Talking Sabermetrics: Third Base Coaches And Sending the Runner Home

Dave Clark seems like a great guy, a coach who is well liked by players. But his decisions as a third base coach are not popular with the fans. At least that is the impression we get from the internet. (And we all know that everything on the internet is true.) AstrosB at SB Nation Houston has mentioned the popular view of Clark, saying that he is "affectionately known as 'Stop Sign.'" And, in fairness to Clark, most teams' third base coaches make decisions that are unpopular with the fans---that's seems to be the nature of the job.

Sabermetrics Looks At Sending The Runner

I admit that I have complained a few times about Clark's third base decisions. But two articles at Beyond the Boxscore make me wonder if Clark is getting a bum rap. BtB's Jim Gentile has some great analysis in "The Slow Death of Aggressive Baserunning" and "Parsing the Demise of the Windmill." These two articles examine the decision to hold or send a runner home from 2d base with a single. If Clark is viewed as Mr. Stop Sign, that may just demonstrate that he is in line with the baseball trend of holding runners at 3d base.

Gentile's articles show that, with runner on second and a single, the hold rate at 3d base has increased continually since the 1950's. The stop sign at 3d base has increased from 25% in the 1950's to 37% in recent years. I don't want to over-generalize, but the growing trend of more conservative hold rates at 3d base seems to cross situational factors like outs, stage of the game, and scoring margins. For example, teams behind by a big margin tend to hold runners more than teams which are ahead by a big margin---and this is true throughout the last 60 years; however, the trend over the years toward more conservative hold rates affects teams on both sides of that scoring margin.

An interesting question is whether the recent trend toward more conservative decisions by the third base coach is leaving too many scoring opportunities on the table. That is difficult, perhaps impossible, to determine. But my gut feeling is that the trend probably reflects improved efficiency in scoring runs. First, Gentile's data indicates that the hold rate trend is accompanied by a slightly growing rate of the runners ultimately scoring. That doesn't prove that the trend is more efficient, but it suggests that it might be. Second, most likely the trend reflects the growing knowledge and experience of managers and coaches over time. Given the broad mix of ages and generations on MLB coaching staffs, it makes sense that experience with the stop sign is handed down from coach to coach, which results in incremental improvement over time. And don't forget that statistics and scouting information are more widely available to teams in the later years. If a team can succeed by bucking the trend, an enterprising manager somewhere would do it.

So, back to Coach Clark. We don't have data on his specific hold rates. But, given the continuing trend to hold more runners, and the possibility that this trend may produce greater scoring efficiency, maybe we shouldn't be so harsh in criticizing Clark for his stop signs. Furthermore, the short LF at Minute Maid Park probably should result in Clark exercising a higher percentage of stop signs compared to third base coaches in other parks.

There could be any number of additional hypotheses for the trend toward more conservative third base coaches. Have outfield arms improved that much? Are players slower? Are ballparks smaller? But I am skeptical of these reasons.

My guess is that the trend reflects a growing acceptance of the value of conserving outs. We know that teams used to be more aggressive in stealing bases and more willing to give up outs to move runners. Sabermetrics probably is one factor in the changing attitude, with its focus on OBP and avoiding outs. The changing run environment over the years, particularly the higher HR rates, may also play a role in the manager's mind set. Losing too many runners on the base paths may reduce the number of 3 run HRs that win games.

A Look At the Stop Sign and Win Probability

An interesting way to analyze the third base hold decision is through the lens of win probability. (This is the foundation of Win Probability Added, WPA, which shows the impact of an event on the probability in real time of winning the game.) And the Win Probability Inquirer tool at Hardball Times can be utilized to analyze the win probability associated with particular base, out, inning, and score situations. Of course, the results will be based on average players and won't take into account all of the specifics, like the base runner's speed, the outfielder's arm, the quality of subsequent batter, and so forth, that are needed for particular decisions. But it might give us some insight on the third base coach's decision.

If I had the time to model every base, out, inning, and score situation, perhaps the exercise would be more useful. But, I am not that interested in third base coaching decisions. As an illustration, I will look at the runner on 2d with the batter hitting a single for a more or less neutral situation: tie game in the middle innings. For each of the three outcomes (hold the runner at 3d, runner scores, or runner is out at the plate), I determine a post- event base-run-out state and use the Inquirer tool to calculate a win probability for the outcome. Note that I assumed that the runner who singles stays at first on an out at the plate; if the complication of a runner moving up on the throw to the plate were to be added into the analysis, it would improve the "send the runner" decision ever so slightly.

The focus of this analysis is to determine the break even success rate required to send the runner ("break even send rate"). This procedure solves for the expected success rate for sending the runner home which will produce a better win probability than holding the runner. For example, if the break even send rate is 50%, then the third base coach should not send the runner unless he believes that the runner has more than 50-50 chance of scoring. The third base coach's assessment of the scoring probability would include all of the relevant factors, such as the location of the batted ball, the outfielder's arm, the runner's speed, etc. As added information, I also show the actual hold and scoring rates based on Gentile's article.

Notice that a very high expected success rate (88%) is required in order to send the runner home with no outs. That means that the third base coach must be nearly certain that the runner won't be thrown out with zero outs. By contrast, the third base coach should just roll the dice with two outs. As long as the probability of being called safe at home is slightly above 1 in 4, the third base coach should send the runner with two outs. Even with one out, the third base coach's threshold before he should send the runner is fairly high, 71% probability.

The MLB Average is also expressed as a percentage, but it is a much different type of value, since it doesn't indicate probabilities, but rather the actual average hold and thrown out rates. (If you want to know the score rates, it's just 1 minus the out percentage.) The high and low break even probabilities are supportive of the high hold rate with zero outs and the low hold rate with 2 outs--definitely an indication that the decisions are at least roughly consistent with win probabilities. If anything, I wonder if there may be some advantage to holding more runners at third base with no outs and perhaps even 1 out. I don't have enough information to be confident in making that statement, but it seems like a 88% probability is a high enough threshold for sending the runner to support more than a 55% hold rate with no outs. On the other hand, the 1% current out rate experienced in that situation is so low that refinements in the hold decision may have only a marginal impact.

This type of analysis also gives me appreciation for the third base coach's job. We know that Dave Clark isn't taking an I-Pad into the third base box so that he can work on a spreadsheet when he makes his decision. (And even if that was possible, how easy is it to distinguish a 70% probability from a 88% probability in real time?) But think about all of the factors which are not included in the analysis, above, but should affect the decision: not just the runner's speed and outfielder's arm, but also the offensive ability of subsequent batters, the way that the ballpark plays, whether the opposing pitcher is tiring, and the ability of the catcher to block he plate.

Do you have any thoughts about the MLB trend toward holding more runners at third base? And for what it's worth, Gentile's data indicates that the AL is slightly less likely to hold runners at 3d base than the NL. (This extends prior to the DH rule, by the way.) Will that affect the Astros' tactics next year?