Coming home with Jackieby Shane Tourtellotte
April 10, 2013
I did something out of the ordinary early last year. Having never written a single piece of baseball research, I yet found myself prodded to action by the page dedicated to Jackie Robinson at Retrosheet. I studied his steals of home, put the numbers through some sabermetric paces, and produced this.
And am I ever glad I did. I offered it to Dave Studeman here at The Hardball Times, and he snapped it up. It got a pretty good reception from readers, nudging me to try my hand at the game again, and again. It wasn't long before I was a regular fixture here.
Obviously, I have a soft spot for my original Robinson article. More, something I wrote near the end has stayed with me, a promise left pending, unredeemed. I said that once fuller statistics were in for Robinson's career, somebody, perhaps I, should examine the following proposition:
Hypothesis: Jackie Robinson accumulated more run-producing, game-winning value with his steals of home than with all his other career steal attempts put together.
When I wrote that, numbers for neither Expected Runs nor Win Expectancy/Win Percentage Added yet existed for the early part of Robinson's career, but there was good reason to hope they would be compiled soon. I composed the original article around Expected Runs, but so far my source for those numbers has not extended them back to 1947.
Baseball-Reference did recently extend its play-by-play records, including WE and WPA, back to the 1947 season. Not all 1947 games had full PBP at the start, but all 1947 Brooklyn Dodgers games did. I suspect we have Allan Roth, the statistician hand-picked by Branch Rickey in 1947 to work with the Dodgers, to thank for that specific completeness. He personally scored every Dodgers game in that era, and his records survive.
Here was the chance to keep my promise. More, I could do it with a superior metric, counting the wins Robinson created with his steals rather than runs. That's what I'm writing about today. Those who don't remember the original piece are invited to go back and read it over now, but if you still recall the main points or just want to dive in here, go right ahead.
I began my cataloguing of Robinson's steals, thinking my path to an answer was clear. Before long, I started hitting the potholes on that road, all due to the problematic definition of a "play."
Example: Robinson is sent running from first on a two-strike count. The batter strikes out, swinging or looking, as Robinson steals the base. Baseball-Reference counts both changes to the base-out state as one occurrence, assigning it a single WPA value. By this measure, Robinson appears to have cost his team win probability by stealing second.
Or perhaps he's caught stealing, in which case the negative values for both the strikeout and the caught stealing are hung around his neck. In both cases, two values are amalgamated, and it is not possible to tease one out from the other, not from B-R's records.
There are several other ways this happens. Robinson steals home, but a trailing runner is put out trying to take home behind him. Robinson steals home, but an attempt to nab a trailing runner at third produces an error and scores that second run. Robinson steals second, but two errors allow him to come all the way home. (All of these scenarios happened during Robinson's career.)
I had two tasks. First, I needed to decide how to handle multi-element plays. Second, I had to find a way to measure the proper WPA for the elements of those plays that I was counting for Robinson.
The first task wasn't that hard. Some applied common sense, plus a judgment call here and there, gave me this list of rules:
1. Errors on a play count in Robinson's favor, unless they were made on attempts to put out other runners. If the catcher chucks the ball into center field, Robinson's play forced that error, and he deserves credit for the extra base he takes. If said catcher chucks the ball into left field trying to cut down Gene Hermanski at third while Robinson steals home, Robinson is not credited for Hermanski coming home. (In that real-life play, the error went to the third baseman, apparently frozen by all the action.)
Likewise, if Robinson was put out on continuing action, that counts against him. Example: On July 9, 1950, he stole second against the Phillies, advanced to third on catcher Andy Seminick's error, then was thrown out trying to take home. I count it all, and the total result is just as though he had been out at second. In the same vein, outs made on other runners going for an added base aren't debited to Robinson.
2. All gains on a multiple steal are credited to the lead runner. If Robinson's heading home on a triple steal, the bases his teammates gain count for him. If he's on first for a double-steal of second and third, it's a null event for Robinson.
This isn't a perfect rule, as sometimes it can make sense to try to put out the trailing runner. Yet consider this: Jackie at first and Player X at second pull a double-steal. Who would Player X have to be for you as the catcher to throw to second? And how many plausible Player X's played for Robinson's Dodgers? In this context, the rule works pretty well.
3. Robinson is not liable for the batter striking out on a busted hit-and-run, but is liable for his own advancement or failure to advance. I could not exclude these plays without hunting down every other missed hit-and-run not resulting in a strikeout and excluding them too.
I will, however, track the two attempted steals of home, both failed, that came on busted plays: a hit-and-run on July 26, 1947 and a suicide squeeze on May 13, 1956. Going for home on such a play is materially different from going for second base. I'll do the calculations according to the above rule, but when appropriate I'll note how Robinson's numbers would look if not charged for those two times he was truly hung out to dry.
4. By a similar token to rule 3, Robinson is responsible for his success on delayed double-steals. I filtered them out at one point in my original article to make a big point on success rates, but I cannot exclude them here. Taking the result at a discount would require me to invent an arbitrary percentage, and I can't see how that's any better than just counting it fully.
Figuring out the WPA of those separated plays was trickier. Baseball-Reference's data no longer helped. I went elsewhere for the numbers—that elsewhere being here.
The Hardball Times website hosts the Win Probability Inquirer. This lovely gadget lets you calculate a team's Win Expectancy in nearly any situation you can dream up, as well as figuring the Win Probability Added going from one situation to another. This is perfect for my purposes: I can include and exclude whatever elements I need to.
The Inquirer also thankfully factors in run environments. I calculated expected runs per game for the Dodgers, by year and park, and ran the scenarios accordingly. As I mentioned in my original article, Robinson's Dodgers were a high-scoring team, both through their offensive punch and the hitter-friendly characteristics of Ebbets Field. This is broadly reflected in the numbers, retarding gains and amplifying losses compared to our modern run environment.
(I did have one hitch here. The WPI allows runs per game values in half-run increments from 3.0 to 6.5. However, the 1951 Dodgers were expected to produce 6.8 runs per game in Forbes Field, Pittsburgh, which rounds to a 7.0! Robinson did make two attempts in that environment, swiping second both times. I did a little extrapolating from 6.0 and 6.5 values to get something that satisfied me. The effect only registered at the fourth decimal place, but I mention it for thoroughness. Oh, and the Dodgers actually over-performed expectations, scoring 94 runs in 11 games at Forbes that season. Helps when you're batting against the '51 Pirates' pitchers.)
I set aside Baseball-Reference's WPA numbers and went entirely with THT's Win Probability Inquirer. Having numbers from two sources would confound the data, whatever work it might save me.
The stolen base and caught-stealing numbers come from the game-by-game records at Baseball-Reference. Allan Roth's records, compiled at Retrosheet's tribute page to Robinson, credit him with one fewer stolen base in 1947. B-R's game accounts show that added stolen base (whichever it is), and I calculated its WPA value, so I am counting it.
When I ran the numbers, I came up with these results. The following table is for regular-season play. I will cover his performances in the World Series at the end of the article.
Year SB CS SBH CSH WPA WPA-2 WPA-3 WPA-H 1947 28 11 3 1 +0.5676 +0.2380 +0.1547 +0.1749 1948 22 14 5 3 +0.1353 -0.1741 +0.0074 +0.3020 1949 37 16 5 2 +0.1900 -0.0534 +0.0888 +0.1546 1950 12 5 1 1 +0.0436 +0.0125 -0.0115 +0.0426 1951 25 8 1 1 +0.1094 +0.1713 +0.0276 -0.0895 1952 24 7 1 0 +0.3294 +0.2387 +0.0125 +0.0782 1953 17 4 0 1 +0.1053 +0.0432 +0.0675 -0.0054 1954 7 3 1 1 +0.1489 +0.0274 +0.0604 +0.0611 1955 12 3 1 1 +0.0658 +0.0145 +0.0505 +0.0008 1956 12 5 1 1 +0.1282 -0.0440 +0.1117 +0.0605 Total 197 76 19 12 +1.8235 +0.4741 +0.5696 +0.7798
I have some notes of interest before grappling with conclusions. First, there was often value in Robinson's stolen-base plays aside from the bases he stole. On 197 successful steals, 26 times he advanced at least one extra base on an error the defense made trying to catch him. That comes to 13.2 percent. The authors of The Book estimated that base stealers get an extra base eight percent of the time, inching up to nine percent for the fastest and most disruptive runners. Jackie's outlying success might be due to inferior steal defense in his day—or maybe he was just that fast and disruptive.
In three of the instances, he made home on a steal of third, and 21 times, he made third after stealing second. And twice, he came all the way home from first base on the steal. How disruptive an opposing team would consider this, one can only imagine.
This aggression didn't come without cost. Twice Robinson was put out on continuing action after a steal (not counting the World Series—the story on that one is coming later). One of those was the Andy Seminick play in July of 1950 that I noted above in my Four Simple Rules. Still, a 26/2 ratio produced plenty of value. You'd break even or better stealing third with two outs at that rate—even if the disruption you caused was to your manager's heart rhythm.
If you're wondering about a breakdown of steal attempts by base: Robinson was 150-54 stealing second, 28-10 going for third, and 19-12 coming home. Percentages are 73.5, 73.7, and 61.3, respectively. Second and third was all the same to Robinson, apparently.
Steals of home would rise if you factored out busted plays like hit-and-runs or suicide squeezes, as I mentioned in my original article. If you do that, though, you need to do likewise for busted plays on other steals. Robinson was 13-10 on busted hit-and-runs that resulted in a strikeout (or in one case a walk when he was going for third). Subtract those from his steals of second and third, and the combined percentage nudges up from 73.6 percent to 75.3 percent. No surprise that Robinson was better when you left him alone to figure out when to go.
Now for the WPA numbers. Robinson's career is seemingly defined by stealing other teams blind. (Okay, we know what it really is defined by, but I meant in the games themselves.) Despite that, the value he produced with stealing is not overwhelming. Some of the reasons I touched upon in the original article: percentages not that far above break-even, a concept nobody really grasped back then, and not tailoring his aggression to game situations, which was understood better but not exactly a science either.
This falls in with the caution modern analysts have raised about the running game. Even a successful base thief won't add a mountain of value, unless with extremely high success rates. Robinson didn't have those, though he did have the opportunism to turn one stolen base into two or three. And stealing home could be very rewarding at seemingly low rates.
Just not quite as rewarding as I had thought. Robinson produced more WPA value stealing home than either second or third, but my hypothesis involved beating the number for second and third combined. He missed that by about a quarter-win: +1.0437 for second and third, +0.7798 for home. My speculation last year was incorrect.
However ... steal attempts are not the only way that base-thieving aggressiveness can register with Win Probability Added. Let's go one level deeper.
Three outcomes (maybe less true than those others)
A runner looking to swipe the next base risks more than being called out 90 feet away. He also risks getting picked off. Robinson encountered this occupational hazard of the base-stealer 23 times during his career. (This counts pure pickoffs, not the pickoff-caught stealing, which is counted with regular steals.) Given his lifetime 197-76 record on steals, this is a substantial number. Add those pickoffs to his caught-stealing numbers, and suddenly his success rate drops from 73 percent to below two-thirds. You can do this with any player, not just Robinson, but his example is illustrative.
Robinson's nemesis on pickoffs was Boston Braves pitcher Warren Spahn. The southpaw picked off Robinson four times over the years, including twice in the first game of a Sept. 6, 1948 double-header in Boston. Quite attention-getting is when those pickoffs happened: the 11th and 14th innings. Yes, Spahn started the game, and was still pitching in the 14th. Jackie's aggressiveness was costly: The Braves won in the bottom of the 14th. For a single season, the Giants' Clint Hartung was the champ, picking Jackie off three times in 1949, but never before or afterward.
From 18 pickoffs at first base and five at second, Robinson lost a total of 1.0432 WPA. He was never recorded as being picked off third. Count this against his other stealing numbers, and Robinson tumbles to just break-even stealing second and third, with his positive stealing-home numbers untouched. I could claim vindication here, but there are positives yet to count.
Pickoffs don't always go right, or even neutrally. A wild throw can give the runner the base he was hoping to swipe, or more. This was a tougher hunt, as pickoff errors are too uncommon to get their own specific search function. I looked up pitchers' throwing errors on the teams Robinson faced, checked fielding logs to see if those pitchers made errors playing the Dodgers, and looked over play-by-play on matching games. Doing this, I was able to find 10 instances where Jackie Robinson took at least one extra base on a pitcher's pickoff error, plus one in the 1952 World Series.
(I did omit searching for pickoff errors by catchers. There would have been a great deal more work for an event that occurs much less often, perhaps never where Robinson was concerned.)
Robinson really made hay when a pickoff went awry. Six of the 10 times, he took two extra bases on the error, including going second to home via Milwaukee Brave Ernie Johnson's miscue on Sept. 11, 1956. He got a bit of revenge against Hartung by going first-to-third on him in a 1949 game, though Warren Spahn never slipped that way.
This restores some of the WPA lost to pickoffs, but not nearly all. In total, it's a 0.2756 WPA gain, just over a quarter what he gave away. There were no errant pickoffs when he was on third, so all of the gain goes into the second-third column.
And again, we're not quite done. One final way a nervous pitcher can reward a potential base-stealer is with a balk. I counted eight times in the regular season, and twice in the World Series, that Robinson advanced on a balk. However, I only count five of the eight (plus the two Series events) in Robinson's favor. On the other three, the base ahead of Robinson was occupied: it wasn't the threat of Robinson (alone) running that triggered the balk.
Robinson accumulated 0.0617 WPA via the five balks: two when he was on first, two on second, and one on third. On that last one, Jackie basically stole home without even having to run. As it was in an 8-1 game in the ninth inning, though, the WPA value is a tiny 0.0003. Almost all of his balk value is in non-home situations.
So let's look at the chart again. Non-steal events are counted with the base ahead that Robinson was potentially stealing. If he's picked off first, that counts with attempts to steal second; if he's balked home, that counts with attempted steals of home.
Events No. WPA Ttl. 2nd 3rd 2nd+3rd Home Steal attempts 273 +1.8235 +0.4741 +0.5696 +1.0437 +0.7798 Pickoffs 23 -1.0432 -0.8241 -0.2191 -1.0432 0 Pickoff errors 10 +0.2756 +0.1412 +0.1344 +0.2756 0 Balks 5 +0.0617 +0.0189 +0.0425 +0.0614 +0.0003 All non-steals 38 -0.7059 -0.6640 -0.0422 -0.7062 +0.0003 Total 311 +1.1176 -0.1899 +0.5274 +0.3375 +0.7801
The steals of home finally do come out ahead, after a fair amount of manipulation. I can't consider this a true vindication of my starting hypothesis, but it's an informative result.
This does assume, though, that all the value of those ancillary plays should be counted toward Robinson's penchant for stealing. That's not quite so: We can record what a pitcher did, but not whether a runner really was planning to take the next base or just wanted a good lead. But even if only half the value of those plays was attributed to Robinson's stealing, it would still nudge his plays for home above those for the other bases combined.
The numbers do emphasize something a bit unexpected, that one could already see in the steal-only table: Robinson was doing some real damage stealing third base. He produced greater value on 38 attempts at third than the did on 204 attempts at second. He did this despite making a fair number of attempts with zero or two outs, the latter especially considered a violation of the proverbial Book and a tactical blunder. As my original article noted, Robinson didn't restrict himself to ideal tactical situations or great leverage:He went when he thought he could take the base.
The secret to his success at third is probably that he had company. In at least nine of his 38 attempts to steal third, the man on first was running with him. (It may be as many as 12: Robinson was caught three times for the third out in potential double-steal situations—twice on strikeout-throw out plays—and what the trailing runner did is not recorded. Accounts in The New York Times mention none of the plays.)
The double-steal of second and third is one of the great untapped percentage plays in baseball. Going with one out, the break-even percentage is just a little over 50 percent, and it's not too much worse with zero outs. Robinson was seven for nine for the times we know he was leading such a double-steal, and all but one came with one out.
The cumulative WPA on those nine plays is +.3339, over half his value for all attempted steals of third. Even if all three unsure plays are debited to double-steals, it still comes out to +.2242. Averaging as much as 3.8 percent of a win for a single play is great percentage baseball.
When it really counted
Now for the World Series. Robinson's stealing record in World Series play is superficially perfect: six for six on steals, plus advancing twice on balks and once on a wild pickoff throw. There are two complicating factors, though. The list here is short enough that I can show you every steal-related play Robinson made in the Fall classic. All games listed were against the New York Yankees.
Date Game H/A Inn. Base/Out Bkn-NYY Took/Play WPA 9/30/47 One A T1 1XX/1 0-0 2nd/SB +.0156 9/30/47 One A T3 1XX/2 1-0 2nd/Bk +.0101 10/2/47 Three H B1 1XX/1 0-0 2nd/SB+O -.0405 10/1/52 One H B6 1XX/2 3-1 2nd/POE1 +.0073 10/3/52 Three A T9 12X/1 3-2 2nd/DS 0 10/5/52 Five A T2 12X/0 0-0 3rd/SB +.0362 10/2/53 Three H B5 X2X/1 0-1 3rd/Bk +.0318 10/5/53 Six A T6 X2X/1 0-3 3rd/SB +.0169 9/28/55 One A T8 XX3/2 4-6 Home/SB +.0527
For the completists—which is probably a pretty good number of you—the balks were charged to Spec Shea and Vic Raschi respectively, and the pickoff error was by Allie Reynolds.
The sore thumb in that list comes in Game Three of the 1947 Series. Robinson stole second, then started for third when catcher Sherm Lollar's throw got past Phil Rizzuto.
Second baseman Snuffy Stirnweiss backed up the heave, though, and got the ball to Rizzuto, who tagged Robinson out as he was scrambling back to second. Not one for Robinson's highlight reel; not even a play most Yankees die-hards would remember, unless Scooter recounted it a time or nine in his broadcast days.
The other obvious oddity is Game Three in 1952. As the back end of a double-steal, Robinson's stolen base gets no credit in my system.
There's some irony that Robinson's two biggest base-stealing plays in the World Series didn't really affect the outcomes of the games. Brooklyn survived his off-base encounter with Rizzuto by winning a 9-8 slugfest. His fabled steal of home in Game One of the 1955 Series closed his team's gap to 6-5, but that was the final score of the game.
Taking just the steals, Robinson is credited with -.0249 WPA for stealing second, +.0531 for third, and +.0527 for the fabled steal of home against Yogi Berra in 1955. Add these to career steals, and ... no, not quite. Steals of home end up a total +0.8525 WPA, against +1.0723 for the rest. I can't drag my hypothesis across the finish line that way. All told, World Series steals produced +.0809 WPA for Robinson.
Throwing in the balks and pickoff error, the value for second in the Series rises to -.0075, and for third to +.0849; the overall Series number goes to +.1301. It's the reverse of the regular season numbers: For the Series, home comes out ahead on steals alone, but falls behind when ancillary plays are included. With so small a sample, there's no actual meaning to that, except for what he did on the basepaths and to the nerves of opposing pitchers. And to the temper of Yogi.
I was wrong: Jackie Robinson did not accumulate more value with steals of home than with all his other steals. A broad look at the record, though, shows that his stealing success came through everything but the standard, mundane steal of second base. He piled up value not only going for home, but by stealing third, often leading double-steals, and with frequent extra bases taken by provoking errors and balks.
His greatest successes came with two of the most untapped percentage plays in baseball: the steal of home, and the double-steal of third and second. These require quite modest success rates to produce win value, and Robinson easily beat those break-evens in a way he had more trouble doing with the workaday steal of second.
Perhaps this is the foundation of Robinson's reputation as a singularly disruptive base-runner. It wasn't that he might steal on you, but that he might steal anything on you, at any time. Second, third, or home; double-steal or triple-steal; first inning or ninth; six runs up or six runs down. If Robinson had a base open ahead of him—and sometimes when he didn't—you could not relax.
Those opposing pitchers did not relax, a truth visible in the pickoff numbers, as well as the pickoff error numbers. You could best him, like Warren Spahn did, but you could never ignore him, not even at the end of his career. In proof of that, I note that Robinson converted the last nine stolen base attempts he ever made.
Let that stand as a fitting conclusion to my twin studies of Jackie Robinson: from his first game to his last, you could never ignore him.
References and Resources
I have updated my original table of Robinson's attempted steals of home plate, to reflect not only the WPA data but new information about pitch counts. There was no natural place for it in the article itself, so it goes here. WPA values are figured in accordance with the rules I listed above.
Date Opp't Inn. Score Base/Out Count SB? WPA The Final 6/24/47 @PIT T5 2-2 X23/1 2-1 Y +.0107 W 4-2 7/19/47 vSLN B1 1-2 1X3/2 2-1 Y +.0795 L 7-5 7/26/47 @PIT T6 3-0 1X3/1 1-2 N -.0279 W 6-4 8/29/47 vNYN B6 5-1 XX3/2 0-0 Y +.0216 W 6-3 7/4/48 vNYN B7 4-8 X23/2 0-0 Y +.0278 W 13-12 7/21/48 (1) @CHN T9 9-3 XX3/2 1-1 N -.0003 W 9-3 7/23/48 @PIT T5 3-0 XX3/2 1-1 N -.024 W 4-3 7/25/48 (1) @PIT T8 6-5 X23/2 2-0 Y +.1032 W 7-6 8/4/48 vCHN B1 1-0 XX3/2 0-1 Y +.0664 W 5-4 8/22/48 vBSN B5 2-2 123/2 0-1 Y +.1156 L 4-3 9/3/48 (2) vNYN B1 0-0 1X3/2 1-2 N -.0485 L 6-3 9/28/48 vBSN B5 6-4 XX3/2 0-0 Y +.0618 W 9-8 (13) 5/17/49 @CHN T8 2-0 XX3/2 0-0 N -.0268 W 8-5 (11) 6/2/49 vSLN B6 3-1 XX3/2 2-0 Y +.0591 L 7-4 (14) 7/16/49 vCIN B2 0-1 123/1 0-1 Y +.0788 L 7-6 (10) 7/18/49 vCHN B6 1-0 XX3/1 1-1 Y +.0379 W 3-0 8/9/49 @PHN T5 5-0 1X3/2 0-0 Y +.0253 W 8-1 8/14/49 vBSN B5 3-1 1X3/2 1-0 N -.038 W 7-2 9/20/49 @CHC T8 4-0 XX3/2 1-1 Y +.0183 W 5-0 6/19/50 vNYN B6 8-4 XX3/2 2-1 N -.0124 W 8-5 (11) 7/2/50 (1) @PHN T4 1-4 XX3/2 2-1 Y +.055 L 6-4 5/2/51 vPIT B2 0-1 1X3/2 3-1 N -.0896 L 4-3 9/26/51 @BSN T8 13-3 1X3/1 ?-? Y +.0001 W 15-5 5/18/52 vCHC B4 2-1 123/2 3-1 Y +.0782 W 7-2 7/16/53 vSLN B7 7-2 XX3/2 0-1 N -.0054 W 9-2 4/23/54 @PIT T6 2-1 123/2 2-0 Y +.1015 W 6-5 (13) 6/17/54 vMIL B2 0-1 XX3/2 0-1 N -.0404 L 6-4 8/28/55 vSLN B2 0-0 XX3/2 3-1 N -.0391 W 6-1 8/29/55 vSLN B6 4-1 123/2 3-1 Y +.0399 W 10-4 9/28/55 @NYA T8 4-6 XX3/2 1-0 Y +.0527 L 6-5 4/25/56 @NYN T2 0-0 XX3/2 1-0 Y +.0822 W 7-2 5/13/56 vNYN B8 6-4 123/1 1-1 N -.0217 W 6-4
Shane Tourtellotte is a long-time, occasionally-nominated science fiction writer, currently living in Asheville, North Carolina. He will tell you all about the baseball novel he’s shopping if you give him an inch.