THE BOOK--Playing The Percentages In Baseball

I’ve been reading these all night (at the expense of studying for my math test!), and I have some general notes to make regarding errors, omissions, good points, etc.  Steve, if you are reading this, I hope you don’t take it like I’m being overally critical, just trying to help make your library the best out there . 

1) There doesn’t appear to be any explanation of Linear Weights in the “Offense” section.  Given that wOBA, wRC+ and wRAA are all just different forms of Linear Weights, it should definitely be discussed. 

2) In the ISO section, I would also link to Patriot’s article about Power Factor, just cause I think it’s an interesting debate. 

3) On BABIP, it’s incorrect to say that it’s not a measure of talent, but luck.  BABIP has some skill and some luck in it, like everything else. It just has more luck involved in it than other stats, but when a player put’s up a .400 BABIP, he likely has an above average BABIP skill.  (Side note: Rally, do you have all of your projections in a CSV file somewhere?)

4) For plate discipline, you should mention that BIS has changed the ways in which they have calculated their stats over the years, you have to be careful when evaluating multiple seasons. 

5) For UZR, it’s using almost the exact same methodology, with a few differences.  I’m just curious, why do you think +/- is more intuitive? 

6) For FIP, you should mention that it literally measures, “how many earned runs per 9 a pitcher would have given up, if performance on balls in play and timing were assumed to be league average”.  It only infers at true DIPS (although quite well!)

7) For tRA, you should mention that batted ball classifications aren’t going to be completely accurate due to bias or just misjudgment.

8) Like BABIP (and HR/FB), LOB% is not a luck stat, just more luck filled than others.  It may seem like a semantic issue, but I think it’s an important distinction to make. 

9) Pitch type Linear Weights are just basically the same thing as wOBA (which is the rate stat form of Linear Weights), except that they are adjusted by count.  I don’t know how you are getting that they are like WPA. 

10) There doesn’t seem to be any mention of the positional adjustment for WAR, although I can understand why you just linked to Dave’s articles and didn’t want to have to explain all of it in such a short place.

11) I liked your explanation of WPA a lot.

12) How does David calculate WE?  Using empirical or Markov? 

13) Regarding Pitch f/x, this isn’t really right “Pitch f/x is a way of visualizing a pitcher’s “stuff”.”.  It also tracks location, and, indirectly, release point. 

Also, Pitch f/x is not really used for making graphs, it’s more used for examining trends in baseball as a function of the attributes that it measures (just like most data).  It can also, as you mention, serve for scouting reports.  You seem to be a bit underrepresented of some of the better applications of Pitch f/x, so here are some links that I think are crucial:

http://www.hardballtimes.com/main/article/how-fast-should-a-fastball-be/
http://www.hardballtimes.com/main/article/searching-for-the-games-best-pitch/
http://www.hardballtimes.com/main/article/fastball-slider-changeup-curveball-an-analysis/
http://www.hardballtimes.com/main/article/pitch-identification-tutorial/
http://www.hardballtimes.com/main/article/squaring-it-up/
http://www.hardballtimes.com/main/article/the-eye-of-the-umpire/
http://www.hardballtimes.com/main/article/the-injury-zone/http://www.hardballtimes.com/main/article/pitch-sequencing/
http://www.hardballtimes.com/main/article/anatomy-of-a-player-league-average-pitcher-part-ii/

Sorry for the obnoxiously long post.

I agree with the comments on pitch-f/x in post #1. 

(Nick, your check is in the mail.)

"12) How does David calculate WE?  Using empirical or Markov?  “

Fangraphs and B-R.com both use my WE and LI numbers.  My process for generating those numbers were to use empirical state-transition numbers at the base/out level, then use recursive probability to flesh that out to inning/score/base/out.  That was for the 5.0 runs per game environment.

Then, with some fancy-schmancy work (one day, when I retire, I’ll donate all my work… this particular one I’m fairly proud of), create a set from 2.0 to 7.0 runs per game, in steps of 0.5.

What you then do is establish the run environment you are in, say 4.7, and then take 60% of the 4.5 and 40% of the 5.0.  There, done.

The ONLY difference between what David and Sean do is figure out what the run environment is for a particular park.

Excellent, thanks Nick! I was really waiting for people to start tearing the Library to shreds yesterday, considering I wrote it all in about 3 days, don’t consider myself an expert on all the stats out there, over-simplified lots of things in the name of brevity, and probably missed a bunch of caveats to throw in there. So then when I didn’t get much constructive feedback minus one or two little things here and there, it felt very odd. I really appreciate the time you took with your response and I’m going to look into making the necessary changes in the areas you pointed out.  Like you said, it’s about making it the best it can be.

You asked a couple questions above, so let me see if I can answer some of them. Let me go through it point by point:

1) Great idea. Will do.

2) Also excellent. Easy to add in and will do.

3) I tried to express that hitters have influence over their BABIP numbers to some extent, but maybe I should change how hard I stress the “luck” aspect. I did pound it in very hard and you’re right, the real answer is a lot more subtle than that. I can try and tweak that a bit so the message isn’t getting lost, but I think it’s also important for newbies to realize how much luck can influence statistics like BA and ERA.

4) Will do. Check.

5) I find the Dewan scale as easier to understand intuitively mostly because I’ve never seen UZR explained in a succinct, direct manner similar to the way Joe Pos explains +/-. All of the links I could find on its methodology seemed more involved than +/-, including the simplest one I could find at Big League Stew. Maybe that’s just me, though. If you wanted to put together an intro for UZR, I’d be happy to put it up on the site.

6) Good caveat. That’s an easy change to make.

7) Good call. Forgot about that.

8) Similar to BABIP, I tried to make the distinction and mention how certain pitchers can influence their LOB% to an extent, although I probably hammered home the luck point too hard again.  I guess what I’m wrestling with is how much accuracy you lose in the name of simplicity.  For newbies getting started on saber stats, giving them the more technical version seems almost a bit much. The important thing for them is to grasp that luck plays a part in baseball, which most people have probably never considered much before.  Yes, babip and lob% are part skill and part luck, but how do you explain the fine line there without confusing?

That said, I think I can make those pages a bit less luck heavy. Check both pages Wednesday morning and see what you think.

9) At least to me it helps to think of Pitch Type Linear Weights in a similar way as WPA, but maybe that’s entirely off. The way I see it, you’re measuring how much run expectancy (as determined by linear weights) changes on a pitch-by-pitch basis.

Woh! So I just re-read Dave Cameron’s explanation of PTLW and just got a whole new understanding of them. So PTLW isn’t like WPA in that it doesn’t account for the situation. Instead, it takes the average run value for each count and measures the change that way. Now that I actually understand PTLWs, I’ll go back and change that.

10) Yeah, my WAR article was already really long compared with the others and there are a million and on links to go for other info, so I figured I wouldn’t get into the details.  I think I briefly mention it when discussing how to calculate offense WAR. i’ll check on that.

11) Well, thanks

12) ...if I said something wrong, let me know or type up something you’d like to see in there. I think I implied that it was empirical, but I don’t really know all the details.

13) My section on Pitch f/x was one of the last ones I did and I figured I would need to beef it up later because it seemed very bare-bones.  Love the links and no worries, I’ll make it better.

Thanks again for the feedback!

Thanks for the response Steve.  12) there was nothing wrong with anything you said, I was more asking Tango for clarification.

John - cash would be better.

Steve - how would you feel about giving a few people access to add on to the Library?  That’s what Tango originally did with the Wiki, but for some reason, it never really got popular.

Yeah, I started thinking about that last week a bit and I think that’s going to be the next area that the Library is going to expand.  I’ve already included a bunch of work from other people on the site, but it’d be nice to have a cast of characters able to add and re-word things here and there. I’ll keep it in mind and probably end up adding some other writers within the next couple weeks.