THE BOOK--Playing The Percentages In Baseball

There’s also comments by MGL and me on Ballhype:
http://ballhype.com/story/the_hardball_times_with_the_game_on_the_line_i_want/

I just want to highlight my answer to MGL’s objection about the head-to-head matchups, specifically Rollins v Burrell.

Actually, I’m looking at the Phillies ballots.  I received 52 of them, and in 40 of them, either Rollins and/or Burrell is on them.

Rollins was the preferred choice of the two 25 times, Burrell 4 times, and tied 11 times.  As I noted, the other 12 ballots had neither of them on.

So, it’s rather clear here that Rollins was much more highly desired, right?

I’m looking forward to the results.

Tango,

Did you refresh Luke Scott’s Marcel for the move from Houston to Baltimore? 

ZiPS, CHONE, and PECOTA all expect Markakis to be a better hitter than Scott, so you may want to re-think your choice there.

Marcel makes zero adjustment for parks.  If Fans think they pulled one over the monkey, that’s a bonus for them.

***

A good day for the Fans:
http://eastwindupchronicle.com/great-moments-in-jeff-brantleys-mouth/

Fans know clutch.

http://www.baseballthinkfactory.org/files/newsstand/discussion/tht_tango_with_the_game_on_the_line_i_want/

I am happy that the BTF thread has not devolved into a horrible clutch thread.  While the main focus on the initial comments is on Nash, I am very happy to have been able to write a clutch article where the two sides did not take that as an opportunity to come out with guns blazing.  Hopefully, that’ll last…

Here’s the head-to-head count for the 30 pairs of players:
teamID fans marcel tie
teamID fans marcel tie
ana 18 5 6
ari 11 0 0
atl 17 7 5
bal 48 2 3
bos 177 40 150
chc 52 22 41
cin 31 21 8
cle 20 8 4
col 9 4 3
cws 30 9 21
det 24 5 7
fla 5 2 1
hou 13 5 3
kc. 6 2 0
la. 26 12 19
mil 31 24 42
min 9 5 4
nym 39 15 10
nyy 80 58 44
oak 22 12 20
phi 27 7 15
pit 8 1 3
sd. 16 8 3
sea 55 2 4
sf. 30 8 9
stl 64 23 23
tb. 12 5 5
tex 14 1 2
tor 30 16 15
was 11 7 3

This is what I did with an illustration.  I looked at all ballots where Utley and/or Howard was selected.  I looked at the rating for each player.  Utley was preferred to Howard 27 times, and Howard was preferred 7 times.  They got an equal rating 15 times.

So, I think that addresses MGL’s concern for those teams where I had to go to the #2 guy.  For example, most wanted Pujols.  If you remove him, you are still left with 110 ballots where Molina and/or Duncan appear.  Molina was preferred 64-23 with 23 ties.

***

And here are the 6 wildcard picks:
teamID fans marcel tie
bos 245 24 74
det 23 3 4
nyy 75 28 41
phi 25 4 11
sea 39 3 2
tor 37 24 1

Other than perhaps Wells/Scutaro, it was a huge landslide as to who the fans preferred.  The Toronto one is more notable in that Scutaro had alot of solo ballots (only he was mentioned) while Wells had alot of “POOR” ratings.

***

Can we agree that the group of 30 is representative of the Fans’ choice?  And that the 6 wildcard picks are also Fans’ preference?

***

(16:12 ET.  Note: above lists corrected, slightly.)

Would you want me to put the wildcard picks up in the scoreboard?  I could include a separate section for them or just tack them on to the usual section.

I would say to include them in the usual section.  Simplicity is best.

The reader can always manipulate the results however he sees fit.

What was the distribution of Poors, Fairs, Averages, Goods and Greats? Did fans tend to be optimistic about clutch ability?

So far, the Fans’ players are taking a beating, but man are they hitting the homeruns.  Joe Crede is the latest hero, after Rollins and Encarnacion hit HR for the fans.

Seems that Clutch = non-strikeout, HR hitter.

I am surprised at the selection of Erstad by Astros fans. That contradicts the Rowand Rule that a fanbase has to grow an attachment to a player before they’ll consider him to be clutch.

Just out of curiosity, how did Alfonso Soriano do on the balloting?

I’m not sure that the distributions will matter, since a good portion only selected one or two hitters.  I didn’t specifically ask everyone to rank everyone.  I can still produce the numbers for you tomorrow, but I just wanted to make that part clear.

Well, for the Astros, Erstad was their SIXTH pick.  They correctly had the same order as Marcel for the top 5, though early on, it looked like they had a preference for Carlos Lee.  This was far more a case of hating Wiggington than liking Erstad.

***

Soriano did not do good at all.  DeRosa was preferred to him.  Here’s their head-to-head:
41 DeRosa, 24 Soriano, 27 tie

http://fangraphs.com/statsp.aspx?playerid=227&position=3B&season=&sort=

Crede’s HR in the top of the 8th that tied the game was NOT part of the clutch pool.  The LI was 1.87, and I drew a line at 2.00.

Perhaps at the end, if the Fans end up losing at the 2.00 threshhold, I’ll ask David to re-run, using 1.50 as the threshhold.

I wonder if the titles on the two scoreboards at fangraphs should be reconsidered.  As it is, it might give the impression that it’s a question of whether Tom Tango’s “better” or “smarter” than the fans...which isn’t really the point.

I’d suggest “Clutch hitters” vs. “Top hitters” or something like that.

Anyway, just a suggestion.
-Justin

So far, in the clutch, Marcel’s guys have 28 of 63 PA as walk or strikeout.  17 of 64 Fans’ PA are walk, K, or HBP.

Looking at the players, it seems that Fans think of Clutch as “contact the ball”.

So far, Marcel has a slight lead…

I asked David to include hits on BIP:
http://www.fangraphs.com/clutch.aspx?type=0

Fans are simply getting killed at the moment with their clutch players, in large part due to the hits on BIP.

Luckily for them, there is no doubt that their ultra-low hits on BIP will rise in the future.

So far, each side has around 110 PA, after around some 110 league games, which is fairly close to what I expected:

.10 clutch PA per PA
x 550 PA per player per 2430 league games
x 110 league games
x 36 players
= 90 PA

Fans continue to get killed:
http://www.fangraphs.com/clutch.aspx?type=0

After around 180 or so PA, the WPA/LI for the better hitters is +1.2 wins compared to the clutch hitters of -0.3 wins.

Tango, your team obviously does not make good enough contact to be suffiently clutch. 46 K’s in 188 PA’s is not acceptable, we need more productive outs!

Best regards,
Cory

You are right.  That’s almost 50% higher than the average rate, and double the Clutchers rate.  The key to being perceived as clutch is to look good making an out.  This is a corollary to the Billy Beane “we’re not selling jeans” law.

I wonder if there’s an enormous subset of players in the minor leagues that do not have the tag of prospect because they strike out too much. 

Here’s a project for someone to do: go through the Baseball America lists of 5-10 years ago, and find 2 similar players in terms of MLE (BA, OBP, SLG), but one strikes out far more than the other.  Come up with one matched pair per team in that time frame.
Q1: Where to BA rank these players?  I’ll bet the K-players ranked several spots lower than the nonK-players.

Q2: How did these players perform in MLB (if they were even called up)?

I think it’s interesting that Tango’s team has 14 IBB v. 3 for the Fans’ team. Maybe there’s a big difference in the number of clutch PA with first base open, but certainly opposing managers perceive the better hitters as more dangerous in those spots to date.

Ooooh… great point.

Tango, how do those IBB affect the WPA for both teams (yours and the fans).

I would LOVE to see someone do that study you suggest.  They can probably do the same thing with other possible biases, such as BA being overvalued, walks being undervalued, HR being overvalued, etc., or even size of the player.

It would be sad if it were true that organizations like BA would significantly over or underrate players based on these things.

As you mention, one of the problems for such a study will be selective sampling - underrated players might not get much playing time, if any, in the majors, whereas the overrated ones will get lots more playing time.  All you can do is make sure you use rate stats in the study and that you don’t use any “cutoff points” (min number of PA in the majors) for the inclusion of players.

BTW, doing those two things in a study, having minimum “something” for inclusion, or using counting stats, like Win Shares, is often a recipe for selective sampling disaster, depending upon what question you are trying to answer of course.

The fans didn’t like Frank Thomas?  Big mistake.  That choice is 1/4 of the difference in WPA so far.

IBB barely affects WPA (probably +.015 wins per IBB), and affects WPA/LI even less (probably +.005).

***

MGL: can you post your MLEs of Ryan Howard?

I moved the Ryan Howard posts here:
http://www.insidethebook.com/ee/index.php/site/comments/ryan_howard/

In addition, read posts 14 and 17 for contemporous accounts of Howard.

What a difference a week makes.  The gap last week was +1.5 wins in 180 PA, in favor of the “best hitters” team.  Now, that gap has closed to +0.5 wins in 260 PA. 

The Fans’ hitters bumped up their BABIP like crazy to .290.

Excluding the IBB, and the two stat lines are very close.  This is actually great news to the Fans.  After all, they knowingly went into this choosing the worse hitter overall, but expecting to at least be as good as the better hitter with the game on the line.

That is, the started with a 20-point wOBA disadvantage, and are close to making up for it with their clutch play.

They are getting good help from their wildcards as well.

Maybe Clutch hitting does exist, and maybe the fans know who it is that has it.  We’ve still got a long way to go…

The Fans have actually taken over!

http://www.fangraphs.com/clutch.aspx?type=0

By Situational Wins (WPA/LI) after 350 PA, Fans’ Clutch players are now at +0.5 wins, while my overall better hitters are hitting league average.

Remember that the Fans were already at a 20 wOBA point disadvantage to start this thing (which is about being 0.5 wins in the hole after 350 PA).  They not only overcame that, but added another 0.5 wins on top of that!

Maybe Fans do know Clutch.

It’s a very tight race (once you remove the IBB).  SLG are a match, the better hitters are getting a few more regular walks (and a whole bunch of IBB).

David at Fangraphs, with blinding speed, added a new column for my clutch project called “OBP2”, which is simply the OBP with the IBB removed in both the numerator and denominator.

Since the great hitters were getting a ton of IBB, greatly inflating their OBP, I thought it would make more sense to make the comparison without them.

We see that the numbers are fairly close, but any edge you want to give would go to the great hitters.  RBI, BABIP, BA, OBP, SLG, WPA, WPA/LI each give the edge to the great hitters.  The clutch hitters however do make more contact (though not more successfully).

But, we’re only 20% of the way through, and things change pretty fast every week.

Now tied in WPA/LI, and Fans/Clutch are LEADING in OPS (excluding IBB).

Fantastic job by the fans here, as they were really behind the 8-ball in their group of players selected.

I’ll still take the non-clutch guys at the end of the season, if anyone wants to wager!