On Un-Earned Runs
Early in my career, I questioned whether the concept of an "error" was useful (as opposed to useless) and whether it was legitimate (as opposed to arbitrary.) To say that I "questioned" the concept is being unnecessarily kind to my younger self. I attacked the concept. The concept of an error, I argued, made an arbitrary distinction between events which were otherwise much the same. Two balls are hit to center field. The center fielder doesn’t field either one. In both cases the batter winds up at second base—but one of these is recorded as an error, and the other is recorded as a double. Why is that?
It is because a third party, not connected to the game in any necessary manner, has made a judgment about what SHOULD HAVE happened in one case—the center fielder should have caught that ball—and has entered his judgment into the record. This is not what records are supposed to be. The statistical record is supposed to be an account of what has happened, not a record of subjective judgments about what SHOULD have happened if everybody had played better.
As far as I know I was the first person to make that argument—maybe I was or wasn’t, who knows—but in any case it has entered the marketplace of ideas, and a substantial number of people now see the issue that way. But in retrospect. . . .welllll.
The principle that the record is supposed to be an objective record of what has happened, without the extraneous input of value judgments, still seems to me to be a valid principle. It is, however, not the ONLY valid principle that instructs the business of record-keeping. Another valid principle is that we want the record books to include as much information as we can get. Another valid principle is that unlike events should not be recorded as if they were the same.
Suppose that there are three ground balls to the shortstop side of third base, hit on the same line with the same velocity and the same number of hops, etc. In one case, the third baseman picks up the ball, throws to first, and the out is recorded. In the second case, the third baseman picks up the ball and throws it to where the first baseman could never possibly catch it. In the third case, the third baseman is in a shift, and there is no possible way he could make a play on the ball.
I trust that you see my points: that the elimination of the concept of the error (a) eliminates information from the record book which would be useful to us if we knew it, and (b) causes unlike events to be recorded as if they were the same. Also, in making that argument, I argued that it was generally and usually difficult to say whether something was a hit or an error. I don’t know, in retrospect, that that was absolutely true. It is often true, a point that has been brilliantly illustrated by Brian Kenny with his "Hit, or Error" videos, but I also think that I over-stated the point years ago.
I am not saying that the younger Bill James was entirely wrong, or that you should not pay attention to him. I was making a useful point, I think. I am saying that you shouldn’t take the point too literally.
This is part of what is going on with the Jack Kralick/1961 problem, one of the largest issues. If you obliterate the reliance on errors then un-earned runs are the same as earned runs. The design of rWAR takes that literally, and treats earned runs the same as un-earned runs. Kralick allowed a very low number of un-earned runs. We are used to evaluating pitchers in large part based on the Earned Run Average, but Kralick looks much, much better when the un-earned runs are treated the same as the earned runs. There are 40 American League pitchers who qualified for the ERA championship in 1961. In ERA, Kralick ranks 18th among the 40. He is near the center of the chart. If you rank them by TOTAL runs allowed per nine innings, however, Kralick leapfrogs Jim Bunning, Bennie Daniels, Don Schwall, Frank Lary, Rollie Sheldon, Bill Monboquette, Camilo Pascual and Jim Archer, and thus moves up to 10th among the 40 pitchers. This is one of the largest reasons for the Kralick anomaly.
There are six reasons for the Kralick anomaly:
1) That there is no pitcher in 1961 who is clearly better than the others, and there is no little group of three or four who can be easily distinguished from the others. This creates a cluster of pitchers who rate as being close to one another. When there is a cluster of players who are about the same, that makes small adjustments relatively larger.
2) The fact that the rWAR system treats unearned runs the same as earned runs, which is not natural to us.
3) The fact that the system makes park adjustments, which most of us still do not do most of the time in making a casual assessment of a player. The intensity of emotion supporting Jacob deGrom in 2018 occurred in substantial part because people don’t make park adjustments in their heads, and thus overrated the extent to which deGrom was dominating the league. (I agree that deGrom probably was the best pitcher in the league, but when you factor in the park adjustment, it’s actually a razor-thin margin, rather than a wide margin. But most people don’t know that, because most people don’t make park adjustments in their heads as a routine matter.)
Kralick benefits from the park adjustment, and this is absolutely legitimate. But whereas Kralick worked in a hitter’s park, all nine of those who were ahead of him worked in pitcher’s parks. Eight of the nine worked in parks with Park Factors in the 80s. Thus, when you park-adjust their RUNS allowed per nine innings, Kralick vaults from 10th in the league to 4th.
4) Park Factors can be measured either as one-year or multi-year elements, neither of which exactly works, and either of which causes problems. But rWAR, I believe, is using a single-year evaluation.
Single-year evaluations of Park Factors can be unreliable because the number of games isn’t large enough for the data to stabilize. The measured Minnesota Park Factor in 1961—118—is almost certainly much larger than the actual effects of the park. The Park Factor for that park, if measured in a five-year window (1961 to 1965) would be 106, not 118. Also, the Angels in 1962 moved from a hitter’s haven to a pitching-dominated park, which should have pushed the relative Park Factor of every other team in the league UP, so even the 106 park factor that you would get for Minnesota from a five-year study may be too high. Kralick is almost certainly getting the benefit of an adjustment that is much larger than it actually should be.
5) We are, of course, not paying attention to his won-lost record, which most of us still look at and still place some weight on in evaluating pitchers.
6) The system is assuming that Kralick suffers from bad defense, and that this is costing him .09 runs per nine innings, or two and a half runs on the season (the category shown as RA9def in the Baseball Reference charts.) When you make THAT adjustment, because the differences among these pitchers are small, Kralick passes two of the three pitchers still ahead of him on the Runs Allowed per 9 innings Park Adjusted chart, moving him into second place. Since Kralick pitched far more innings than the other pitchers with comparable RA9 park-adjusted, he shows as the most valuable pitcher in the league.
But the RA9def adjustment is highly suspect. I am not saying that pitchers don’t benefit or suffer from good or bad defense behind them, or that we should not attempt to adjust for that. I am not saying that the Baseball Reference adjustments are always wrong or generally wrong, or even that two and a half runs is not a reasonable estimate in the case of Jack Kralick. What I am saying is that it is EXTREMELY difficult to know what the actual impact of the defense on the pitcher has been, even now when we have tons of data to work with, and it is highly speculative in regard to any one particular pitcher when working with the much more limited data of 1961.
What I said before was that there is a very narrow pathway that leads to the conclusion that Jack Kralick was the best pitcher in the American League in 1961—or the conclusion that Don Cardwell was the best pitcher in the National League and the best pitcher in baseball. There is ONE pathway through the numbers that leads to that conclusion. To reach that point, you have to believe:
1) That won-lost records shouldn’t count,
2) That earned runs should be counted the same as un-earned runs,
3) That runs allowed averages must be park-adjusted,
4) That the Minnesota Park Effect in 1961 was actually +18%, rather than that it merely happened that there were some high-scoring games in Minnesota in 1961, and
5) That despite the fact that he was charged with only four un-earned runs, Kralick suffered from very poor fielding support.
My point is not that any of those assumptions is wrong. It is that all of them are questionable at one level or another, and that all five of them are necessary to reach the conclusion that Kralick is the best pitcher in the league. The one pathway that happens to lead to that conclusion is the one pathway chosen by the rWAR system.
And that is just what happens when you design rating systems. You HAVE to make choices that are not absolutely 100% right 100% of the time. You choose the set of assumptions that you think are 51% right in one case and 55% right in another case and 80% right in one case and 99% right in another case. And sometimes the sum total of those assumptions is great, and sometimes it is good, and one time in a hundred the sum total of those assumptions is going to lead you out in the forest and drop you off on Jack Kralick’s doorstep.
Kralick and Curt Simmons
In the National League in 1961 there was a pitcher, Curt Simmons, who is in many respects very similar to Jack Kralick, and in other respects exactly the opposite of him.
Simmons had a career winning percentage of .513 and a career ERA of 3.54. Kralick had a career winning percentage of .521 and a career ERA of 3.55. Simmons, one of the first Bonus Babies, had had a long career. He had come to the majors at the age of 18 in 1947. He had perhaps the best fastball in baseball as a young pitcher, but after a few good seasons he had suffered for years with not very good Philadelphia Philly teams, had worn out, gone to the minors to get re-started, and had wound up in St. Louis. Kralick was a young pitcher with no past to speak of, but in 1961 he was exactly the same thing that Simmons was: a left-hander with an average/below average fastball who could win a little more than half of his games because he got ground balls and didn’t walk people and he knew how to pitch.
And, as Kralick is seen by one pathway through the numbers as the best pitcher in his league, so too is Simmons. The National League ERA was almost exactly the same as the American League ERA, 4.03 and 4.02, and Simmons also was working in a hitter’s park. The Busch Stadium Park Factor in 1961 was a whopping 130, the highest in baseball although, as was true with Minnesota, it wasn’t actually that much of a hitter’s park, you just don’t get a true read with one season’s data.
Simmons had a 3.12 ERA in 30 games, 196 innings. With a 4.03 league ERA and a Park Factor of 130, a 3.12 ERA is really good. His ERA+ was 141, which was the best in the National League. Simmons was at that level quite a bit; he had two seasons when his ERA+ was 144. His ERA+ was third in the league in 1960 and 1963, second in the league in 1954, and among the top ten in the league eight times.
The 1961 St. Louis Cardinals, for whom Simmons labored, had had trouble at shortstop since Marty Marion got hurt more than ten years earlier. The situation had gone steadily downhill, as it will if you don’t decide what you want to do at a position and commit yourself to it, and by 1961 it had reached the point of being laughable. The Cardinals kept switching between shortstops all year. For the season, Alex Grammas played 351 innings at shortstop for the Cardinals; he is listed in many sources as the Cardinals regular shortstop because he played in more games at shortstop than anyone else, but half of those games were late-inning defensive replacements, and when he started the game, he often came out for a pinch hitter. He was 35 years old, all year, and he had been a pretty good defensive shortstop, but you know, he was 35 years old, and had spent several years as a utility infielder. He had been traded from Cincinnati to St. Louis and then back to Cincinnati and then back to St. Louis. He got 351 innings at shortstop, Bob Lillis had 384, Daryl Spencer had 317, Jerry Buchek got 230, and Julio Gotay got 88.
They were all terrible; they were all second basemen trying to play shortstop except Gotay, who was a legitimate shortstop except ridiculously error prone. Among the five of them they made 53 errors. Gotay fielded .804 at shortstop, Buchek .912, Lillis .928, Spencer .956 and Grammas .960. The 53 errors was 19 more than any other team in the National League—at shortstop or any other position. It’s a LOT of errors, and also, Cardinal second basemen made 27 errors, which was second in the league at the position.
They particularly liked to make errors when Simmons was on the mound. It’s not surprising; left-handed ground ball pitcher, you’re going to get GB6 when he is on the mound. Simmons was a really good hitter; not that that is relevant but it is part of the story. He had hit over .200 ever year in 1956, 1957, 1958, 1959 and 1960, and in 1960 he hit .303 in 77 plate appearances with 8 walks giving him a .378 on base percentage. He was probably the best hitting pitcher in baseball in 1961, I don’t know.
In San Francisco on May 31, Simmons was leading 2-0 going into the bottom of the sixth. Ground ball to short to lead off the inning, shortstop (Bob Lillis) made an error to get the inning rolling. He got one out but then walked Joey Amalfitano, which meant that he had to pitch to Willie Mays with two men on. Mays hit a double. There were two un-earned runs in the inning, and Simmons wound up losing the game 3-2 although he pitched pretty well.
My point is, it’s not ALL his fault, but it isn’t all the shortstop’s fault, either. When the shortstop makes an error leading off the inning with Mays and Cepeda due to bat, that’s not helpful, but when you walk Joey Amalfitano to pitch to Willie Mays, that’s not the shortstop’s fault.
His next start was June 5 against the Cubs. Simmons was leading 4-2 heading into the 7th. He gave up singles to the first two hitters, both very weak hitters, and then the ball was tapped in front of the plate, probably a bunt. The catcher threw to third base and threw wildly, E-2. Simmons came out of the game, still leading 4-2 but with the bases loaded and nobody out. The bullpen allowed all three runs to score, all of them un-earned, although the Cardinals rallied to win the game anyway.
Again, you can’t really say that Simmons is responsible for those three runs, one of which reached on an error and none of which scored while he was on the mound, but you can’t say that he wasn’t responsible at all, either. He was SOMEWHAT responsible.
Simmons’ next start, June 10 against Cincinnati. Gene Freese hits a double to center with a runner on first. That run is earned, but Curt Flood throws wildly to home plate, allowing Freese to take third on an E-8, and he will score on a sac fly, so that’s three straight starts with unearned run(s).
His next start, June 17 at Pittsburgh. In the first inning Ken Boyer botches a ground ball to put Dick Groat on first with Roberto Clemente coming up. Clemente hit .351 that year, with 23 homers, but Groat attempts to steal second and is out, so that one is no harm/no foul. Simmons gives up a couple of runs early but hits a leadoff double in the third inning, setting up a 2-run inning to give the Cardinals a 3-2 lead.
Same game, bottom of the third, one out, Bill Virdon singles. Dick Groat is up. Groat is very slow and hits a ton of ground balls, and he answers with a double play ball to the shortstop, Julio Gotay, but Gotay rushes the pickup and kicks the ball away, E-6. Runners on first and third, and that brings Clemente to the plate again. Clemente grounds out, but the run scores, un-earned run, tie game.
Still the same game, bottom of the sixth, still 3-3. With one out the pitcher, Bob Friend, hits a ground ball to Gotay. Gotay drops it again, another E-6. Virdon singles and Dick Groat hits a fly ball; un-earned run, 4-3 game.
Still the same game, eighth inning, game still 4-3. Simmons gives up leadoff singles to Groat and Clemente, and leaves the game. He is out of the game but still responsible for the runners on base. The reliever strikes out Dick Stuart, and Don Hoak hits a ground ball to Gotay, should be an inning-ending double play, but Gotay boots it again, his third error of the game, all three of them leading to un-earned runs.
Simmons has one start without an unearned run, then on June 27 faces the Braves. Frank Bolling hits a ground ball to short to start the sixth inning, but the shortstop (Alex Grammas) boots it, giving Simmons the opportunity to face Eddie Mathews with a man on base. Mathews homers, so there’s two runs, one of them un-earned.
Simmons next start is the first game of a Fourth of July double header against the Phillies. First batter of the game, Tony Taylor, hits a single, but the second hitter hits a double play grounder to the shortstop, Jerry Buchek. Buchek boots it, setting up a three-run inning, all three runs un-earned because of Buchek’s boot.
Simmons, however, draws a leadoff walk in the fifth inning, scores on Bill White’s homer, the Cardinals lead 4-3. In the sixth inning Simmons bats with two out and a man in scoring position, hits a run-scoring single to make it 5-3. By the top of the eighth the Cardinals have a big lead.
Eighth inning, two on, two out. Ground ball to short. Alex Grammas is playing short by now, as a defensive replacement for Buchek, but Grammas commits another error, loading the bases with two out. Simmons leaves the game, but all three runs eventually score, all three un-earned. Simmons is credited with the win despite giving up six un-earned runs in the game. Simmons has given up 16 un-earned runs in seven starts.
On the season, Simmons was charged with 23 un-earned runs, an exceptionally high number for 196 innings. Kralick was charged with 4 un-earned runs, Simmons 23—despite which Baseball Reference, making their absolute best and serious efforts to assign credit and responsibility, says that Kralick’s defensive support was poor—negative .09 runs per nine innings—while Simmons’ defensive support was exceptionally good, positive .30 runs per nine innings, or positive six and a half runs over the course of the season.
Of course, what Baseball Reference means by "defensive support" is not just errors—is not, and should not be. There is a great deal more to defense than just errors. The Cardinals, despite playing in what appears to be the best hitter’s park in baseball, despite a below-average number of strikeouts by pitchers and a far worse-than-average number of walks, have led the National League in ERA. You have to explain it somehow. Baseball Reference credits the Cardinal defense with saving about 45 runs. But is that true?
Well. . ..it’s a bit of a reach. The Cardinals committed 166 errors, second-most in the league. They have three regulars who are outstanding defensive players—first baseman Bill White, third baseman Ken Boyer, and Center Fielder Curt Flood. But they have no regular catcher; their catchers are not as defensively troublesome as their shortstops, but they’re definitely not good. They have a 40-year-old in left field, Stan Musial, and a first baseman in right field, Joe Cunningham. Their second baseman, Julian Javier, is extremely quick but a rookie, and error-prone. It is VERY difficult to see how that team gets to be 45 runs better than average in the field.
Their top four starters, on the other hand, are Bob Gibson (251 major league wins), Larry Jackson (194 wins), Curt Simmons (193 wins) and Ray Sadecki (135 wins in an 18-year career.) Those guys were all in the majors a long, long time for a reason. The 130 Park Factor isn’t actually right, and the 45 runs for the defense isn’t actually right. It is NOT a convincing analysis.
But the real point I am trying to get to is this. Baseball Reference WAR assumes that the pitcher is AS MUCH responsible for an un-earned run as he is for an earned run, and thus makes no distinction between the two. Using that assumption, they reach the conclusion that Kralick is the most valuable pitcher in the American League.
ERA+ assumes that the pitcher is not at all responsible for un-earned runs, 100% off the hook. Using that assumption, ERA+ concludes that Curt Simmons was the most effective pitcher in the National League.
If you evaluate these pitchers based on ERA, Simmons is half a run ahead, 3.12 to 3.61. If you evaluate them by TOTAL runs allowed per nine innings, Kralick is almost as far ahead, 3.76 to 4.18.
But if you assume that the pitcher should be held 50% responsible for the un-earned run, not 100%, then Simmons and Kralick are almost exactly the same—3.65 (Simmons) and 3.68 (Kralick). So what is the best answer?
It’s obvious, isn’t it? The best answer is that the pitcher does not bear the SAME responsibility for an unearned run as for an earned run, but he bears SOME responsibility for the un-earned run, which we will call 50% until we have some better answer.
And that’s what the Game Score system does. It charges the pitcher twice as much for an earned run as for an un-earned run. That was my point in writing this. I am arguing for my system.