WhartonBall

Right Idea.... Wrong Reason

2008-04-29T12:41:00.000-07:00

Darren Everson reports in the WSJ of a growing trend among managers to try new things. One idea is to not bat the pitcher last in the lineup. I am not sure this really makes sense since the negatives probably outweigh the positives. A second idea is to have the "Starting" pitcher begin in the second inning. I have long argued that this is a really good idea, but not at all for the reason cited by Doug Melvin (Brewer's GM):

The thinking is that starters are typically among a team's best
pitchers, yet nowadays they often pitch only through the fifth or sixth inning, well before many games are decided. By having them pitch later, they'd be around for the higher-leverage innings.

Right Idea. Wrong Reason. The concept of leverage is important only for the purposes of resource management. The star closer is a scarce resource useful only for one inning at a time. So you would not want to waste it in a low-leverage situation. Ergo, you save the closer. What is forgotten in this calculus is that you inevitably fail to use the closer in many games where his use in an early inning could have changed the game most.

Now consider the first inning. Because team's lead with their best hitters the most runs are scored in the first innings. Since all runs count the same (regardless of the inning in which they are scored) the first innings has the highest expected leverage. This means that if you could go back and erase a single inning and replace it the inning most likely to make the most difference is the first inning. Ultimately what this means is that you want to use a star reliever at the start of the game to face the most feared hitters. Then your starter should take over in inning two. A second star reliever (if you have it) should be used for high-leverage situation towards the end.

Update of SAFE values

2007-12-07T08:22:00.000-08:00

I've updated our SAFE fielding values after doing some modifications to the methodology, such as improved measures of shared consequence, park dimensions, and individual fielder curves:

SAFE Website

Nice Fielding Comparison

2007-11-13T11:30:00.001-08:00

Nice blog entry and article on comparing current systems for quantifying fielding:

On Baseball and The Reds

Red Sox Nation my Ass!

2007-10-26T14:07:00.001-07:00

If you happen to read this:

NY Times Article

You may be surprised to read this:
"Boston's 2004 win ended an 86-year championship drought and exorcised what generations of fans had called a "curse."But it also spawned "Red Sox Nation," an ardent base of traveling supporters who sometimes outnumber home team's fans at Boston road games.Not even the revered New York Yankees played in front of bigger regular-season crowds on the road this year."

While technically true the red sox "out drawing" the Yankees away from home is due entirely to the fact that the Red Sox play the Yankees away at Yankee stadium (big park) and the Yankees play the Red Sox away at Fenway (Small Park).

According to ESPN the Yankees outdraw the Redsox with respect to total percentage of available seats filled. It was not indicated on the site whether this a straight percentage average by game or the faction of total seats sold/ total seats available. Anyway the Yankees lead the world at this number 83%. The RedSox are second at 81%.

see:

ESPN Link

No matter how you slice it. Red Sox fans can't do math.

NESSIS 2007 Talk

2007-10-15T20:17:00.000-07:00

The organizers have posted video of my recent talk on SAFE at the New England Symposium on Statistics in Sports

Preseason Predictions

2007-09-25T12:10:00.000-07:00

Great blog entry on comparing pre-season predictions of team records by several analysts (mostly from ESPN):

Vegas Watch

Further evidence that Steve Phillips really has no idea what he is doing.

2007-08-20T10:47:00.000-07:00

My recent article with Matthew Carruth on throwing ability has gotten a bit of attention and led to some interesting discussion:

Phil Birnbaum's Sabermetrics Blog

Evaluating Catcher Throwing Ability

2007-04-22T19:30:00.000-07:00

Our last post was on an evaluation of outfielder throwing ability, where success/failure on individual throwing events were tabulated for each outfielder in terms of runs saved/cost. We now apply the same methodology to catchers: the Run Expectancy Matrix provides us with expected runs for each game situation, which we can use to calculate a runs saved/cost value for a particular successful or unsuccessful play. For catchers, this is the ability to throw out base stealers as well as keeping players from trying to steal. We tabulate the runs saved/cost for every catcher in each of four seasons of ball-in-play data (2002-2005) and then average across all four seasons (taking into account differences in opportunities between seasons).

The table below gives the five best and five worst catchers, in terms of average runs saved/cost in throwing events.

Ten Best Catcher Arms

Name	Runs Saved
1 . Schneider Brian	6.554
2 . Molina Yadier	4.406
3 . Hall Toby	4.258
4 . Ardoin Danny	3.669
5 . Miller Damian	2.998
6 . Molina Bengie	2.835
7 . Blanco Henry	2.377
8 . Rodriguez Ivan	2.372
9 . LaRue Jason	1.816
10 . Wilson Vance	1.772

Ten Worst Catcher Arms

Name	Runs Cost
1 . Piazza Mike	-4.975
2 . Varitek Jason	-3.199
3 . Martinez Victor	-3.001
4 . Zaun Gregg	-2.78
5 . Fordyce Brook	-2.679
6 . Fasano Sal	-2.61
7 . Lieberthal Mike	-2.505
8 . Bennett Gary	-2.231
9 . Wilson Tom	-2.111
10 . Hundley Todd	-1.87

The full results for catchers can be downloaded here.
The use of the expected run matrix to evaluate the run consequence of stolen bases leads to an interesting consequence: in terms of expected runs, it is more valuable to throw out a baserunner attempting to steal than preventing a baserunner from attempting to steal. Catchers that have a reputation for throwing out baserunners will not have a great run contribution because baserunners will attempt to steal less often. An example is Ivan Rodriguez who is considered to be the best catcher in the game. However, Rodriguez is ranked as only the eighth best catcher by our analysis, in part because he has one of the lowest proportion of steal attempts against him (3.5% of baserunning opportunities) among regular catchers. Clearly, the optimal situation for a catcher is to have a high success rate on throwing out baserunners but without the reputation for doing so, so that baserunners still attempt to steal at a substantial rate. An extreme (and not recommended) implementation of this strategy would suggest that catchers could deliberately fail on a throwing attempt in an relatively unimportant game situation in the hopes that baserunners would then be more likely to attempt (and be thrown out) in a more important situation.

Preceptorial Lecture

2007-04-03T14:10:00.000-07:00

For those of you at my recent Preceptorial lecture on baseball statistics, I have posted my lecture notes online:

Preceptorial Notes

I removed the slides on our projection system that I didn't get a chance to cover in the lecture. The removed slides were very "work-in-progress" anyway, and I'm hoping to post several more refined blog entries on our projection project in the near future.

Evaluating Outfielder Throwing Ability

2007-03-18T14:55:00.000-07:00

All of our previous SAFE fielding evaluations have only focussed on the ability to make outs on balls in play. However, another dimension of fielding is the evaluating performance of throwing events by individual players. We have just finished an analysis that quantifies the relative throwing ability of each individual outfielders by tracking all throwing opportunities in the 2002-2005 seasons and evaluating the relative success of each fielder relative to the league average on the scale of runs saved/cost.

The Run Expectancy Matrix provides us with expected runs for each game situation, which we can use to calculate a runs saved/cost value for a particular successful or unsuccessful play. For outfielders, we want to measure ability to throw out baserunners trying to take extra bases, as well as keeping baserunners from trying to take extra bases. We tabulate the runs saved/cost for every outfielder in each of four seasons of ball-in-play data (2002-2005) and then average across all four seasons (taking into account differences in opportunities between seasons).

The table below gives the ten best and ten worst outfielder arms, in terms of average runs saved/cost in throwing events.

Ten Best Outfielder Arms

Name	Runs Saved
1 . Edmonds Jim	8.718
2 . Jones Jacque	8.659
3 . Taveras Willy	8.074
4 . Johnson Kelly	7.772
5 . Sullivan Cory	7.372
6 . Chavez Endy	6.524
7 . Guerrero Vladimir	6.197
8 . Hidalgo Richard	6.181
9 . Hunter Torii	5.974
10 . Walker Larry	5.852

Ten Worst Outfielder Arms

Name	Runs Cost
1 . Brown Emil	-16.783
2 . Pierre Juan	-10.769
3 . Lawton Matt	-9.428
4 . Sanchez Alex	-7.793
5 . Holliday Matthew	-7.341
6 . Crawford Carl	-7.111
7 . DeJesus David	-6.996
8 . Williams Bernie	-6.785
9 . White Rondell	-6.559
10 . Magee Wendell	-5.973

The full results for outfielders can be downloaded here. I'll post another entry soon that applies the same methodology to the throwing ability of catchers in base-stealing opportunities.

Interview with Balls, Sticks & Stuff

2007-02-09T07:38:00.000-08:00

I was just interviewed by the website Balls, Sticks & Stuff on our SAFE method for evaluating fielding. The full interview can be found at:

Balls, Sticks & Stuff Interview on SAFE

Fielding Adjusted Pitcher ERAs

2007-01-20T12:08:00.000-08:00

In a previous blog entry, we discussed team-level measures of SAFE, our Spatial Aggregate Fielding Analysis measures. One way in which these measures can be used is to re-evaluate the past performance of pitchers and account for the defense working behind them. We take the runs saved/cost by each team's defense in a given year and partition those extra runs among the pitchers based on their number of innings pitched. We then recalculate each pitchers ERAs with these extra runs saved or cost factored in. The full set of results for the 2002-2005 seasons are available for download at:

Pitcher ERA Adjustments

As a preview of the results, we give a table of the top ten relief pitchers in terms of the percentage change between their old and new (adjusted) ERAs.

PLAYER	TEAM	YEAR	IP	New ERA	Old ERA	Diff	%Diff
Dan Plesac	Phi	2003	33.1	2.19	2.7	-0.51	-19.0
Mariano Rivera	NYY	2005	78.1	1.14	1.38	-0.24	-17.4
Josh Hancock	Cin	2005	14	1.63	1.93	-0.30	-15.7
Derrick Turnbow	Ana	2003	15.1	0.68	0.59	0.09	14.6
Mariano Rivera	NYY	2004	78.2	1.67	1.94	-0.27	-14.1
Fernando Cabrera	Cle	2005	30.2	1.67	1.47	0.20	13.3
Tom Gordon	NYY	2004	89.2	1.94	2.21	-0.27	-12.0
Brian Shackelford	Cin	2005	29.2	2.15	2.43	-0.28	-11.7
Jorge DePaula	NYY	2003	11.1	0.70	0.79	-0.09	-11.5
Matt Miller	Cle	2005	29.2	2.02	1.82	0.20	11.1

We see that most pitchers which show improvement in their ERA played for the Yankees or Reds, the two teams with the worst defense in the 2002-2005 seasons. This same trend is seen in the top ten starting pitchers in terms of the percentage change between their old and new (adjusted) ERAs.

PLAYER	YEAR	TEAM	IP	New ERA	Old ERA	Diff	%Diff
Aaron Harang	2005	Cin	211.2	3.51	3.83	-0.32	-8.5
Brandon Claussen	2005	Cin	166.2	3.88	4.21	-0.33	-7.9
Jason Schmidt	2004	SF	225	2.95	3.20	-0.25	-7.8
Shawn Chacon	2005	NYY	151.2	3.19	3.44	-0.25	-7.4
Kevin Brown	2004	NYY	132	3.79	4.09	-0.30	-7.3
Jason Schmidt	2003	SF	207.2	2.17	2.34	-0.17	-7.1
Jon Lieber	2004	NYY	176.2	4.04	4.33	-0.29	-6.8
Brett Tomko	2004	SF	194.0	3.77	4.04	-0.27	-6.6
Randy Johnson	2005	NYY	225.2	3.54	3.79	-0.25	-6.5
Ramon Ortiz	2005	Cin	171.1	5.02	5.36	-0.34	-6.4

No 20 Game Winners

2007-01-06T12:21:00.000-08:00

An interesting and surprising fact about the 2006 season is that neither league had a single pitcher that won 20 games. Chien-Ming Wang of the Yankees and Johan Santana of the Twins both won 19 games. No one in the NL won over 16 games. Has it ever happened before that there were no 20 game winners in either league? Disregarding strike-shortened seasons, the short answer is no. However, there have been several occurrences in baseball history where one of the leagues has gone without a 20 game winner. These seasons are given in the table below (excluding strike-shortened seasons):

Year	League	Top Pitcher(s) in Wins	Top Wins
1931	NL	Elliott, Hallahan, Meine	19
1955	AL	Ford, Lemon, Sullivan	18
1960	AL	Estrada, Perry	18
1982	AL	Hoyt	19
1983	NL	Denny, John	19
1987	NL	Sutcliffe	18
2006	AL	Santana, Wang	19
2006	AL	Harang, Lowe, Penny, Smoltz, Webb, Zambrano	16

We can see from the table that not having a 20 game winner is pretty rare, and not having even a 17 game winner in the NL has never happened before. Was 2006 just an unlucky season for pitchers, or is this a manifestation of an overall decreasing trend in wins by starting pitchers? The graph below shows the highest number of wins by a pitcher in each league over the years. Not surprisingly, there has been a substantial decreasing trend, with a very slight reversal of that trend in the 1950-60 period.

The overall trend is similar in the following figure, which shows the average win totals of the top 25 wins by a pitcher in each league over the years. The increase in pitcher wins during the 1950s and 1960s is more dramatic when looking at average over 25 pitchers, as well as the corresponding return to the downward trend from the 1970s onwards.

I think we can expect a few more seasons without 20 game winners in the future. Thanks to David Gurian-Peck for his help with this analysis.

SAFE for Entire Teams

2006-11-22T19:07:00.000-08:00

There have been several recent posts about our Spatial Aggregate Fielding Evaluation (SAFE) model and results. We've previously applied SAFE to the analysis of fielding performance for individual players. We've now improved our results to include SAFE measures for entire teams. The full results for teams can be seen on our SAFE website:

SAFE Website

The table below gives an abbreviated set of results: the SAFE measures for each team averaged across the 2002-2005 seasons. As a reminder, SAFE measures give the runs saved or runs cost relative to an average team. The White Sox, Devil Rays and Mariners come out on top, while the Yankees, Reds and Giants stand out as being exceptionally bad.

Team	SAFE 02-05	Team	SAFE 02-05
Chicago White Sox	15.6	Milwaukee Brewers	-0.1
Tampa Bay Devil Rays	14.0	Texas Rangers	-1.4
Seattle Mariners	13.4	Chicago Cubs	-2.7
Atlanta Braves	12.6	Philadelphia Phillies	-5.2
Anaheim Angels	11.0	Colorado Rockies	-6.8
Los Angelos Dodgers	10.0	San Diego Padres	-7.0
Minnesota Twins	9.3	New York Mets	-9.2
Cleveland Indians	8.8	Boston Red Sox	-10.3
St. Louis Cardinals	8.3	Florida Marlins	-10.4
Pittsburgh Pirates	7.6	Arizona Diamondbacks	-11.2
Houston Astros	7.0	Montreal Expos	-17.3
Oakland Athletics	3.9	Kansas City Royals	-18.0
Baltimore Orioles	3.5	San Francisco Giants	-22.2
Toronto Blue Jays	1.0	Cincinnati Reds	-23.8
Detroit Tigers	0.3	New York Yankees	-35.4

The Playoffs are a Coin Toss

2006-10-29T12:19:00.000-08:00

The improbable success of the St. Louis Cardinals this October has many people talking about the inherent randomness of the playoffs. We examined the outcomes of every playoff series since the start of the modern world series in 1903. We tabulated the number of times that the "better" team won a playoff series, with "better" being defined as the team with the superior winning percentage in the regular season. The proportion of series won by the better team are given below:

All Series since 1903: Proportion = 0.529
Only 5 Game Series: Proportion = 0.512
Only 7 Game Series: Proportion = 0.533

We see that all the proportions are pretty close to 0.5, which suggests that a playoff series isn't much different from a coin toss. In fact, none of these proportions are significantly different from a null hypothesis of p = 0.5. Not surprisingly, we also see that 5-game series are even closer to being a coin toss than 7-game series.

We can also look at the trend over time by calculating the proportion of the better team winning as a 20-year moving average around each year. The plot of this moving average proportion is given in the plot below:

We see that the proportion of times that the better team wins seems to be generally decreasing over time, though there are have been some other interesting dips in the past. Finally, we can look at some of the greatest playoff upsets, in terms of the winning team having a much lower regular season winning percentage compared to the losing team. A table of the top 5 greatest upsets is given below:

Year	Round	Winner	%-age	Loser	%-age
1906	WS	Chicago White Sox	0.616	Chicago Cubs	0.763
2001	ALCS	New York Yankees	0.594	Seattle Mariners	0.716
1973	NLCS	New York Mets	0.509	Cincinnati Reds	0.611
1954	WS	New York Giants	0.630	Cleveland Indians	0.721
2006	NLCS	St. Louis Cardinals	0.516	New York Mets	0.599

A special thanks to Matt Koizim for assembling the data and running the initial analysis.

2004 ALCS Win Probabilities

2006-10-23T19:20:00.000-07:00

I recently happened upon the following very interesting analysis of the 2004 ALCS between the Yankees and the Red Sox.

Win Probability charts for the 2004 Red Sox-Yankees ALCS

The site includes a chart of the overall probability of winning the series as a function of time, as well as the same plots for each individual game. Also included is the WPA (Win Probability Added) values for individual players in each game. Very interesting stuff.

For a brief explanation of WPA, check out the Wikipedia entry:

Wikipedia: Win Probability Added

SAFE for fly balls

2006-10-15T19:33:00.000-07:00

I've added our SAFE methodology for fly balls-in-play (flys, pop-ups, and liners) to the SAFE website, as well as our full results for infielders and outfielders based on the 2002-2005 seasons. The website link is still:

SAFE Website

Still to come is a discussion of overall team fielding, but in the meantime, let us know what you think.

SAFE for grounders

2006-10-09T20:48:00.000-07:00

In my introduction to SAFE, I described some of the issues with quantifying fielding ability that we are attempting to address with our spatial modeling for fielding ability. I have put together a website for SAFE that describes some of our methodology and results. Check it out at:

SAFE Website

So far, the website only discusses our methodology and results for the fielding of grounders by infielders. I will add our analysis of outfielders at a later date. I will also eventually add a discussion of overall team fielding. Your comments are welcome!

Fielding as a Continuous Measure

2006-08-24T19:17:00.000-07:00

One of the aspects of baseball that is the hardest to quantify and evaluate is fielding ability. Most events in baseball, such as hitting events, are discrete which makes them easy to tabulate and model probabilistically. The central difficulty with fielding is that we are trying to evaluate players on a continuous playing surface where we must take into account not just whether a successful play was made, but whether a successful play was possible. The much-maligned error statistic is a subjective attempt at discretising this phenomenon: players are assigned an error if the official scorer deems that their unsuccessful play should have been successful. However, tabulating errors isn't a good measure of ability without a corresponding measure that credits a player for making a play that most players wouldn't have.

Recent techniques such as Ultimate Zone Rating or the Plus-Minus system from The Fielding Bible are based on the tabulation of both positive and negative fielding events. These statistics are more detailed and accurate measures of fielding ability. These techniques are also getting a bit more attention is the regular media, as evidenced by this recent Yahoo! Sports article. However, despite being obvious improvements on previous methods, both of these approaches are still based on dividing the baseball field into discrete zones and vectors, and tabulating events within each zone. Ideally, the baseball field could be treated as the continuous playing surface that it actually is, instead of a set of zones or vectors.

Over the next few weeks, we will present results from our own approach to this problem, which we call SAFE: Spatial Aggregate Fielding Evaluation. We use the same raw BIS data used for The Fielding Bible. However, instead of tabulating fielding events within discrete zones, we fit continuous probability distributions to each fielder based on their past fielding events. The closest technique (at least in spirit) to our approach is the work by David Pinto. Once we have probability models for each player, we use numerical integration to calculate the expected runs that they save or cost their team.

We'll give a more detailed outline of our methodology, along with lots of results in the upcoming weeks. As a teaser, though, we'll just say that Adam Everett really is a good shortstop and Derek Jeter really is a bad one, but the results are not as dramatic as The Fielding Bible would lead you to believe!

Welcome!

2006-08-22T14:54:00.000-07:00

The primary goal of this blog is a discussion of interesting and novel statistical research for the study of baseball. The founders of this blog are two professors (Shane Jensen and Abraham Wyner) in the Department of Statistics at the University of Pennsylvania. We are overseeing an active baseball statistics research group that started last year in collaboration with ESPN.

Our plan is to post our own research as often as possible, and we would really like to hear feedback as well as tips on other interesting baseball research out there on the web.

More soon...