Do Power Hitters Choke in the Clutch?
by Cyril Morong
A study called “Clutch Hitting: Fact or Fiction?” By Andrew Dolphin suggests that they might. It is at
Dolphin used a Monte Carlo simulation. Since I have not used that technique, I can’t comment on it.
I have a different data set, all the hitters with at least 6000 plate apearances (PAs) between 1987-2001. The clutch situation I looked at was “Close and Late (CL),” or 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.
The first thing I did was to calculte a hitter’s SLG DIFF (slugging percentage differential) by subtracting their NON CL SLG from their CL SLG. The correlation between a player’s overall AVG and his SLG DIFF was .196. So the r-squared was .038, meaning that 3.8% of the variation in SLG DIFF across players is explained by their overall AVG.
I then ran a regression in which SLG DIFF was the dependent variable and AVG and SO% were the independent variables. SO% is SO/ PA with intentional walks taken out.
The equation was
SLG DIFF = -.10 + .307*AVG + .107*SO%
The r-squared was .065. Neither AVG nor SO% were significant.
So it looks like high average hitters are more likely to maintain or increase their SLG in the CL, but the effect is not significant. A .050 increase in AVG means a .015 increase in SLG DIFF.
I then ran a regression in which AVG DIFF (CL AVG – NONCL AVG) was the dependent variable and overall SLG and SO% were the independent variables. The equation was
AVG DIFF = .017 - .069*SLG - .003*SO%
The r-squared was .064. SLG was significant, with the T-value being –2.10. The effect may not be large. A .100 increase in SLG means a .007 decrease in SLG differential. So power hitters are more likely to see their AVG decline when it is CL, but the effect is small.
Actually, the higher the SLG , the higher the CL AVG. This can be seen in the following regression equation
CL AVG = .248 + .064*SLG - .049*SO%
Although the effect is small and the r-squared is just .027.
But which player’s are most likely to see their SLG fall in the CL? The ones with the highest NONCL SLGs. We can see this in the following regression equaiton
SLG DIFF = .061 - .251*NONCLSLG + .107*SO%
The r-squared was .217. The T-value on NONSLG was –4.31, so it was significant. The effect may be large. A .100 increase in NONCL SLG means a -.025 fall in SLG DIFF. The highest drop off in SLG was about .119. So .025 could be a fairly large part of that. About half the players had a SLG DIFF between -.2 and -.65