), don't I want someone at the plate who is more likely to put the ball in play? Mightn't I even prefer the loathsome Juan Pierre? (OK, not Juan Pierre, but someone who might consistently get the ball at least into the outfield)
posted by Carlo at 10:11 AM on June 10, 2011
I get this. Here's what's bugging me. Sometimes I feel that the best "hitter" may not be the best guy to have at the plate in a given situation. If there's a man on third and one out, I don't think I want Jim Thome or Adam Dunn at bat if it's a critical run. Yes they are more likely to hit a two run homer every so often, but if that is a critical run (let's say to tie the game in a late inning), don't I want someone at the plate who is more likely to put the ball in play? Mightn't I even prefer the loathsome Juan Pierre? (OK, not Juan Pierre, but someone who might consistently get the ball at least into the outfield)
posted by Carlo at 10:11 AM on June 10, 2011
Yes, your intuition is correct. There are some game situations in which a better chance at scoring one run is more valuable than a chance at scoring multiple runs, while the general saber-situational wisdom is based on just increasing the average run expectancy (the mean number of runs that we'd expect to be scored) in a way that doesn't take account of the value of, e.g., the tie-breaking run. There's a good chapter about this, "When Is One Run Worth More Than Two?", in Baseball Between the Numbers. And apparently Nate Silver invented a number called ORVY — One-Run Value Yield — to help quantify when it's a good idea to play for one run rather than many.
Though it must be out there somewhere, I don't know of anything directly tying this to the various batting stats as you're doing (which would be useful in selecting pinch-hitters, for instance) but it's at least intuitively clear that, in one-run situations, K rate becomes a more important factor.
posted by RogerB at 10:32 AM on June 10, 2011
Though it must be out there somewhere, I don't know of anything directly tying this to the various batting stats as you're doing (which would be useful in selecting pinch-hitters, for instance) but it's at least intuitively clear that, in one-run situations, K rate becomes a more important factor.
posted by RogerB at 10:32 AM on June 10, 2011
The real answer, though, and the reason the thing you're talking about isn't more widely tracked in "the sabersphere," is that hitting with runners on base is simply not a separate skill from hitting. By far the best explanation for any difference between a player's batting stats with bases empty and batting stats with runners on is simply random variation.
This is probably true, but it's not a given. At bats are not completely independent trials. A lead-off type hitter might not be able to drop a bunt for a hit, or a line-drive type hitter might have to try to hit a fly ball to get a sacrifice fly. And the pitcher might throw different sorts of pitches, such being more likely to go for a strikeout than to just throw strikes to keep their pitchcount down, or the manager would be much more likely to bring in a situational reliever for a good matchup against a particular hitter when runners are in scoring position. If batting is substantially more difficult in at bats that are inherently more likely to result in runs being scored, then looking at total batting stats would favor batters who bat in a smaller percentage of those situations or who perform better in the easier but less important situations.
posted by burnmp3s at 10:55 AM on June 10, 2011
This is probably true, but it's not a given. At bats are not completely independent trials. A lead-off type hitter might not be able to drop a bunt for a hit, or a line-drive type hitter might have to try to hit a fly ball to get a sacrifice fly. And the pitcher might throw different sorts of pitches, such being more likely to go for a strikeout than to just throw strikes to keep their pitchcount down, or the manager would be much more likely to bring in a situational reliever for a good matchup against a particular hitter when runners are in scoring position. If batting is substantially more difficult in at bats that are inherently more likely to result in runs being scored, then looking at total batting stats would favor batters who bat in a smaller percentage of those situations or who perform better in the easier but less important situations.
posted by burnmp3s at 10:55 AM on June 10, 2011
A quick answer from my household baseball stats person:
There's no single stat that tracks this. What you're asking is too difficult to figure out because it's not specific enough -- for example, how do you want it weighted and why do you want those weights? (It's a thing where you'd need to do some research to find out the useful weights.)
You can measure increased run expectancy, but that seems like it doesn't answer your question. The question you're asking depends a lot on what the score is - eg if it's the bottom of the ninth, do you need one run or two to win?
Run expectancy is a general stat (not individual): Tangotiger has done some work on this. For example here's a chart showing run expectancies from 1999-2002 (a different offensive environment so these numbers wouldn't apply directly to today) -- it shows, what is the chance that one run will score, two runs will score, etc for different game situations.
You could combine individual players' likelihood of hitting eg a single or ground ball etc, with the run expectancy for a situation, to see which player will fare better in the situation.
posted by LobsterMitten at 7:23 PM on June 10, 2011
There's no single stat that tracks this. What you're asking is too difficult to figure out because it's not specific enough -- for example, how do you want it weighted and why do you want those weights? (It's a thing where you'd need to do some research to find out the useful weights.)
You can measure increased run expectancy, but that seems like it doesn't answer your question. The question you're asking depends a lot on what the score is - eg if it's the bottom of the ninth, do you need one run or two to win?
Run expectancy is a general stat (not individual): Tangotiger has done some work on this. For example here's a chart showing run expectancies from 1999-2002 (a different offensive environment so these numbers wouldn't apply directly to today) -- it shows, what is the chance that one run will score, two runs will score, etc for different game situations.
You could combine individual players' likelihood of hitting eg a single or ground ball etc, with the run expectancy for a situation, to see which player will fare better in the situation.
posted by LobsterMitten at 7:23 PM on June 10, 2011
Response by poster: I think the first part of my question CAN be answered specifically since it does not assign a value to the situation (in a given situation, what percentage of the time does an individual batter drive in the run...or not). I'd like to be able to look it up somewhere, but it doesn't seem to be an option.
The second part (weighted value to RBI situations) seems more of a subjective endeavor, though the win probability added referred to in the ORVY article that RogerB referenced might be a good place to start.
Thanks to everyone for the thoughts.
posted by Carlo at 8:33 PM on June 10, 2011
The second part (weighted value to RBI situations) seems more of a subjective endeavor, though the win probability added referred to in the ORVY article that RogerB referenced might be a good place to start.
Thanks to everyone for the thoughts.
posted by Carlo at 8:33 PM on June 10, 2011
This thread is closed to new comments.
posted by burnmp3s at 9:16 AM on June 10, 2011