After Sabre-school Special, Oct 31 (October 31, 2003)
I'm really really behind, so here goes...
Clutch hitting, again
People get nervous, or get cool under pressure. A person's response is always dependent on his environment. The result, the outcome of his action cannot be something random, centered around his normal abilities. He has different abilities for different environments.
Rising (or failing) to the occasion exists, because the things necessary to make that happen exist.
DER, again
DER = fielding+pitching+park+luck
Assume luck = 0.
Park is already provided by me at my site as an adjustment, so use that.
So, let's say you have a DER of .740, and the league is .700, and your park is -.020. You take half your park adjustment (half your games at home), and you get -.010.
DER over average = .740 - .700 = .040
= fielding+pitching+park
= fielding+pitching-.010
So, fielding+pitching=.050
What I would do, if you have no further information, is split the difference in 2, and make fielding=pitching=.025 (it's the safe way).
DER, BP adjustments
They're trying to find the pitcher's skill at BIP, but are not regressing it (based on your description).
The other problem is that the following equation is not being respected:
DER = Pitcher + fielders + park + luck
As we have determined from the Allen/Hsu study, luck is the biggest component of DER.
Defense, pitching/fielding split
50% (though you can argue 60%) of BIP and 100% of non-BIP goes to the pitcher. Almost 75% of PA are BIP.
A safe BIP leads to about +.55 runs and an out BIP leads to about -.30 runs (more or less).
A safe non-BIP leads to about +.55 runs (numbers dependent on the freq of BB and HR for a particular pitcher) and an out BIP leads to about -.30 runs.
So, about 70% of the difference between team runs allowed and league average runs allowed can be attributed to the pitcher. (.6 x .75 + 1.0 x .25 = .70)
Defense, pitching/fielding split, once again
The problem is with your initial train of thought. This idea of "breaking things up" is not a good one, and one which is perpetuated with Win Shares.
Things work in a marginal additive manner. So, the way to approach things is to first adjust out the park factor. And then, once you've adjusted that out, split the difference with what you don't know.
If LA has a DER that is .050 worse compared to the league, but that their park helps them by +.010, then we know that the pitching+fielding is .060 worse than the league. Ideally, you would know what the fielding portion is, but since you don't, you split the difference, and make it +.030 for each.
Don't think of things in terms of "% of responsibility" IF you already know what that responsibility is (e.g., park).
If you DIDN'T know the park factor, then you would take that .050 and say .020 pitching, .020 fielding, and .010 park. That is, in this case, you are saying that the park hurts the DER by .010.
Park factors, custom
If you take say a park like Fenway, and you have alot of righty-FB pitchers there, this will skew the results of the true impact of Fenway. The most accurate thing to say about this example is: "Given that 60% of the balls in play were thrown by righty FB pitchers, Fenway impact ALL pitcher by +.020". Ideally, what you would do is break down your park factors by the GB/FB and LH/RH of your pitchers, and have FOUR separate park factors. You can break it down even more by fastball/curve, or great pitcher / bad pitcher, etc.
That is, suppose that LA has a ton of pitchers who
are:
- great
- FB
- RH
- fastballers
then any park factor that you use should be tailored to that profile. If you just use a general park factor for DER, then yes, it's very possible that the adjusted DER will not accurately portray what really happened.
(This is similar to not applying the Pac Bell PF to Bonds, because that PF was established using all players, and what the heck does Benito Santiago, or anyone, have in common with Barry Bonds?)
Runs Produced
Runs Scored = .26 1b + .43 2b + .61 3b + 1.00 HR + .26 BB plus other stuff
RBI = .21 1b + .42 2b + .63 3b + 1.63 3b + other stuff
With R+RBI, the HR is worth 2.63. With R+RBI-HR, it's 1.63. Which looks more right to you?
If you go on my site
http://www.tangotiger.net
I have a long dialogue, plus a good evidence, on this subject.
*
I don't think runs produced will answer any meaningful question. It's a useful quick and dirty stat, and you'll get into trouble if you go beyond that. The productivity of Bonds v Pujols can best be answered by creating a model that replicates their conditions as best as possible... and then removing them for some neutral player (say Manny Ramirez). How they perform relative to Manny will show you something that RP just won't. Of course, you also have the # of game to contend with.
*
I use something like RP - AB/10 = Runs Created.
If you want to create a rate stat, you would do RC /
(AB-H).
SB break-even point
I believe it's 72% for a 5 RPG, but I'd have todouble-check. Drop 4% for every 1 RPG change, so thatat a 3RPG, it would be about 64%. Again, just a ruleof thumb.
According to the Run Expectancy matrix that I published here for1999-2002
http://www.tangotiger.net/RE9902.html
you can derive the break-even points on SB as being73.8%. The run environment for that time period was5.00 runs per game.
I also published somewhere, but I can't find where onthe web, the RE matrix for 1974-1990. The SBbreakeven point was 70.6%, for a run environment of4.3 RPG.
Important note: I am only considering the possibilityof safe at 2B, or out. With a SB, the catcher canthrow it away, etc, etc. A runner can also be safe ona CS, too. When you consider all that, the break-evenpoint drops about 2-3% points.
To get a better understanding of why the breakevenpoint changes as it does, consider that the marginalrun impact of a SB adds about .16 to .17 runs,regardless of run environment (add about .02 to .03for the "extra stuff" that might happen). However,the CS's value changes drastically as the runenvironment changes. Why? Because outs are precious,especially in the higher run environments. Youwouldn't steal in a softball league, but you shouldsteal alot more often with Pedro on the mound.
Shaping fans' perceptions with stats
All I do is present a perspective that is grounded in logic, and is useful with as few limitations as I think is worth it.
I don't think I'm in a position to say what the average fan wants or knows what he wants or is willing to accept. I'm not really interested in shaping other people's minds, other than those who wish to put the extra effort to look at different perspectives.
What I write is appreciated by my readers, who, while they may be very small in number, are very big in digging their feet in. I'm happy to leave it as such.
Isn't LWTS too hard to handle?
No, I don't think that's too much to handle. A fan doesn't figure an ERA himself, though he understands what a 3.20 ERA is. Similarly, if you tell the average fan that a player is +60 runs per 162 GP above average, he can appreciate that as well.
I think people can accept Leveraged Index (LI), without knowing how I figured it out.
OPS does a pretty good job in any case to convey the value of players to the "common" fan. Even "Total Average" is a pretty simple concept, and should be easy for the common fan to embrace, if he so chooses.
A hitter's goal should be to contribute to the scoring of his team... and not to collect H, 2B, R, RBI, HR in and of themselves.
His goal is to contribute to TRYING to help his team to WIN. All those little pieces, H, HR, walks, errors all contribute in some way or other, and those contributions are dependent on the context of when it happened (men on base? 2 outs? late and close?).
OBA and SLG are statistically significant, as is ERA. Therefore, they have great meaning. Maybe not to the level you want it, but it is at a pretty high level.
Walking Bonds
Check out www.baseballprimer.com, click on Authors, click on Tangotiger, and I have 2 articles entitled, "Dear Mike, you should walk Barry Bonds when..."
There are plenty of situations to walk Bonds.
At Primate Studies, several weeks ago, I posted articles by Mark Pankin and John Jarvis who also look at the same issue. Jarvis I think won "presentation of the year" for SABR 33.
--posted by TangoTiger at 01:00 PM EDT