Not So Sweet Surrender
“A good sacrifice is one that is not necessarily sound but leaves your opponent dazed and confused.” – English chess prodigy Nigel Short
“Sure, I screwed up that sacrifice bunt, but look at it this way. I’m a better bunter than a billion Chinese. Those poor suckers can’t bunt at all.” – John Lowenstein, former Baltimore Orioles outfielder
As I perused the various sections in The Bill James Handbook 2006, I came across the “Managers Record” chapter towards the back of the book. There each 2005 manager is listed along with a variety of statistics spanning his entire career that serve to give you a profile of his tendencies. These include number of lineups used, percentage of platoon advantage in his lineups, pinch hitters, pinch runners, defensive substitutions, quick and slow hooks, long outings, number of relievers used, stolen base attempts, sacrifice attempts, intentional walks and pitchouts.
There’s some very interesting information embedded in these tables, including the fact that Bobby Cox of Atlanta and Bob Melvin of the Diamondbacks led the majors in starting players with the platoon advantage at 69% and 68% respectively, while Jack McKeon and Alan Trammell did so only 43% and 49% of the time.
I also noticed that Dusty Baker led the world in pitchouts with 70 (the average was about 23), while Frank Robinson of the Nationals called for only four. What’s even more interesting is that the difference apparently had no effect on the running game; the Cubs caught 40 of 90 (31%) would-be base stealers while the Nats caught 41 of 76 (35%). That’s probably not much consolation to the Dodgers, however, who nabbed just 34 of 130 base stealers (21%) and called for just 17 pitchouts.
Robinson though wouldn’t be outdone in his penchant for giving up outs and putting runners on for free. He called for 115 sacrifices and 77 intentional walks, both of which led the majors.
All of which leads me to the main topic of this article.
What really caught my eye in the Manager’s Record were the sacrifice attempts called for by each manager, a category added in 2006. Last May on my blog I speculated a little on how often sacrifice attempts were successful in an attempt to provide some context to this comment Bill James made in an interview on Baseball Digest Daily:
“…the general argument against the bunt seems unpersuasive to me. The essential argument against the bunt is that the number of expected runs scored after a bunt attempt goes down in almost all situations when a bunt is used, and the expectation of scoring one run goes up only in a few situations.
But this argument is unpersuasive to me, because it assumes that there are two possible outcomes of a bunt: a ‘successful’ bunt, which trades a base for and out, and an “unsuccessful” bunt, which involves an out with no gain. In reality, there are about a dozen fairly common outcomes of a bunt attempt. The most common of those is a foul ball, but others include a base hit, a fielder’s choice/all safe, a pop out, a pop out into a double play, an error on the third baseman and a hit plus an error on the third baseman, or the second baseman if you’re talking about a drag bunt.”
In that quote James calls into question the standard argument used against the bunt based on run expectancy and scoring probability tables. For those who aren’t familiar with that argument it goes a little like this.
Using play-by-play data one can calculate on average how often and how many runs were scored after each of the 24 possible base/out combinations. For example, in the period 1999-2002 the run expectancy for a team with a runner on first and nobody out was 0.953 runs with a scoring probability of 0.437. In other words teams averaged about a run once this situation obtained with a 44% chance of scoring one or more runs. The entire scoring probability table is shown below.
Base/Out 0 1 2 Empty .293 .173 .077 1st .437 .283 .136 2nd .632 .406 .223 3rd .864 .662 .263 1st/2nd .641 .426 .231 1st/3rd .876 .655 .285 2nd/3rd .856 .695 .276 Loaded .872 .670 .325
Using a simple formula that Pete Palmer and John Thorn introduced in The Hidden Game of Baseball one can then calculate the break-even percentage for a particular strategy. That formula is:
Break Even % = (Pv – Fv) / (Sv – Fv)
where Pv = Present Value, Fv = Failure Value, and Sv = Success Value.
In the case of the sacrifice with a runner on first and nobody out, the scoring probabilities are Pv = .437, Fv = .283 (runner on first and one out) and Sv = .406 (runner on second and one out). When you do the math you get a break-even percentage in excess of one (actually 1.5). When the result is greater than one it means that even with a success rate of 100% a sacrifice attempt costs you more in terms of scoring probability than you gain. The same calculation can be applied to run expectancy to determine when a team should sacrifice when the objective is to maximize the number of runs they score.
Once all of these calculations are done (and if you’d rather than do them by hand you can download my strategy application) it turns out that for the 1999-2002 period a sacrifice is worth the risk only in the following situations:
and when…
Of course the above is based on the assumption that an average hitter is at the plate. Pitchers are the most often called upon to sacrifice, and they are anything but average. We can adjust for that as well by adjusting the probability of success based on an average pitcher’s performance. When that is done the following situations make a sacrifice attractive:
Base Out Score Maximize 1st 0 .679 No 1st 1 .820 No 1st/2nd 0 .359 .621 2nd 0 .563 .939
As you’ll notice there are still a minority of situations where sacrifices appear to make sense. However, James contends that sacrifice attempts are more valuable than you would think at first glance because they have ancillary affects other than simply giving up an out.
And that brings me back to the sacrifices-attempted column in The Handbook.
The nice guys at Baseball Info Solutions tell me that in calculating the number of sacrifices attempted they use the following definition:
As you might guess, and as James noted, this definition doesn’t cover the full gamut of sacrifice attempts, because this calculation is based on play-by-play results and not observation. Notably this does not include the situation when the batter struck out while attempting to sacrifice, nor does it include sacrifice attempts with runners on third (suicide and safety squeezes) or sacrifice attempts by position players when there are fewer than two outs (something we regrettably saw Willy Taveras do multiple times during the postseason).
So I decided to take another crack at this and try and determine just how often sacrifice attempts are successful and how this jives with the run expectancy and scoring probability break-even points.
So How Many Sacrifice Attempts Were There?
To do so I went through the play by play data for 2005 and created my own set of plays that could reasonably be termed sacrifice attempts. These included:
As you can see, this is a more liberal definition than that used by The Handbook, but it’s one I think more accurately captures the actual number of attempted sacrifices. There are still holes however, as it fails to account for bunt attempts by non-pitchers with one out where a sacrifice was not credited by the official scorer. There were 81 such occurrences with the following outcomes:
Single 39 Double Play 5 Force out 11 Ground out 18 Pop out: 6 Batter interference 2
With these it’s difficult to make a reasonable assumption as to whether or not a sacrifice was attempted, so I’ve chosen to leave them out. There are also likely some non-sacrifice attempts lumped into the first category above, so my hope is that these will have the effect of canceling some of those out.
When you total it all up that means there were 2,355 sacrifice attempts in the majors in 2005 (The Handbook total was 2,138), which, broken down by team looks like this:
Team SacA WAS 132 COL 126 FLO 125 SFN 118 HOU 113 SLN 112 ATL 108 MIL 108 ARI 101 SDN 101 CHN 98 NYN 96 PHI 93 PIT 90 LAN 86 CIN 80 CHA 74 SEA 61 KCA 60 ANA 60 MIN 60 DET 57 BAL 54 CLE 54 TBA 51 NYA 44 TOR 30 OAK 30 BOS 22 TEX 11
I mentioned at the beginning of this article that Frank Robinson’s Nats attempted 115 sacrifices according to The Handbook. The total here is 17 higher since my criteria is a bit more liberal. Can you believe that Buck Showalter called for just 11 sacrifices?
The players who were called on to sacrifice most often were:
Name SacA Omar Vizquel 25 Chris Carpenter 20 Coco Crisp 19 Brandon Webb 19 Luis Castillo 19 Livan Hernandez 18 Juan Pierre 17 Neifi Perez 17 Willy Taveras 17 Andy Pettitte 17 AJ Burnett 16 Nook Logan 16 Ramon Ortiz 15 Royce Clayton 15 Jack Wilson 15 Jamey Carroll 15 John Smoltz 15
Why did I have a feeling that Neifi Perez would be high on the list? Incidentally, in his 17 attempts Perez executed 12 times, struck out once, popped out three times and grounded into a force out once.
And when broken down by base/out situation the totals for the majors look like this.
Base/Out 0 Pct 1 Pct 1st 1001 0.425 321 0.136 2nd 415 0.176 5 0.002 3rd 0 0.000 22 0.009 1st/2nd 404 0.172 105 0.045 1st/3rd 21 0.009 51 0.022 2nd/3rd 1 0.000 3 0.001 Loaded 0 0.000 6 0.003 1842 0.782 513 0.218
As you would expect, there were many more attempts with nobody out than with one out, and the most attempts were with a runner on first and no outs.
How Successful Were They?
Now that we’ve got a data set with which to work we can look at how often sacrifices were successful.
As a first pass we can count those sacrifice attempts as successful if the sacrifice was successfully executed or if no outs were recorded on the play by the defense. In other words, this includes all the attempts where the sacrifice was put down successfully plus all those attempts where the batter reached safely because of an error, fielder’s choice or beating out the bunt for a single. This then encompasses most of the outcomes that James had envisioned.
When we use those two criteria we find that sacrifices were successful 76.2% of the time in 2005. The teams ranked by success rate were:
Team SacA Succ Pct KCA 60 50 83.3% SEA 61 50 82.0% BOS 22 18 81.8% TEX 11 9 81.8% ANA 60 49 81.7% BAL 54 44 81.5% CLE 54 44 81.5% HOU 113 92 81.4% DET 57 46 80.7% SFN 118 95 80.5% TBA 51 41 80.4% NYA 44 35 79.5% MIN 60 47 78.3% TOR 30 23 76.7% OAK 30 23 76.7% CIN 80 61 76.3% FLO 125 95 76.0% SLN 112 85 75.9% COL 126 95 75.4% PHI 93 70 75.3% ATL 108 81 75.0% ARI 101 75 74.3% MIL 108 80 74.1% NYN 96 71 74.0% WAS 132 96 72.7% CHN 98 71 72.4% SDN 101 73 72.3% LAN 86 62 72.1% PIT 90 63 70.0% CHA 74 50 67.6%
So despite getting a lot of practice, the Nationals were not stellar at giving themselves up. And despite being in the American League where the success rates are higher (since position players apparently are better bunters than pitchers), the World Champions also had some problems, although it didn’t really hurt them since they relied much more on the home run, on which they scored 42.4% of their runs, good for fourth in the majors.
Incidentally, I’ve always wondered what the success rate is when asking pitchers to bunt with two strikes. Well, in 2005 they were asked to bunt with two strikes 207 times and were successful (they didn’t strike out) just 65 times or 31%. Non-pitchers weren’t any more successful, however, and were 14 of 43 for 32%.
We can then fill in each cell of the base/out matrix with the success rate that teams actually experienced in 2005 as follows:
Base/Out 0 1 1st 0.780 0.695 2nd 0.882 0.800 3rd N/A 0.864 1st/2nd 0.696 0.514 1st/3rd 0.810 0.882 2nd/3rd 0.000 0.667 Loaded N/A 0.333
So what does this tell us? Well, when we compare this against the previous table that indicated when sacrifice attempts are generally worth it, we can see where the league beat the break-even percentage and therefore where teams generally are making good decisions. Those turn out to be (in bold in the previous table):
And of course squeeze plays are always a good bet when you need a single run since they’re often successful (over 86% of the time with a runner on third).
The most interesting of these is the large difference between the break-even success rates with runners on first and second of .359 and second with nobody out of .563 and the actual success rate of .696 and .882 respectively. This may indicate that managers don’t risk attempting to move the runner to third even with a below-average bunter as often as they should, especially if that run means tying the game or going ahead.
Now of course as discussed previously, these break-even percentages were based on a pitcher hitting, so the percentages would rise somewhat if we weighted them by the actual number of sacrifices attempted by pitchers versus position players. If we looked only at position players then there wouldn’t be any situations where the league as a whole would beat the break-even percentage.
The Wisdom of Weaver
In the end what this tells me is that Earl Weaver’s fourth and fifth laws as laid out in his book Weaver on Strategy: The Classic Work on the Art of Managing a Baseball Team, still makes a good deal of sense when applied in general. Those laws are:
But as in most of life, although you may live by a few general principles there will always be times you have to adjust when the context demands. After all, even Earl Weaver asked John Lowenstein to bunt now and then.
References & Resources