Sunday, April 6, 2008

Is David Ortiz A Clutch Hitter?

Many fans are aware of clutch hitting stats like hitting with runners in scoring position and hitting when it is close and late (CL = Situations when the game is in the 7th inning or later and the batting team is leading by one run, tied, or has the potential tying run on base, at bat or on deck). And many fans probably recall David Ortiz having many late inning clutch hits. But statistical expert Bill James has come up with a new way to judge clutch hitting which can be read in the article Mr. Clutch Big Papi, Chipper, Pujols come through when it counts.

James says ""Clutch" is a complicated concept, containing at least seven elements:

1. The score,
2. The runners on base,
3. The outs,
4. The inning,
5. The opposition,
6. The standings,
7. The calendar."

He then shows what Ortiz did in these situations from 2002-2007 in 394 ABs. He batted .322 with a .413 OBP and a .678 SLG. His overall numbers from 2002-2007 in AVG-OBP-SLG were .298-.394-.597 (thanks to the Lee Sinins Complete Baseball Encyclopedia). So he did somewhat better in the clutch situations than he normally did, but was this difference big enough to be significant?

To answer that, I used a technique of calculating a "Z-score" that Pete Palmer usded in his article Clutch Hitting One More Time. It involves how much better or worse a guy does in the clutch compared to how he hits in other situations. It also takes into account what the normal league difference is. For example, from 1991-2000, the normal dropoff in CL situations was about .012 in AVG (probably because the good closers are brought in and the pitcher has the platoon advantage). If a player gets a Z-score of 2 or more the probability of doing so is under 5% or significant (the Z-score could be negative, meaning that he does worse in the clutch).

Suppose that in the James Clutch situation, AVG also falls .012, like in the CL situations (I have to make an assumption since James did not report what the normal dropoff is in AVG-more on this later where I assume other dropoffs besides .012). But I found that the average difference between all ABs (which includes CL) and just CL ABs from 1991-2000 was .010. A quick check at Retrosheet has it about .012 from 2002-2007. That is pretty close and so the CL vs. NON-CL differences will be similar in both periods.

Ortiz batted .294 in the non-James clutch situations. If the dropoff is .012, using the Palmer technique, I got a Z-score of 1.59, or well below the significant 2.0 level. But as I said above, I don't know the normal dropoff. To get it up to 2.0, the dropoff would have to be .022. That seems like alot since it is nearly twice the CL dropoff. Can the James clutch situation be that much harder than the CL? It is possible since 2000 that CL has gotten tougher since so many more relievers are being used. But the dropoff has to be pretty big to make Ortiz clutch in AVG. Also, if Ortiz had batted .294 in the James clutch situations, he would have had 11 fewer hits than he actually did. That means 11 over 6 years, or just two clutch hits a year (if I put in a dropoff of .012, it would be 16 clutch hits, still under 3 a year).

I did the same thing for HR frequency (HR%), extra-base hit frequency (XB%) and OBP. For HR%, his normal was 7.0% while it was 8.9% in clutch situations, about a 27% increase (8.9/7 = 1.27). I had a normal dropoff in in HR% of .003575, so Ortiz would be expected to have a 6.64% rate (.07 - .003575). He gets a Z-score of 1.48. That 0.003575 is how much lower the average player's HR% is in CL situtaions than in other situations from 1991-2000. That dropoff might be different now. But how much bigger would it have to be to get Ortiz to a Z-score of 2.0 and be significant? The normal dropoff in HR% in the James clutch situation would need to be .0115, like falling from 3.0% to 1.85%. That is pretty severe and is 3 times what I had for the CL dropoff.

Moving to XB%. He had a rate of 14.8% under normal circumstances while he a a 17.5% rate in the James clutch situations, for an 18% higher rate. The Z-score was 1.96 using the normal dropoff of about .013. So this is very close to being significant. His extrabase hit performance may truly be clutch.

Now to OBP. Just using hits and walks (James does not show HBP, SFs, etc. in Ortiz's clutch stats), Ortiz had a normal OBP of .392 and .417 in the clutch. The Z-score was 1.01. Not even close to significance (my data had no dropoff in OBP in CL situations but if IBBs were taken out, it drops off .012). But even so, it would take a normal dropoff of.024 to get Ortiz up to a Z-score of 2.0. And James does not list the IBBs in his clutch data.

So, in general, Ortiz's performance in the clutch is very good but not significant. I also find interesting that about 12.5% of his PAs came in the clutch. In my 1991-2000 data, I had the average guy getting about 15% of his PAs in the clutch. So those two are close and I think using the CL differences is reasonable as a benchmark to calculate Z-scores. Bottom line, how surprising is for a guy who normally bats .294 in 2,756 ABs, to hit .322 in some randomly selected 394 ABs? Not very, even if you assume the average player hits alot lower in those ABs.

Also, some of Ortiz's hits may have ended games. In some of those cases, had he made an out, the next batter (or the batter after that) might have won the game for the Red Sox. Or even if he did not end the game, taking his hit away might still result in a Red Sox win. I once tried to calculate how many games clutch hitters win and very few clutch hitters add alot of wins. See How Many Games Do Clutch Hitters Really Win?.


Anonymous said...

Cyril, you've also got the idea of how transience vs. persistence plays into account in clutch discussions which I feel is arguably the most important. If someone is truly a clutch hitter his numbers will correlate well year to year. According to The Hardball Times, Ortiz was the 5th unclutchiest AL player last year. Would I want him on my team at the plate in a crucial situation? Yes, but I think it is more of his performance in clutch situations being good than him being a devout clutch hitter.

Cyril Morong said...


Thanks for dropping by and reading this entry. What you say suggests a good analysis for Ortiz. I should look at how his clutch performance differed from his non-clutch performance in each of the 6 years that James lists. Then if he does better in the clutch each year than the non-clutch, then he might be a clutch hitter. But, as you say, he was unclutch last year. His clutch numbers that James shows are little lower than his numbers overall.

But we really need the clutch data on a large number of players and calculate Z-scores for everyone. Then see what % of them have significant Z-scores. If it is more than expected, then some of those guys might be clutch. But some guys do better in the clutch due to being lucky and some might do better due to actually being clutch. It is hard to tell who is who (as Willie Rundquuist pointed out).