clutch study

[cummings2]

O.K. Dean, I've been thinking of the IBB element a little bit.

First of all I think the way you are going about it is very good and should yield very intriguing numbers. Nice going there.

Now with the IBB thingie what I was thinking is that it could be used as an element to determine the margin of error.

For example, in the record books the top 5 IBBs are

Most Intentional Walks

Ichiro Suzuki - 37
Ichiro Suzuki - 33
Ichiro Suzuki - 29
Barry Bonds - 29
David Ortiz - 28

Most of my leagues (I am going through the few I've played and gathering some data) the top IBBs getters are around the high teens to low 20's.

So what I was thinking is this: If one takes a range of estimated IBBs issued to previous batter it could yield a margin of error so the result of your correlations would read: x.xx (+/- .xx). The value of this margin of error would be a bit generalized and it's only apllicable use would be while forming specific lineups where you can see for yourself the likelyhood of the IBB to come into play and then discard the +/- or consider it.

So, if you project that the range to be from 0 IBBs to 20 IBBs to the previous hitter over the course of a season then...

O.K. I'm getting lost here... the aim of this would be to have a margin for the ammount of PAs with Clutch in play. So, the range of error (from IBBs) would be applied to the Overall Clutch Percentage. So the Overall Clutch Percentage for all hitters would remain the same but now it would have a range of error...

then, if one projects no IBBs to the previous hitter the range is subtracted to the OCP and then correlated to the ILP (Individual Lineup Percentage) -To use some of the schnazzy names I came up with earlier on-

and the same for considering a high ammount of IBBs but in this case the margin of error would be added to the OCP.

As I said, I am getting lost here. Maybe -hopefully- what I am trying to say makes more sense to you than to me.

I need a ciggy and coffee...and to put my glasses back on.

...And if none of this makes sense: then forget I said anything at all :wink:

[MARCPELLETIER]

Dean,

I still don't understand your correlation. To do correlations, you need two sets of data, or two matrices. What data got into the first, and what data got into the second?

[Mean Dean]

[b:cfda9a4e7f]Cummings[/b:cfda9a4e7f]: Okay, it would theoretically be possible for me to account for that; I do know how often a player hit in a particular lineup spot, so if I know Ortiz hit 3rd 80% of the time and Manny hit 4th 80% of the time, I guess I can figure that 64% of Ortiz's IBBs were in front of Manny. I kinda doubt I have the spreadsheet ability to actually pull it off, but I'll see what I can do.

[b:cfda9a4e7f]lucky[/b:cfda9a4e7f]: The first group is the % of the player's PAs that were clutch opportunities. The second group is the % of PAs the player got while hitting 1st in the lineup. (Repeat eight more times for hitting 2nd, 3rd, etc.)

[tersignf]

This is very interesting.
Thinking in terms of self-consistent systems, such as in electromagnetics, another consideration is the control group used, with respect to their OB avg mostly I would think.

Basically, my point is, rather than strictly correlating clutch opportunity to lineup position, another degree of freedom that could affect the results' accuracy if not accounted for would be the likelihood of players getting on base ahead of the position in question.

Although this is indirectly addressed by correlating clutch PAs within the same control group based on lineup position, it seems there would have to be an assumption made regarding a target OBA in each of those positions--to limit one more degree of freedom. I suppose one could also say XB opportunity and/or SBs could also have such effects, but I think they would be small enough such that they wouldn't have to be declared as an assumption to validate a result set.

[Mean Dean]

Hmm. That is a good thought, because the game does break down stats by lineup position, and it would be relatively easy for me to figure out the stats of the "average" #1/2/3/4/etc. hitter. Figuring out the hitter who actually hit in front of a given player is a pain in the butt... but averages, I can do. I will see what I can do with that. Thanks.

[bleacher_creature]

So Dean...

What are the current conclusions? Others have stated in the past that the #5 spot gets the most clutch opps. It sounds like your study says 3 & 4 do. The vets here seem to quickly jump to the thought that, "well...it depends on the lineup..."

BTW, I like the idea of knowing which spots in the order have the highest % of opps. Maybe because it is simple to understand? :?

[Mean Dean]

No, my study did have 5th as the spot that was best correlated with clutch opportunities. I feel relatively confident that clutch is most important at 5th, and that it is least important in the top third of the lineup. Wouldn't be too comfortable getting much more specific than that, but, I bet that statement at least is true.

[tersignf]

Absolutely Dean, and your qualitative simplification makes sense for good reason.

The 5th slot is maximized in terms of OBA in front of it combined with slugging in front of it. Just like any optimization problem. Of course getting past qualitative is difficult in terms of closed end solution, but I still think an approach is to make an assumption regarding OBA and SLG in front of the position in question, as it related to opportunities for hitting with 2 outs and at least a runner as far as 2nd.

It just seems to make qualitative sense to be 5th--consider:

If I had to put a weight toward OBA versus SLG as a determining factor in front of the position in question, I'd say it's probably something like 75/25 OBA/SLG, with the weighting regarding position of the OBA/SLG becoming quite difficult to reason through qualitatively. Would seem SLG influence on clutch oppty is maximized at the position right in front of the hitter for obvious reason (actually SLG here should be defined as the chance for an XBH that is NOT a HR), and OBA more important 3-4 slots in front (most common scenario to get 2 on then 2 more are out).

Either way, 5th makes sense because your top slugger is typically 4th while your top OBA guys are 1-2.

[tersignf]

Further, as an extension, think of high OBA/high HR guys who usually hit near 4th...I bet you would see a bimodal distribution of opportunities versus a smooth variation through the lineup for clutch opps, i.e., 2 spikes with the 2nd spike in the bottom third of the order.
I think that's what some of the numbers have shown empirically.

If I had to guess had some kind of relation (whether we thin of it parametrically or as a regression etc), I'd think lineup position could be the independent variable, and the coefficents would be some combo of a weighted OBA and XBH (both minus HRs) that accounts for the 4 positions in front of the lineup placement in question.

If we could keep it in simple cartesians as a curve in the x-y plane/2-d space, it seems clear your x-axis is position and your y is clutch chances.
I'd think about making a guess at what the curve would look like (thinking a double hump) and then see what gets us there.

I don't know--just musings and habit...when a problem seems intractable, I start with what I know and what's simple, then refine it. We may end up in spherical coordinates with a Fourier series when all is said and done!

Frank