clutch study

[Mean Dean]

I took data from four complete 2004 draft league seasons, in order to try to figure out which spots in the lineup would be most affected by clutch hitting.

What I did was, for each player, I figured out what % of his total plate appearances were clutch opportunities. e.g., if he hit 500 times and had 50 clutch opportunities, that was 10%. I then found out nine sets of correlations: the correlation between that percentage, and the percentage of times he hit 1st in the lineup; the correlation between that percentage, and the percentage of times he hit 2nd; etc.

The correlations were, from highest to lowest:

5th (correlation = .13)
9th (.12)
4th (.09)
8th (.06)
6th (.04)
7th (-.01)
3rd (-.09)
2nd (-.13)
1st (-.15)

Any thoughts, or comments on how this study could be improved?

[cummings2]

Dean, I am really interested in your study, but I am bit slow and thick.

Let me see if I understand this, you took four complete leagues (48 teams) and you took from all those players the clutch data you described, roughly about 432-450 cases, obviously, some cases it's the same player in a diffferent league, given that its only four leagues, the info for each player can not be repeted more than 4 times, that is there can be no more than 4 sets of data for any given player. So you collected the info of no less than than 108 players.

Then you determined for each of those no less than 108 players the % of PAs where the clutch came into play. This is for lack of a better name the Overall Clutch Percentage (OCP) for the given player.

Then for each player you correlated the ammount of times they hit, 1st, 2nd, 3rd, etc... in the lineup.

So, to simplify it and using a singular example, you collected Jim Thome's info and came up with his Overall Clutch Performance.

Then you calculated the % of times he batted 1st, 2nd, 3rd, 4th, etc... for a lack of a better term I'll call this the Individual Lineup Percentage (ILP)

THEN you correlated the OCP and ILP the result of that correlation is what you listed?

Did I get this right?

So, lets say that in this ficticious case Thome's OCP was 15%, that is that 15 percent of his PA were clutch opportunities:

If Thome were to bat 5th the % of PAs with clutch in play would be 28?
and If Thome bats 9th the % of PAs with clutch in play would be 0?

So lets say a Hitter will have 600 PAs, in this case I can expect 128 PAs to be clutch opportunities, in other words 64 rolls on the hitters card to be clutch situations?

If I did get this right and I just repeated what you initially wrote I apologize but I want to make sure I understand before I comment.

Very interesting Dean.

[worrierking]

Dean why a correlation and not just raw percentages? Nice work, BTW. Thanks for sharing.

[cummings2]

The result of an IBB would alter results, perhaps not of much relevance but since you asked for any thoughts towards improvement:

Along with keeping tabs of PAs in which clutch was a factor, if you keep track of how many of those came into play because the previous PA result was IBB and how many of the Overall Clutch Times resulted in IBB then you'd probably have a more accurate list of results. One not only with the context of batter and spot in the lineups but also factoring IBBs (Spot in the lineups not in the context of orders but of hitters) since it's been argued that clutch is one of the elements used by HAL to issue IBBs maybe it is worth at least looking into.

Have to run, will comment on this later.

[Mean Dean]

My goal was not to determine anything about what actually happened in the clutch plate appearances, but simply to look at how many clutch PAs each lineup spot gets. About 13% of the total PAs were clutch opportunities. Now, obviously, that doesn't mean 13% of PAs are actually affected by the clutch rating, because not every roll in the clutch is affected by the clutch rating. But, that's not really the point; what I think a person making a lineup wants to know is, how often will each lineup spot come to bat in a clutch situation.

Let's take Thome as an example, since you used him. In one of my Thome seasons, he came up 602 times, and had 89 clutch opportunities. So, about 15% of his PAs were clutch PAs, a slightly above-average percentage. He hit 2nd 7% of the time, 3rd 47% of the time, 4th 37% of the time, 7th 8% of the time, had a couple of at-bats 6th and 8th, and never hit 1st or 5th. The goal of my study was to try to find a connection between which players had more/fewer clutch opportunities than normal, and where those players hit in the lineup. Thome had a slightly above-average % of clutch PAs, and hit mostly 3rd and 4th; that's a very small data point towards concluding that 3rd and 4th get a lot of clutch opps. (If he had a higher % of clutch PAs, or if he hit more exclusively in one spot, it would be a stronger data point.) He had that above-average % of clutch PAs without hitting 5th once, so this season actually very slightly decreases the correlation between clutch PA % and hitting 5th. [quote:aefacace27]Dean why a correlation and not just raw percentages? [/quote:aefacace27]Can you be more specific?

[cummings2]

You're right, my boo-boo. It's clear from your post that the data you are after is % of times that clutch comes into play. Not in the result of clutch situtations.

I guess where my rationale was going is in finding how many times (or % of times rather) the clutch came up due to IBB to the previous hitter.

To continue using Thome as an example. Using the greatest %: He batted 3rd 47% of the times, let say that the batter that was batting 2nd was Ichiro or some monster Avg. hitter like that. It could be argued that some % of his clutch opportunities came up because Ichiro was walked intentionally, where as if the 2nd spot were a weaker hitter (like Adam Kennedy) Thome's overall number of clutch PAs would go down. In the case of his batting 8th and with the low numbers of IBBs likely issued the numbers would probably be quite low, so low they couuld be discarded, but with his 47% of PAs batting third even a low number of IBBs issued to the 2nd hitter would likely affect the results. Just a thought though.

The nature of why HAL issues IBBs is up for debate in another thread but it can be argued that a greater number of IBBs are issued with 2 outs, therefore rendering the following AB as a clutch opportunity.

My mistake was in suggesting factoring IBBs before the PA and for the PA. You are right, what I should've written was factoring how many of those clutch PAs were as a result of an IBB to the previous hitter only.

Hope this makes sense.

[pedakrla]

Dean, this is great stuff. Thanks for sharing. Would it be possible to express the data in terms of

1st position in order--x% of PAs were clutch
2nd position--x% were clutch
and so on......

Thanks again.

[Mean Dean]

[b:c7079dd16b]Cummings[/b:c7079dd16b]: I see. In order to break this down to what people did in front of/behind specific hitters, I'd need play-by-play data, which I don't have. This is definitely going to be a generalization type thing. I'd imagine that the clutch opportunity distribution is different on the Pirates or Royals than it would be on the Red Sox or Yankees, because you have different hitters, batting in different spots, who make outs a different % of the time. If anything, one thing that is surprising me (besides that the 9th spot is the second most closely correlated) is that the differences between the various spots aren't really all that big.[quote:c7079dd16b] Would it be possible to express the data in terms of

1st position in order--x% of PAs were clutch
2nd position--x% were clutch
and so on...... [/quote:c7079dd16b]Nope; if I did, I wouldn't have done this ;) That's pretty much exactly what I'm trying to figure out, by figuring the correlation of clutch PA % with % of time batting in each lineup spot.

[worrierking]

Dean, I was asking the same thing as Dave. Why is this a correlation to the batting order position rather than a list of what % of PAs occur in a clutch situation per batting order position?

I'm sure your explanation to him makes sense but I still don't understand it. Stupid Brain Cells! :P

[Mean Dean]

Because I didn't have that data :) The game can tell you a player's total clutch AB/BB, and it can tell you his AB/BB in each lineup spot. It doesn't track the combination of those two, clutch AB/BB in each lineup spot -- that's what I'm trying to figure out. The only direct way to figure it, as far as I can tell, would be to go through the play-by-play and actually count them. I'm obviously not gonna do that for almost 10,000 games played, so that's why I tried to get at the issue indirectly by using correlation.

I do get the feeling there is a more direct way to get at this, but I can't think of it.