Stats question

[Nutriaitch]

Ok, I'm not a big time stats guy. I know the common stuff like BA, ERA, stuff like that, but OPS+ and some of the others I've seen here are out of my league. I've heard talks of stats showing how many wins a hitter is worth (or something to that effect). My question is this: Does that take into account the hitters batting before them in the lineup?

Or I guess I'm really asking if there is something out there that adjusts a guys RBI totals by the OPB of the guys in front of him?

Seems to me that a guy like A-Rod or Manny have more RBI opportunities than guys on teams with low OBP like the White Sox or D-Backs last year.

[skyking162]

If you want adjusted RBIs, try this (plus the glossary): http://baseballprospectus.com/statis...php?cid=204019

If you want to start ignoring RBIs altogether in favor of linear weights, that would be even better.

[baseballPAP]

First things first....forget RBI. RBI is a stat that has almost nothing to do with a players hitting ability....it is almost entirely a team offense stat. Meaning, you are correct, Manny and ARod do have more chances.

OPS is simply slugging + on base %. OPS+ is how a player's OPS compares to his league, with 100 being average. So, Adam Dunn's 136 OPS+ last year means he was 36% better than average. In terms of batting average, that would correspond to Dunn hitting .362 in the NL last season. I only use BA because you said you're familiar with it, as it is almost as bad a stat as RBI.

OPS itself isn't perfect, but it is useful enough to be a starting point. Generally, real stats guys use linear weights, which are based upon assigning a value to every event....singles are worth about .4 runs for example. Then, over the course of a season, a player's value can be determined, with 10 runs equalling roughly one win. This is usually compared to the league average player.....and it gets more complicated from there.

[Nutriaitch]

If you want to start ignoring RBIs altogether in favor of linear weights, that would be even better.

linear weights?
Other than seeing that on here a few times, I've never even heard of linear weights. Is there a site somewhere than I can go to educate myself about it? By educate, I mean more than just a site with player stats. Looking for something that'll break it down and explain it to me.

[weskelton]

Nutriaitch,

This is now the best place to start learning about just about anything related to sabermetrics. Probably more than you ever wanted to know.

http://www.tangotiger.net/wiki/

[skyking162]

http://tangotiger.net has some fantastic articles. I'd start with "How Are Runs Really Created" -- it's a 3 part series. The math might be tough to follow the first 10 times through, but you'll get the idea without the details.

[brett]

First things first....forget RBI. RBI is a stat that has almost nothing to do with a players hitting ability....it is almost entirely a team offense stat. Meaning, you are correct, Manny and ARod do have more chances.

I find it amazing that we rate hitters based not on actual run production, but of their rate of bases produced per out, then we turn around and do the exact opposite for pitchers-rating them on runs allowed rather than bases allowed per out.

There are reasons for it of course, that the pitcher is responsible for all events.

[Tango Tiger]

http://tangotiger.net has some fantastic articles. I'd start with "How Are Runs Really Created" -- it's a 3 part series. The math might be tough to follow the first 10 times through, but you'll get the idea without the details.

Thanks for the kind words.

I agree that the 3-part Runs Created article is the best place to start.

[Stat One Author]

If you're looking for a streamlined, equitable, and telling statistic to measure a hitter's value, try P/E Average. For more about P/E Average, go to www.statonebaseball.com

[Tango Tiger]

Craig, if the quote below is accurate, get ready for a pounding from the regulars here. This is the same kind of cr-p I was inventing when I was a teenager. Basically, add up all the numbers you can see, divide by something, and voila, a useless metric.

This is exactly the kind of thing sabermetrics is against.

And, I'm being extremely nice right now.

http://www.kansascity.com/baseball/story/563017.html

The P/E average is simply a measurement of a player’s production numbers plus efficiency numbers divided by plate appearances.

For production numbers, Messmer adds RBIs and runs scored, subtracting home runs to eliminate the overlap. He calls this “net runs.”

For efficiency numbers, Messmer adds total bases, walks, hit-by-pitch and steals, subtracting caught stealing. He calls this “complete bases.”

The calculation: (net runs + net runs + complete bases) / plate appearances.

Messmer says a P/E around .800 or .900 is OK, anything above 1.000 is pretty good, 1.200 is an MVP-type season, and 1.500 is a number reached in a season only 23 times.

A-Rod, if you’re curious, had a 1.415 P/E average last year, far and away best in baseball.

Is it perfect? Of course not. Just on the surface, Messmer’s stat doesn’t take into account park, league, or era factors, which makes comparisons a little dubious. And with the net runs doubled, P/E is still weighted slightly toward guys in big-time lineups.

[Patriot]

Is that 13 or 14 documented separate "inventions" of bases/(PA or outs)? I've lost count...

[misterdirt]

The amazing thing is that Craig actually got a major publisher to publish his "invention".

[skyking162]

This is exactly the kind of thing sabermetrics is against.

And it's exactly what hinders real analysis from catching on.

[weskelton]

Is that 13 or 14 documented separate "inventions" of bases/(PA or outs)? I've lost count...

Ah, but it's the novel introduction of 2*runs produced to the numerator that makes this stat!

I actually picked up Craig's book at one of the major bookstores a few weeks back. Of course, after taking a quick look at the details, I put it back down.

Honestly, I can't fault Craig for putting together what was probably a labor of love, but I don't think we've broken any new ground here. My first recollection of Total Average (one of the early documented inventions) is summer of 1981.

[Patriot]

Good heavens, it never stops.

http://bleacherreport.com/articles/1...ntroducing-UVI

(Total Bases + Walks + HBPs + SB - CS - .25(Bunts + Sac Flies) - GIDP - .1K)/PA

This is hailed by its author as "the stat that will revolutionize baseball". He is either a not-so-clever satirist or a fool; my money would be on the latter.

[SABR Matt]

Oh the humanity.

*headdesk*

I'm really starting to get tired of "new" offensive stats that add nothing to the discussion.

[Tango Tiger]

Wow. I mean, really. Doesn't anyone read any more?

Hey, I have a revolutionary idea: if you take two positive integers, and you put (I will call it "add") them together, you get a new number that is greater than either of the two.

***

No one should be allowed to make up a new stat, until they read The Hidden Game of Baseball. That should wipe out 90% of the "advanced metrics" that are out there.

Patriot, maybe you can write an article for Hardball Times, based on your blog post. THT gets great exposure, so at least people will have a reference point, and you will save all of us alot of pollution.

[Patriot]

That's a good idea. Perhaps you would like to be a co-author (if Studes & Co. would take my submission). I only ask because, to go along with the recounting of all of the different versions of bases/X out there, it would be interesting to have an actual count of bases produced/event, and I don't have the Retrosheet bonafides to compile that myself.

In my post, I did a lot of superfluous regressions and the like that wouldn't fit in a THT piece anyway, so instead I would want that to point out that if you actually account for all of the bases created by a single, rather than just those for the batter himself, you will get a much better weighting of the events. From there, I would close by saying that while that is better, all bases are not equal in terms of run expectancy, and offer LW as the end of the rainbow.

I think that approach would be more effective, since it would give a novice reader who was pointed to the piece a bridge to the true "advanced metrics".

[weskelton]

Patriot, maybe you can write an article for Hardball Times, based on your blog post.

Which blogpost are we referring to? Link please. Thanks.

[Tango Tiger]

Sure, I can generate that for you when I get home (not need for me to be co-author).

Just working the numbers logically, presume you have .30 runners on 1B, .20 on 2b, and .10 on 3B for every PA (a bit higher in the higher run environments, a bit lower otherwise).

A triple and HR therefore will add 1 one base times .10, 2 bases times .20, and 3 bases times .30 for a total of 1.4 bases per hit.

A double would be 1x.1, 2x.2, natch, plus 3x.30x.40 and 2x.30x.60 (a double will score the runner on 1B around 40% of the time). That gives you 1.2 bases per double.

A single:
1x.1
2x.2x.6 + 1x.2x.4
2x.3x.3 + 1x.3x.7
= 0.8 bases per single

A walk will add 1 base per runner on 1B (happens 30% of the time), an extra base when you have runners on 1B and 2B (7.5%) and a further base when bases are loaded (2.5%). In all, a walk = 0.4 bases added

So, in terms of "bases added", both for the batter and the runners themselves, that gives you:
walk: 1.4
1b: 1.8
2b: 3.2
3b: 4.4
hr: 5.4

Treating 4 bases = 1 run, you can divide all the above by 4 to give you:
walk: 0.35
1b: 0.45
2b: 0.80
3b: 1.10
hr: 1.35

That gives you the general basis. You can use empirical data to get you more precise numbers (which I can give you when I get home, or if Matt or the other Retrosheeters have their DB handy, they can post them here).

It's no accident that these numbers are VERY close to Linear Weights.

Since not all bases are created equally, Linear Weights comes in to save the day to gives us a more precise measure. But, we're only talking about 5% correction factor to one side or the other.

[Tango Tiger]

Which blogpost are we referring to? Link please. Thanks.

http://walksaber.blogspot.com/2008/0...orse-pt-4.html

[weskelton]

Thanks Tango. I had looked over there, but was thinking it was more recent.

[Patriot]

Thanks, Tango. The empirical numbers would be great, but I should probably pitch this to THT before anyone wastes their time figuring them.

I knew they would be "close", which would be my intent in writing this--pointing out that the bases gained stats don't account for all bases, and that if you do, you are looking at roughly LW relationships.

[Tango Tiger]

I'm sure studes would go for it. He wrote a similar article on this idea:
http://www.hardballtimes.com/main/ar...luate-hitters/

You should probably consider being a regular contributor there. Studes is amassing quite the list, and it always culminates in a great Annual.

[Patriot]

Studes gave the green light. Thanks for your suggestion/help.