Friends, Cincinnatians, countrymen, lend me your ears. Too often, lately, have we fought amongst ourselves. We are brothers and sisters. Mothers and fathers and children. We should be friends and not fight amongst ourselves.
Still, there is a rift within the community and it is centered upon none other than Joey Votto, the Prodigal Son. It pits us against one another and must be dealt with. This, fair countrymen, is my modest attempt…
The arguments we have been having are statistical arguments. I know many of you don’t think of yourselves as “stats” people, but you are. What I mean by that is that you have stats that you trust (like average with RISP and RBI) and you think they are better than the stats some of the rest of us trust (OBP, WAR, etc.). The argument we are having is over which of these stats better tells us how good a player is. I hope we can agree on that. If not, then I don’t know how this discussion can move forward.
The argument many, including myself, would make is that RISP and RBI aren’t good stats because though they might tell you what has happened, they don’t tell you how likely something is to happen in the future or how much control a player really had over that.
Now, before I start my real argument, I want everyone to pause and ask yourselves a question: Can you be persuaded? That is, is there anything that can convince you that how someone has hit this season with RISP or how many RBI a player has are not numbers that really tell you much of value about that player. If not, then you aren’t really willing to engage in a debate. You are taking these stats as articles of faith and we can’t have a constructive conversation about it. You should probably leave now. (For the record: Yes, I can be persuaded and am willing to change my mind. Doing so requires statistically significant data.)
Okay, let’s get started…
Clutch Hitting Is Real
If you’ve been paying close attention to what I’ve had to say about clutching hitting lately, you’ll have noticed a lot of qualifiers, “most,” “nearly all,” that kind of thing. This is because clutch hitting does exist. At least, a little.
Brace yourselves, there’s about to be math (though, frankly, I’m an English teacher. If I can handle the math, you ought to be able to).
Still the best discussion of clutch hitting that I am aware of is from a book called Baseball Between the Numbers that the Baseball Prospectus team put out in 2006. The particular chapter is written by Nate Silver. And here, basically, is what he did…
In looking at what constitutes clutch hitting, Silver did not use numbers like average with RISP. The reason here is fairly simply and I’m sure most of you see it now. A solo homer in a tie game in the ninth is way more clutch than a bases clearing double in 14-3 blowout. So he looked closely at those plays that actually affect how likely a team is to win a game. This is called Win Expectancy and you can see a version of it in the Fnagraphs charts that are posted at the end of many recaps here at RN. For instance, here’s yesterday’s game:
You’ll notice that Votto grounding into a double play really killed the Reds chances of winning.
Silver looked at many players and many seasons in compiling his data (this is the value of advanced stats, they let you look with many pairs of eyes instead of just one). And he did find evidence of clutch hitting. In the seasons he looked at (1972-2005), Mark Grace provided the most clutch value with an extra 13.68 wins happening over the course of his career because he hit better in the clutch (it comes out to about one win per 650 PAs). That’s significant, but it’s not gigantic.
The next important point to make here is that there are extreme seasons. For instance, at the time the book was written, David Ortiz had also been good for about one clutch win every 650 PAs, but much of that came from his remarkable 2005 season, when he produced 3.63 clutch wins.
So while clutch hitting exist, what Silver found is that a good clutch hitter enhances his offensive value about as much as a good base runner. It matters, but not a ton. Additionally, the in-season samples are so small that it fluctuates a lot from year to year so that in a given season, only about 10 percent of what you see is actual clutch hitting ability and the rest is chance.
So, has Joey Votto hit poorly in the clutch this year? Yes, he has. In fact, as some of you have pointed out, Fangraphs has a stat showing this. However, in their explanation of the stat, it’s made clear that this doesn’t have a predictive value except over very long periods of time. Therefore, when discussing whether or not Joey Votto has any clutch ability, we should look at his entire career and not 2/3 of a season. What we see then is that, yes, he does seem to have some positive clutch ability.
But don’t read anything into this year. It’s almost all noise and how he has hit in the clutch so far this season tells us so little about his abilities that we can’t make any kind of worthwhile inferences about how he’ll do for the rest of the season.
But there has been some new research since the Silver piece discussed above. Most notably, we have a better sense for how long it takes certain numbers to stabilize. That is, how long does it take before we can reasonably trust a number? You might be balking at this internally, but we all account for this mentally at the beginning of every season. For example, when a player goes 5 for 5 with 2 homers on opening day, no one assume he’s going to bat 1.000 with 324 home runs. That’s all stabilization is. It just tells us when we can actually expect that a player’s numbers have useful meaning.
Over at Fangraphs, you can find an article that tells you how many Plate Appearances a player needs for certain stats to being to stabilize. The threshold it uses is such that in the given amount of plate appearance, you can assume that half of what you see is the player and the other half is chance. This was determined by looking at mammoth amount of data and it’s a good gauge. A fair bit of it “feels” right, too. For instance, swing % stabilizes in 50 plate appearances. That is, if you’ve been watching games for, say a week and a half, and it feels like a player has changed his approach and is swinging at different pitches, you might very well be correct. That kind of change can be seen over a very small sample.
Let’s go back to Average with RISP. If you look at the figures in the linked Fangraphs piece, you’ll notice the absence of batting average. That’s because batting average doesn’t stabilize to the stated threshold in a full season (650) or less of PAs. Maybe that doesn’t make sense to you, but let me put it this way: In 1987, Tony Gwynn hit .370. In 1988, he hit .313. He won the batting title both years, but that isn’t a stable batting average. The National League hit .261 in 1987 and .248 in 1988, so 1988 was clearly the tougher year for hitters, but it doesn’t explain away all the difference.
Still, BA does stabilize over time. I mean, we know Tony Gwynn was a better hitter than Zack Cozart. But would it surprise anyone if one year Votto hit .290 and Cozart hit .295? It would be odd, certainly, but we’ve seen stranger things. So BA takes a while to line out and even one season of it doesn’t tell you everything you need to know about a player’s true talent level.
Then you take average with RISP, which is a subset of batting average, and well, you can’t tell anything from one year. The sample is just way too small. In his career Joey Votto has 914 PAs with RISP. That is maybe getting to the level we need to have some kind of reliability. In those PAs, he has hit .346 compared to .303 in all other situations. So, over his career, he has been better.
The problem is that we have to take such a wide view to get a good sample, other stuff changes. Votto now is not Votto when he was 24 or 25, so those samples aren’t especially valid, but take them out and we don’t have enough information to tell us anything.
And then there are the inherent RISP problems. Fielders are not ideally positioned as they may be holding runners on. Good pitchers allow fewer batters to reach and thus there more opportunities against bad pitchers with RISP. Also, if runners are on, the pitcher is laboring more, can’t throw balls in the dirt, etc. Which is why the league, as a whole, will tend to hit just a little better with RISP than without (.256 w/ RISP this year vs. .249 otherwise). In general, expect a four or five percent offensive improvement relative to average with RISP from any given player (that’s basically like saying an average player will have an OPS+ of 105 when batting with RISP). It’s tiny, but it needs to be accounted for.
So what does all this tell us?
1. Don’t worry about numbers with RISP. The samples are too small to tell us much.
2. Clutch hitting is an ability, but it only becomes apparent over the course of a career.
3. Joey Votto has done poorly in clutch situations this year, but that doesn’t tell us anything about how he’ll do during the rest of the season because the sample is too small.
4. Over his career, Votto has performed well in clutch situations and, at this point, his career has been long enough that this does tell us something about what we can expect from him going forward.
I don’t know if I’ve persuaded any of you, but this article is offered without malice. Questions are welcome and I’ll do my best to answer, but if you come in making assertions with no evidence, I may get a little prickly. Tomorrow, I’ll tackle RBI.