Quantifying Coachers, Part I
"The employment of one of the side who are in to watch the movements of the field and advice the runner accordingly is a quaint device of American acuteness." - quote from an English newspaper during an 1874 tour by American ballplayers as recorded in Sporting News, February 25, 1909
Unlike today, however, it seems the primary job of the coaches was to "disconcert the opposing players - generally the pitcher - not to 'coach' or assist the base-runner" as Sporting News put it in 1893. As a result, the primary qualifications for a coach of that time was a megaphone like voice (yes megaphones were tried in college games in the early 1900s but fortunately never found a foothold in professional baseball) and a cruel disposition. In fact, it was the abusiveness of coachers like Charles Comiskey and Bill Gleason, who would stand on either side of the catcher commenting on everything from his skill as a catcher to his breeding and personal habits, which precipitated a move to first ban coaching altogether but then to restrict coaches to boxes down the line beginning in 1886. In addition to their primary job as unsettlers of the opposition, third base coaches would also attempt to get opposing fielders to mistake them for a runners, a ploy was which was severely hampered following the 1886 rule change. Be that as it may coaching did eventually come to be taken more seriously with Arlie Latham the first full time coach hired by the Cincinnati Reds in 1900. And Latham was apparently performing the modern function since in July of that season Sporting Life reported that "Manager [Bob] Allen says he is delighted with the coaching of Latham. He says the baserunning of the team has improved 100% after the veteran got on the lines". As is true for advances in other fields Cincinnati's experiment proved to be an early, if successful, trial balloon and it would another decade before the idea took hold. In between there were still calls to ban coaching leading non other than Henry Chadwick in 1904 to denounce coaching as it had "degenerated into a dirty-ball method of annoying the pitcher". Other innovations in the game including increased managerial strategy and signaling finally drove the need to move beyond the "old school of clowns" as Christy Matthewson famously said of this earlier period in Pitching in a Pinch. Not surprisingly it was John McGraw who hired Latham and Duke Farrell as full-time coaches in 1909. By 1912 Sporting Life noted that Latham "does get a percentage of runs across - runs that might not otherwise be made". From there it was generally recognized that coaches paid dividends and although for some time there was apparently a subset of coaches who seemed more preoccupied with rattling the opposition, coaching as a profession gained strength and was here to stay by the early 1920s. Their duties have expanded over time as well. As just one example the first base coach now routinely times the pitcher's delivery to the plate relaying that information to the runner. It's now been almost 100 years since full time coaches were employed and their performance is routinely scrutinized although not very often quantified. The question then (first suggested to Dan by Rich Lederer of all people) from an analyst's viewpoint is two-fold. First, is the job of coaching quantifiable? In other words, can we create a metric or metrics that measure the success and failure of this component in a reasonable way? And second, if it is measurable, can some coaches be said to be more skilled at this half of their job than their peers? In this article and the one to follow we'll take a crack at answering both questions for third base coaches related to their secondary job (relaying signs being the primary) of directing traffic on the bases.
When totaled, these give us a fairly complete picture of the contribution made by a player on the bases beyond what would have been expected given their opportunities. And therein lies the rub. The methodology that underlies these metrics isn't a simple totaling of the number of bases gained in these situations but rather an application of changes in the expected number of runs across several dimensions including the base/out situation (the Run Expectancy matrix), handedness of the batter, and the position of the fielder who fielded the ball. By calculating how often runners typically advance in a whole host of scenarios (for example with a runner on second and nobody out a runner will advance from second to third 43% of the time when the ball is fielded by the shortstop but 97% of the time when handled by the second baseman) and translating those percentages to runs using the Run Expectancy matrix we can credit or debit a runner for each and every opportunity they have on the bases. Totaling the credit assigned to each opportunity (and not crediting the runner for advancing the minimum number of bases) for players allows us to assign a number of theoretical runs above and beyond what a typical player would have contributed given the same opportunities. Yes, theoretical since these metrics, being based on models like the RE matrix, don't actually measure the precise number of runs contributed by a runner but rather can be thought of as an accounting of the decisions made by runners and coaches, that put their teams in more or less advantageous situations throughout the course of a season. That accounting is performed in terms of runs. As mentioned above we then adjust for park effects where necessary. For example the spacious Coors Field outfield allows for easier advancement than the smaller Fenway Park. Already many of you can see where this is going. EqHAR, by measuring runner advancement on hits, may be an appropriate methodology to apply to third base coaches since it measures an aspect of the game in which third base coaches are directly involved. Looking more closely, EqHAR is composed of three basic scenarios. A third base coach may be active in each of these scenarios but as will be obvious it typically depends on where the ball is hit. When a batter singles or doubles with a runner on first base, the runner typically makes his own decision about whether to advance if the ball is hit to left field or within his field of view in center field. On the other hand he'll usually pick up his third base coach if the ball lands in right field. Likewise when on second base ball hit to the outfield typically results in the runner typically taking matters into his own hands only if the ball is hit to left, but rely on the coach if the ball is hit to center or right. By using these general rules as a guide the analysis can be restricted in this sense to plays that fall only into these categories but also include scenarios when multiple baserunners are on base. One might argue that these categories are either too restrictive or not restrictive enough and we have sympathy with both arguments. For example, with the runner on first on a single fielded by the centerfielder there are certainly occasions when the runner picks up the coach. Conversely, with a runner on second and the batter singling to left there are definitely times when the runner knows the ball will be difficult to handle or is running with the pitch and so heads home without consulting the coach. This analysis will not include those events. And these events of course do not include runners attempting to advance on ground ball and fly ball outs nor does it include runners attempting to stretch doubles into triples or triples into inside the park homeruns. The thought was to error on the side of caution and include only those events where it seems the third base coach would be most likely to have influence. Further, these scenarios will include times when runners run right through the stop sign given by their frantic coach only to get thrown out. Through no fault of his own, the coach will be still be debited for plays like these. Surely this is far from a perfect system but given the granularity of the play by play data available and absent video inspection of each play, this seems like a reasonable approach for a first pass at creating this kind of metric. The primary advantage to using the methodology described above as opposed to simply counting the number of runners that were thrown out on each coach's watch is that this system also gives appropriate credit when a runner advances successfully. The system also takes into consideration how difficult the advancement event was and gives more credit when a runner takes a base in a higher reward situation. While keeping runners from getting thrown out is clearly a major component of the job, knowing when to take risks based on game situation is a secondary component and one that this metric captures. Given the above caveats we ran the EqHAR framework for third base coaches for 2006 with the following results.
Team Name Opp OA EqHAR Rate ANA Dino Ebel 238 3 10.3 1.19 PHI Bill Dancy 262 5 7.8 1.15 HOU Doug Mansolino 214 1 5.6 1.11 TBA Tom Foley 163 1 5.3 1.15 DET Gene Lamont 240 5 5.0 1.10 FLO Bobby Meacham 199 4 2.3 1.05 NYN Manny Acta 228 3 2.3 1.05 KCA Luis Silverio 237 4 2.0 1.04 WAS Tony Beasley 239 6 1.5 1.03 COL Mike Gallego 247 3 1.5 1.03 ARI Carlos Tosca 275 6 0.5 1.01 MIN Scott Ullger 222 3 0.5 1.01 BAL Tom Trebelhorn 296 3 0.3 1.01 MIL Dale Sveum 214 5 0.3 1.01 SDN Glenn Hoffman 231 4 -0.2 1.00 TOR Brian Butterfield 237 6 -0.4 0.99 CLE Jeff Datz 274 5 -0.7 0.99 CIN Mark Berry 217 5 -0.8 0.98 SLN Jose Oquendo 230 5 -1.1 0.98 PIT Jeff Cox 230 3 -1.2 0.98 SEA Carlos Garcia 226 6 -1.5 0.97 SFN Gene Glynn 220 3 -2.2 0.95 TEX Steve Smith 234 5 -2.5 0.95 CHN Chris Speier 199 6 -2.9 0.94 ATL Fredi Gonzalez 231 6 -3.3 0.94 NYA Larry Bowa 289 5 -4.1 0.93 OAK Ron Washington 245 7 -4.9 0.89 LAN Rich Donnelly 260 9 -6.0 0.90 BOS DeMarlo Hale 248 5 -7.6 0.86 CHA Joey Cora 234 9 -7.7 0.86 This table includes the number of hit advancement opportunities (Opp), the number of times runners were thrown out advancing (OA), the EqHAR for those opportunities, and a Rate statistic that is the ratio of EqHAR to the expected number of advancement runs given both the quantity and the quality of opportunities along the axes mentioned above. This is important since you'll notice that while Baltimore and Tom Trebelhorn had 296 opportunities, Tom Foley in Tampa Bay had just 163 and all other things being equal, more opportunities means a higher EqHAR. It should be noted that the coach was assigned all plays for the 2006 season for his team since there is no easily accessible record of when a third base coach was not on the field for his team. For example, although Chris Speier took a several day leave of absence beginning July 20th after being arrested for DUI earlier that week, the opportunities during that time are credited to Speier. Through this analysis the coaches were assigned opportunities based on their team's media guides for the respective seasons. So under this measure Dino Ebel of the Angels played a part in helping his runners to the tune of just over 10 additional theoretical runs (the second highest of any single season from 2000 through 2006) while Joey Cora was complicit in costing the White Sox the equivalent of almost 8 runs. Intuitively, this range seems to be within the bounds of believability. Interestingly, newly minted managers Ron Washington (-4.9) and Fredi Gonzalez (-3.3) don't come out very well although Manny Acta (+2.3) does. But is this really a fair gauge of a third base coach's influence? We'll answer that question along with the two we started this article tomorrow.
Neal Williams is the president of the Rocky Mountain chapter of the Society for American Baseball Research.
Schrodinger's Bat: Hit the Ground Running |
Comments
It would be nice if the defensive and baserunning gurus provide us with home and away splits. seeing is believing and I am not believing the adjustments made for park effect are accurate, especially where Fenway is concerned. So how bad are Manny or Demarlo Hale? I think Manny is an average defensive LFer except last year when he was bothered with his knees, and Demarlo Hale is an above average 3rd base coach (and well above average in holding runners at 3B to prevent them getting thrown out at home). The home and away splits may prove me wrong, but until I see them, I am not buying into any of these stats where Fenway is involved.
Also, the Red Sox had some of the worst base runners in the league last year and some players like Manny do not even pay attention to the 3rd base coach, so I will be interested in your answers rommorow. The run environment of a team and park, not to mention lineup position, are also factors in deciding to risk advancing players an extra base. A lead of batter for the Red Sox is unlikely to risk advancing a base with Papi and Manny up next unless it is a sure thing, while teams like the Tampa Bay Devil Rays have no such constraints.
Posted by: Paul Todd at March 14, 2007 2:07 AM
Theres the numbers verifying what i knew to be true last year(and probably voiced here on occassion).
CHA Joey Cora 234 9 -7.7 0.86
Thank God his 'promotion' by his hetero-life partner Ozzie takes him off of that bag.
Posted by: Eric at March 14, 2007 11:31 AM
and it was more that 8 runs for sure.
more like 20
Posted by: Eric at March 14, 2007 11:32 AM
Thanks for the itneresting history on coaches. I suspect the propensity to send runners home has much to do with 1.) The speed of the player 2.) Game situation 3.) Throwing ability of the outfielder and cutoff man and 4.) Who's up next and overall team offense.
Posted by: Al Doyle at March 14, 2007 12:09 PM
But is this really a fair gauge of a third base coach's influence?
I hope the answer to that is no. If the algorithm put Ron Washington is at the bottom of the list, the algorithm isn't passing my smell test.
Washington was fabulous at gauging the four things Al Doyle listed above. I've watched the A's closely for years, and there aren't many times the A's have had a runner thrown out that I thought Washington made the wrong decision given the context. I'm sure my fellow A's fans agree, and how often do fans think their team's third base coach is good?
Posted by: Ken Arneson at March 14, 2007 1:18 PM
Paul, for Hale his rate was .86 on the road and .86 at home and so there was essentially no difference. But your point about what he had to work with is good one. Tomorrow's article will correct for this (as it will for Ron Washington and even Joey Cora).
Posted by: DanAgonistes at March 14, 2007 4:45 PM
I love the idea.
It seems to me that you'd want to regress on the non-3B-coach-influenced baserunning measures onto the 3B-coach-influenced measure, to account for the player's speed.
For example, the 80s Cardinals probably did well, and the 3B coach gets a bonus because of that. However, using a speed-factor, as I describe above, should handle part of that.
I agree about the home/away: I just don't trust any analyst, on any of this, be it my favorites like Dan or MGL or anyone else. Even myself.
Posted by: tangotiger at March 15, 2007 9:42 AM
I am curious as to why you chose to assign responsibility to the coach with a runner on 2nd (and a single) only on hits in the players view?
Conventional wisdom, which is basically true (from playing, coaching and watching thousands of games), is that the third base coach is ALWAYS responsible for sending runners home or not, regardless of where the ball is hit.
Yes, it is true that a player is more likely to disregard or oppose the coach when he can see the ball, but doing so is rare, regardless of where the ball is hit.
I think your results will be more reliable (larger sample size) if you used ALL batted balls when a runner is sent home or not, regardless of where the ball is hit in the OF.
Posted by: MGL at March 15, 2007 10:10 AM
Tango, I'm not sure I understand your comment on regressing the non coach side. Since both coach and non coach opportunities use the same baserunners, isn't the speed factor accounted for?
I'll publish the park factors I used on my blog.
MGL, I did *not* assign responsbility to the coach for that scenario (runner on 2nd and the ball hit to left) since the ball is in view which is what I think you meant. I was thinking anecdotally about players running through their coaches signs but I didn't run the numbers the other way but will to see how it turns out.
Thanks
Posted by: DanAgonistes at March 15, 2007 10:42 AM
Dan,
I didn't see your part 2 until now, and you in fact addressed my issue in part 2.
Tom
Posted by: tangotiger at March 15, 2007 2:12 PM
Dan,
What I meant is that the coach is ALWAYS responsible for sending the runner HOME or not, regardless of where the ball is hit in the OF.
First to third is another story. In general the coach is responsible on a ball hit out of the runner's view, but many players look over their shoulders and make their own decisions.
Posted by: MGL at March 15, 2007 9:32 PM
MGL, thanks for clarifying and I'll see who changing the assignment works.
Posted by: DanAgonistes at March 16, 2007 11:54 AM
While I thought the whole thing was a great idea and great work, the fact that there are weak y-t-y correlations suggests that all we are seeing are random flucs anyway.
Posted by: MGL at March 16, 2007 1:15 PM
To address the year-to-year correlations, use an AR1 covariance matrix to see what the intra-class correlation is (as you have multiple years). That'll give you a better idea of how much this is an endogenous skill and how much it's just random fluctuation.
Posted by: Pizza Cutter at March 18, 2007 3:34 PM
Methodological concern: not all "extra bases" and potential throw-outs are created equally. Suppose that the RF finally picks the ball up in the corner with the runner already 20 feet to home plate. Even Victor Martinez himself could score from there. Should the third base coach get credit for that? In theory, it should all even out in the end, and yes to test this, you'd need to video inspect all such plays over a season...
Posted by: Pizza Cutter at March 18, 2007 9:37 PM
Agreed Pizza Cutter. Or alternatively use zone information (which we don't have at this point).
Posted by: DanAgonistes at March 20, 2007 11:43 AM