THE BOOK--Playing The Percentages In Baseball

Just in time for Hamilton’s arbitration hearing…

By the way, do you ever sleep?

Will you be publishing all of your base running projections? I’d be interested in seeing how you project the Phillies, particularly Jimmy Rollins and Domonic Brown.

In case it’s not clear, that was mgl…

The base running numbers for all players/teams should be up on FG this year, updated at the same time as all their other numbers.

Rollins is projected at +2.9.  I don’t use minor league numbers, so I don’t have anything for D Brown.

Surprisingly, Longoria, who is a big man, is +2.6 in base running true talent.  That really enhances his already high value.  Ryan Zimmerman, another big third baseman, is also a good base runner, +3.2.

Kotchman is still playing evidently.

http://www.mlbtraderumors.com/2011/01/rays-to-sign-casey-kotchman.html

Speaking of big third basemen, how does Scott Rolen do?  I always have heard that he is an excellent baserunner for his speed.  Heard the same thing about Pujols too, though he is very aggressive.

According to Dan Fox’s baserunning reports, Scott Rolen grades out close to average:
http://www.baseballprospectus.com/statistics/sortable/index.php?cid=146291

2010: -1.5 runs
2009: +1.5 runs
2008: -1.0 runs
2007: +1.6 runs
2006: +0.8 runs

I have Rolen at 2.3 runs per 150 (projection/true talent) and Pujols at .1.  Rolen used to be a SS, believe it or not.

I find it much easier to believe that Rolen was a SS in high school or college than he is a 3B in MLB.  He is so !@#$% smooth, surehanded, and strong, that it’s shocking he is at 3B.  I can only imagine that he was blocked at SS somehow that he thought going to 3B would be faster.

To a smaller extent, I have the same question with Zimmerman, Beltre, and a few others.  I’m astounded by the quality of fielding talent at 3B this past decade.

Anyone have pre-MLB history they can share on these guys: why 3B not SS?

According to:
http://www.thebaseballcube.com/players/r/scott-rolen.shtml

Rolen’s been a 3B since he was 18. 

His high school:
http://www.jockbio.com/Bios/Rolen/Rolen_bio.html

Having reached his full height, Scott moved to third base full-time. His coach, Terry Gobert, believed his teenage star had a brighter future at the hot corner. Gobert was also reluctant to have his best relief pitcher at a short, where his arm would endure more wear and tear.

Scott settled in quickly at thrid. His range was so great that Japser’s shortstop could poisition himself close to second base, and the second baseman could cheat toward first. Scott made the whole infield better—a rare feat at this or any other age.

Scott was voted the top high school baseball player in Indiana that spring,making him the unprecedented king of two major sports simultaneously.

Thanks MGL!

It is always nice when statistical analysis matches up with my eyes, though I learn more when it doesn’t.

Thanks MGL.  I look forward to your base running numbers going onto FanGraphs. I assume that SB/CS will be pulled out of Batting WAR and moved into a new Base running WAR component on FG.

BTW, those base running numbers above do NOT include SB/CS numbers, which are pretty straightforward.  They are movement on batted balls only (no advances on WP and PB either).  I don’t know whether FG will include SB/CS in the base running numbers or they will be separate.  I assume they will be combined.  So once you have offense, fielding, and base running, what you have left for a player’s total value is the most important part, of course - heart, grit, etc.

MGL - Would it be possible to give us a little more information on your methodology and possibly run through the calculations on Figgins as an example?  Thanks.

Sure, Peter.  I’ll be back later.

SuperLWTS from 10 years ago, including a quick explanation for baserunning:

http://www.baseballthinkfactory.org/files/primate_studies/discussion/lichtman_2001-08-27_0/99slwts.html

How do you control for the fact that a guy like Figgins tends to be running in front of guys who hit the ball hard? It’s a lot easier to go first to third when you’re running in front Russ Branyan than when you’re running in front of Ichiro. Unless you can control for this, I don’t see how base runner is a reflection of the runner rather than the hitter (or the base coach, for that matter.)

phil: you are correct, which is why you want a WOWY approach.

That said, over a 4 year period, you are mostly (but not totally) going to have that cancel out.  You’ll have the perpetual pairing in Philly (Rollins-Utley-Howard), but even then, it’s not like EVERY single that moves Utley from 1B to 3B is because Howard hit it.

So, objection is valid, and it’s up to us to show you the degree to which this is an issue.

Good point…

Indeed, the flip side of my comment (#18) needs to be addressed by saberists: wOBA assumes that all singles advance base runners equally, but I’m willing to be there’s a significant difference in the runner advancing value of Jim Thome’s singles and Chone Figgin’s singles. Maybe the effect is only a few runs per season for the most extreme hitters, but that’s near the value of base running skill.

phil: this is handled by RE24.

MGL also breaks down his singles by infield/outfield.

There are plenty of different metrics that treat singles differently and you should use those.

This is not a wOBA issue, as it takes the path of least resistance.

Harder struck hits may reach the outfielder quicker, resulting in fewer advances. Balls hit away from the outfielder, where perhaps a faster batter might make a double, would show up as an advantage to the baserunner’s ratings. It probably should be a function of the score, base/out (LI) and the description of the batted ball...we have GB/LD/FB, but there we have a lack of data on hardness and location and use single/double instead. With field f/x, I’d look at where the ball was hit and how long it took the fielder to get to the ball.

The net result will be a rounding error.  This is something that is more fun to talk about and figure out then to actually see the results.

A typical batter has some 50 singles hit with a runner on 1B or 2B.  If he can get an extra .04 runs per single, that’s +2 runs.  And that’s going to be the extreme point.  Virtually all batters will be +/- 1 run.

I’d be more worried about park effects than batter effects.

The reason I asked MGL what his model looked like is that there are a lot of factors that can be considered as the last few posters have pointed out.  Trying to include all of them makes for a very complicated model that has very small sample sizes for each situation.  So there is a tradeoff.  I was interested in which factors MGL had decided were important enough to include in his model and which he had decided were inconsequential enough to ignore. 

I briefly looked at Figgins baserunning for the last few years.  One thing that struck me was the number of times that he went 1st to 3d because he was running on the pitch.  Much higher than I would have expected.  The success of those plays is highly dependent on Figgins speed, but there is also the ability of the batter to put the ball in play and the team philosophy of how often to use runner in motion plays to consider.

The other point to consider is the run value for advancement in different out states.  Some analysts give the same run value for taking an extra base with 2 outs as with 0 outs. Others give the actual change in RE for the specific before and after baseout states.  I think that it is important to do the latter when assessing a runners actual past contribution to his team’s offense.  But it could be argued that that is not the best way to evaluate a player’s “true” baserunning skill as the situations that a runner faces varies with his batting order position and the skills of the batters that follow.  Since MGL is projecting a player’s future ability I wondered how he had decided to assign his run values.

Not sure if my comment #20 was understood. I certainly wasn’t being very clear. The thought was that, ignoring the issue of base running and returning to hitting, we may undervalue certain hitters ability to advance runners while not advancing themselves, and likewise of certain hitters to advance themselves while not advancing the runner. Not all single advance runners equally. Perhaps LHB have an advantage because the throw from right to third is harder than the throw from left to third; but wOBA won’t capture that. RE24 seems like a better metric, but it won’t sort the value of the batter from the runner he advances.

@Brian, #22: yes, harder hit balls may reach the OF faster and make it harder to advance. But the OF may also play deep to avoid extra base hits, making a ducksnort harder to play, making the throw to third more difficult, etc. I’m not saying this is enough in take MGL’s results as a good approximation, but I would like to hear about how this gets handled in the model and whether we have any good estimates of the value of this effect.

It’s noteworthy that all of MGL’s top base runners bat 1, 2 or 3 for their teams. Now, that may just be a fact about managers valuing base runner skills in lead off hitters, but it’s hard to ignore that they may have benefited from running in front of Vlad, Hunter, Hamilton, Cruz, Reynolds, Upton, and Pujols.

"It’s noteworthy that all of MGL’s top base runners bat 1, 2 or 3 for their teams. Now, that may just be a fact about managers valuing base runner skills in lead off hitters, but it’s hard to ignore that they may have benefited from running in front of Vlad, Hunter, Hamilton, Cruz, Reynolds, Upton, and Pujols.”

I have to check, but I use the location of the batted ball (and outs of course) to determine the baseline advancement.  I also use park effects. Of course, that does not consider the depth of the outfielders (based on the batter), which is obviously a consideration.  Maybe I’ll tweak the methodology, but I doubt it will make that much of a difference.

I also do not consider whether the runner was running on the pitch, as Peter mentions, which is an important consideration.  In fact, I probably overvalue the fast runners and undervalue the slow ones.  I am not sure if I even have that data (whether the runners are moving on the pitch).  I suppose I could separate the 3-2 counts, certainly with 2 outs.

Obviously, ideally, you would want to control for everything (outfield arms, etc.), but for one thing, as Peter also mentions, the more you try to control for, wither the smaller the sample sizes or the messy and complex (if you use smoothing functions ore multiple regressions) the methodology becomes.

As Tango says, I doubt that a basic methodology is going to be more than 1 or 2 runs different for all players than a more granular one, and probably less than 1 run for most players.

BTW, this is one reason why it is so valuable to vett methodologies on forums like this!  There are so many things that one person will overlook when constructing a methodology (or doing an analysis), but when you get 20 really smart guys together, hardly anything gets overlooked!  I have tweaked UZR many times based on feedback from all the amateur sabers out there!