Tangotiger Blog

Wednesday, October 28, 2015

Why Baseball Prospectus, why?

?I've had a few dozen email exchanges with Pizza Cutter over the years.  He's produced tremendous research in the past that can be referenced for years to come.  So, I have high expectations from him.  His latest article and my comments that I posted:

The premise to the article is great.

The process to get the answer is great.

The entire payoff was this: "The results? Nothing. Extra days of rest earlier in the season didn’t help a pitcher in September (or hurt him). Everyone pitched about how their seasonal stats would predict he would, at least in the aggregate."

Not a single piece of data to hang our hat on. If Verducci did this, we'd blast him for it. Baseball Prospectus? Hugely disappointed. At a pure minimum, can't you just provide a link to the results of the data?

I was just going to leave it there, but I find his response disingenuous:

Tom, it would be a link to a bunch of regression output that showed non-significant finding after non-significant finding. There's only so many times someone can read p = .872 until they glaze over. There just wasn't anything there to show.

So, instead of just talking about the results of his work, we're left to talking about HOW to talk.  All I'm asking is for him to provide this:

Russell, you can easily show the "before" data, in terms of IP, and K and BB and ERA, and you can show the "after" data in the same way, and split it based on 4 days rest and 5+ days rest.

I should point out that my comment has already received a -2 rating, which just goes to show that people prefer nice responses that contain nothing, than harsh responses that have legitimate complaints.

(I had a long extra stuff written, but I decided against publishing it.)

Sunday, October 25, 2015

John Lackey: Matt Harvey before Matt Harvey

?Jeff Zimmerman pointed out that John Lackey came back from Tommy John surgery in the 2013 season.

Lackey had 29 starts in his comeback season.  So did Harvey.  Lackey threw 2873 pitches, which was just 75 more than Harvey.  Lackey started one postseason game in the LDS and one in the LCS, throwing a total of 192 pitches.  Harvey has also thrown one in each, totalling 194 pitches.

Lackey ended up with two more starts in the World Series, including the Game Six finale, with another 200 pitches (plus another 17 pitches in relief in Game 4).

For all we know, Matt Harvey's been on the John Lackey rehab plan all along.

Wednesday, October 21, 2015

Letting your starter bat and keep pitching, or bring in the replacements?

?MGL takes a look to see how much of an impact each decision has, and how much you have to assume in order to make the decision break-even.

Let's see if I can try to figure it out in one paragraph: a pitcher-as-hitter is some .13 runs per PA worse than league average, or .013 wins per PA, when LI = 1.  If LI = 2, then it's .026 wins.  If LI = 3, then it's .039 wins.  So, that's your cost of letting him hit.  What's the gain for letting him pitch?  If you assume that an almost-great pitcher gives up runs at 80% of league average, and his replacement is league-average, and an average pitcher gives up about 0.45 runs per inning, then you gain 20% of that, or .09 runs per inning, or .009 wins per inning.  If he pitches 2 innings, that's .018 wins at LI = 1.  At LI = 2, it's .036 wins.

All in all, prima facie, you have to work it out, but it may be break-even.  But MGL went through the machinations more closely, and he shows that the decision is pretty straight forward.  All to say: everyone should roll up their sleeves and work it out.  And don't forget the huge PH penalty too!

Wednesday, October 14, 2015

Talking pitching with Collin McHugh

If you want a pitcher to answer "how do you feel?", then listen to a press conference with professional writers.  But if you want a pitcher to talk about the evolution of his pitches, then here's another terrific piece by blogger Eno.

Tuesday, October 13, 2015

Do long layoffs favor the pitcher?

?This is an opening study, but it's clear that you need sample size.

Monday, October 12, 2015

Should we evaluate batters by Linear Weights or RE24?  And is the same answer for pitchers?

?Jonathan asks the question:

So, ask yourself this: if wOBA / TAv are the standard means of evaluating batters, shouldn’t the fundamental measure of pitcher value be the extent to which they limit batter wOBA / TAv? Of course it should.

Not so fast!  With linear weights (or wRAA as you will find it on Fangraphs, which is Runs Above Average based on wOBA), we treat all PA the same, regardless of the base-out situation.  A HR is +1.4 runs whether the bases are empty or the bases are loaded. 

With RE24, we apply a different run value for a HR based on the base-out situation.  A HR with the bases empty is +1.0 runs, while a HR with runners on base will be higher, and much higher with the bases loaded. 

Can you make a case for one over the other?  Sure, it depends on what you are after.

Now, to be logically consistent, must you do the same for pitchers?  No, it's not a necessity.  It depends on the reason you do it for batters.  If the reason you prefer Linear Weights to RE24 for batters is that the batter is not "responsible" for the base-out situation he sees, and so, it is "unfair" in terms of the number of opportunities faced, then that's a reasonable choice.  This applies especially for leadoff hitters.  It also presumes that a hitter won't change his approach based on the base-out situation, which of course is ludicrous.

And it gets to the point I keep making, that do we want to assign the impact of an event to a hitter simply because he happens to be involved, even if he may not "own" everything about the change in that event?

But for pitchers, it's different.  If Verlander walks the bases loaded and then allows a HR, before striking out the side, that's 4 runs allowed (or +3.5 runs above average).  If we followed Linear Weights, we'd give him +1 run for the 3 walks, +1.4 runs for the HR and -0.8 runs for the three K, for a total of +1.6 runs above average.  Where did the other 1.9 runs go?  Well, they went in how Verlander sequenced the events.  He owns that and no one else.

If you include balls in play, the fielders also take their share of the credit for that, but overall, the pitcher is going to own more than 50% of the sequencing, maybe closer to 75%.

Monday, October 05, 2015

Game Scores for 2015

?I posted the enhanced Game Score metric here:

• 40
• +2 outs
• +1 K
• -2 walks
• -2 hits
• -3 runs
• -6 HR

It's pretty straightforward, owing a great deal to Bill James, but shaped by Pete Palmer and Voros McCracken.  You can read the link for more background.  The three main areas of improvement is how it starts off each start at 40, not 50, how it better handles the walk, and that it uses the HR.

You can also align it to exactly 50 as league average by setting the constant for each year.  In 2015, you'd use 38 instead of 40.  Here are therefore the 10 best starts of 2015:

• 109    Max Scherzer    2015-10-03
• 104    Max Scherzer    2015-06-14
• 103    Chris Heston    2015-06-09
• 102    Max Scherzer    2015-06-20
• 102    Jake Arrieta    2015-08-30
• 102    Corey Kluber    2015-05-13
• 101    Clayton Kershaw    2015-09-29
• 101    Carlos Carrasco    2015-09-25
• 101    Cole Hamels    2015-07-25
• 99    Madison Bumgarner    2015-09-12

Game Scores actually have a fairly linear relationship to wins.  Obviously, at the most extreme it'll breakdown, but it does a pretty good job overall to represent a pitcher that averages a Game Score of 65 will win 65% of the time.

Thursday, October 01, 2015

Mariano Rivera v Clayton Kershaw

?Rivera is a no-doubt HOF, and would be so even without the post-season.  Rivera has 1284 IP in the regular season, with a 2.21 ERA.  Kershaw, from 2010-present has 1329 IP with a 2.24 ERA.  Rivera has 291 free passes (walks excluding intentional, but including hit batter), while Kershaw is at 329.  Kershaw is ahead in Ks by 281.  HR?  71 to 79, Mariano to Kershaw.  998 hits allowed by Mo, and 968 by Kershaw.  A .701 win% for Kershaw and 652 saves for Mo.

And I'm not even considering how Mo would have it easier in his workload.  But I'm also not considering the eras they each played in.  Regardless, the point is that if you can fairly compare a pitcher to Mariano Rivera, you have yourself a no-doubt HOF.

Do we really need Kershaw to have 10 years of league average performance in order to put him in the HOF when he, right now, has demonstrated a similar level of accomplishment (in the regular season) as Mo?  (And like I said, even without the post-season, Mo is in.)

This is the Bobby Orr point, that at some point, you just don't need to show any more.  You've shown enough to be considered one of the greatest players of all time.  That's enough.

Sunday, September 27, 2015

Solution to MLB’s problem of pitchers missing a year

?Just as we have concussion rules, and how many days a player has to miss, so too can you do something for pitchers who miss a year of play: they can only pitch at most once a week.

I know what you are thinking: it's going to mess with the rotation. However, that is demonstrably false.  I already showed that back when this was an issue with Strasburg a few years ago.  If you make such a pitcher the "Saturday night special", and only make this pitcher on Saturdays (has to be Saturday), the rotation can remain almost perfectly intact.

In addition, I'd put in a pitch count limit of 110 pitches for any one start, 200 pitches for 2-consecutive starts.  So, flexible enough that we don't put in a 100 pitch cap, but strict enough that it'll average out to 2600 pitches in 26 starts for the season.

Felix Sale

?Felix Hernandez has pitched 201.2 innings and allowed 180 hits, 55 of which were for extra bases.  So has Chris Sale done exactly that. 

One of them has been assigned 78 earned runs, and the other 79.  One of them has allowed 22 HR and the other 23.  One of them has faced 826 batters and the other has faced 827.

Felix has been assigned an 18-9 record, while Sale is 12-11, even though both received 4 runs of support from their offense. 

My expectation of the BBWAA voter mindset is that they will see through the W-L record and give them each down-ballot votes, enough that one of them will finish fifth and the other sixth.  All that's left is to figure out who will finish ahead of whom.

Wednesday, September 23, 2015

Pitches v IP

?I agree with the researcher that it's silly for us to talk about pitches per start, and then suddenly shift to innings per season, when discussing workload.

In terms of his regression line: it MUST have an intercept of 0!  Which means that the slope is simply the league average number of pitches per IP.  That is Pitches = 16 x IP, or whatever it is.

This is an example of a runaway regression, where the intercept is trying to capture something else, and is forced to deliver what it's delivering because of the straight line relationship.  For example, a good pitcher needs fewer pitches per inning simply because he faces fewer batters per inning.  So, Pitches = 3.8 x PA.

So, we can account for the relationship between Pitches and IP by NOT making it a straight line, but forcing the intercept at the only known and accurate point: 0,0.

Sunday, September 06, 2015

What is the Tommy John plan for Matt Harvey?

?Back in 2012, they had a discussion for Strasburg, that had cited Zimmerman as the model.  Zimm threw 2458 pitches in his comeback year.  Matt Harvey is at 2459 pitches.  Strasburg ended up at 2607 pitches.  David points out Garcia at 2609 pitches a few years before that. Harvey is also wanting to pitch in the playoffs.  So, whatever limit there is, it's probably an evolving thing, and he might be setting the new standard.  Other than Tommy John himself, who probably threw over 3000 pitches in his comeback year.

So, which of the two should throw more pitches: a young guy who hasn't built the endurance, or the soft-tosser who comes back at age 33?  I have no idea.  You tell me.

Sunday, August 23, 2015

DRA component run values

?Jonathan was kind enough to send me a link to their latest breakdown, one that we asked for in some form.  I'm writing this as I'm looking at the data, so, I don't know what we're going to find.  The focus will be on Pedro 1999 v 2000. 

In 2000 Pedro is noted as being 56 runs better than average. In terms of the main components that that number is adjusted for, it's +4 runs, meaning that Pedro's performance in 2000 was +52 runs, with about 4 runs uncovered via context. 

Now, let's look at 1999, where he's noted as "only" 40 runs better than average, and his context was -1 runs.  Meaning that he's a +41 runs before context, and when considering context, it's actually +40 runs.

So, the pre-context different of Pedro was 11 runs (+52 minus +41), with the advantage in the 2000 season.  That the 2000 season is ahead of the 1999 season suggests that the central component is Runs Allowed. 

However, each of those numbers is about 25 runs lower than would be suggested by Runs Allowed.  This would ALSO suggest that there's a regression toward the mean built in to all the runs allowed figures.  This is interesting, because this is what should happen! 

That is, assume that ALL PERFORMANCE was luck.  Given 200 IP, some pitchers will allow 3 runs per game and others would allow 5 runs per game. Just by luck, because that's the assumption. Do we want to represent that as +1 runs per game better for one pitcher and -1 runs for the other pitcher, just because they happen to be on the mound when luck happened?  Or, we'll just show then as 0, just like we would if the pitchers were an actual pitching machine.

But, if we assume that it's MOSTLY skill, we still want to regress PARTLY.  And this is what DRA is doing, it seems.  A vote for DRA is a vote for regressing the observed results by removing the portion that is luck-based.

The question therefore is, since we've gone down the regression rabbit hole, would be to determine if there's too much, or not enough regression. That's the first thing you need to do to evaluate DRA, since this seems to be the primary driver to the results.

So, I would therefore ask Jonathan and crew to also include the luck values in runs as another component to display, to make it very clear what is going on here. 

Monday, August 17, 2015

Corey Kluber: second verse, same as the first

This researcher argues that Kluber's component stats are the same, and the difference is that his team allows 1 more run per game when he's on the mound, and scores 1 fewer run per game in his starts.  Sabermetrics is about discovering the truth of the impact of his performance.  So, what is his impact?

Thursday, August 13, 2015

Game Theory on pitch selection

?Good stuff from Neil.  He has a good foundation for the framework.  As the commenters pointed out, the count will matter especially, and the rest of the game state (which you can approximate through Leverage Index) will matter to some extent.

For example, no one swings at 3-0 pitches, so, all those pitches are noise.  At 3-1 counts, pitchers are highly incentivized to throw strikes, and so, will go with a fastball, even if that's not necessarily their most effective pitch in a "neutral" count.  At 0-2, different pitchers will respond very differently.  A fastball just outside the strike zone (chase-able at 0-2) is likely more similar to a curveball in the strike zone than a fastball down the middle (in terms of effectiveness).  But Neil would lump in all 0-2 fastballs together.  And worse, he lumps in all fastballs across all counts.  Which makes the ultimate conclusion irrelevant.

While I like the basic idea that Neil has, in terms of laying out the groundwork, the house needs a frame to make it effective.  And we're a long way to get there, based on what has been shown.

Saturday, August 01, 2015

Bill James on FIP

?Well-said:

What we're always trying to do is see through the illusions created by the numbers and see what is underneath and real and the fielding independent pitching numbers are quite helpful in that respect because it's a systemized, organized effort to filter out the things that are in the pitcher's record which aren't real. They're not related to his skill, it's just something that happened. That's tremendously helpful and tremendously significant.

If you go back to the oldest way of looking at a pitcher — 1975 — pitchers were evaluated by win-loss records. You'd have a pitcher sometimes who might have an ERA of 4.80, but they scored a ton of runs for him and he finished 17-9. People actually thought that he was a great pitcher because he had this ability to pitch well enough and win.

In the modern world, we know that it's nonsense and they just scored a lot of runs for him. Even the dumbest guy in baseball knows that win-loss records aren't that reliable because the offense doesn't even out for people. That's a circumstance-dependent record. ERA is a circumstance-dependent record. But even if you filter out the illusions in ERA and the illusions in run support, some guys are just lucky. Fielding independent pitching stats are an effort to filter that out and to the extent that they're successful, it's tremendously useful to do that.

Friday, July 31, 2015

Poz v Steamer

?Poz applied his personal WAR system, a mish-mash of whatever he happens to be using at that moment, however arbitrary, biased, capricious and inconsistent he makes it.  His final number is just an ordinal ranking.  And he wraps it with wonderful words.

Steamer is systematic, consistent, and cold.  And comes without commentary.

I have a little system to convert ordinal rankings into WAR, so I did that for Poz.  And I compared the two.  So, where do they diverge?

Steamer loves: Tanaka, Carrasco, Pineda, Salazar, all of whom who didn't make it on Poz's top 30.  Plus Jose Fernandez (8th for Steamer, 23rd for Poz).  Steamer expects 7 WAR from these 5, while Poz is implied at 5.

Poz loves: Lance Lynn, Shelby Miller, Mark Buehrle, all of whom were ranked at best 45th in Steamer.  Plus Greinke (3rd for Poz, 14th Steamer), Cueto (7th, 20th).  Poz expects 7.5 WAR from these 5, while Steamer is at 4.5.

So, all we have to do is for these 10 pitchers is add up their WAR from Aug 1st to end of season, and see how they do. 

What kind of stakes can we have for Poz and Steamer (Jared)?

Starter-to-reliever conversion

I like the idea of using the playoffs to get over the selection bias issue.  Of course, sample size is our enemy.

?

Thursday, July 09, 2015

Rays and Times Thru The Order

?One of the big revelations from The Book back in 2006 was the Times Thru The Order Penalty.  That each time through the order, the pitcher would take a big hit.  Passan is looking at the Rays buying into the idea.  The Rays have been sensational the 1st time thru (best performance in MLB).  They've also faced 457 batters the third time thru, which is fewest in MLB so far,  (However, if you include 3rd AND 4th time thru the order, Royals faced the fewest batters.)

Friday, June 19, 2015

Evolving Charlie Morton is a bbFIP machine

?bbFIP is Batted Ball FIP, which includes the batted ball components, in addition to BB, HB, and SO.  Sal notes that Morton's core FIP numbers (SO minus BB+HB), which were never great to begin with, has gotten a bit worse; but he's making up for it by becoming a GB machine.  Noting that GB rates need much fewer observations in order for the signal to exceed the noise, he thinks this is a real evolution in Morton.  He further supports that by the change in his arsenal, notably reducing the fastball frequency in favor of the sinker. He also noted the drop in HR/FB, which may or may not be as real a change as other parts of his game.

In any case, I love it when pitchers evolve.  I'm looking forward to how Felix is going to evolve when he hits his late 30s.  By that point, he may become a late 20s Doug Fister. 

Frank Tanana's transformation preceded my following baseball, and his transformation was earlier and more necessary.  But I can imagine something like that as well.  Just going by supposition here.  I didn't verify anything.