Standing Sabermetrics on its head
To see what falls out of its pockets
Alright, I’m assuming that we probably can’t do this this year (2008), because we won’t have time to organize it before the season starts, but for next year.
What if we converted the MVP argument from an after-the-fact review into a season-long chase? You see what I’m getting at here? We go to games, we keep stats, we follow the players. When the season is over, or nearly over, we sit down, look at what we have, and try to figure out who the MVP must have been.
What if, instead, we keep track of value as it happens?
How would we do that?
By a system of organized observation.
Aren’t statistics just a system of organized observation?
Statistics are a system of organized observation of events. This would be a system of organized observation of value.
Isn’t the value in the events?
Events of entirely unlike value are recorded in the statistics as equals. Suppose that the game is 12-1 in the eighth inning, nobody on, a ground ball is hit to the second baseman and he flips to first. This is entered in the statistics as an assist by the second baseman. Suppose that the game is 4-3 in the ninth, men on second and third, the batter hits a rocket toward right but the second baseman knocks it down, scrambles to his feet and gets the out at first. This also is entered in the statistics as an assist by the second baseman. The statistical observation—the “event”—is the same. But the value is different by a margin of a thousand to one.
So you’re keeping track of. . .key defensive plays?
No, I’m proposing that we keep track of value—observed value. Directly observed value.
How do we keep track of directly observed value?
I thought you’d never ask. OK, here’s what we do. We have somebody designated to cover every game. . .hopefully somebody at the game, but in any case somebody very carefully watching the game.
Like an official scorer?
Sort of. Very much like an official scorer, except that he’s not official and he’s not exactly a scorer.
He’s an unofficial non-scorer.
He’s an unofficial evaluator, designated by us. . .designated by the DOV project. We’ll call him a scorer, although he’s not really keeping score.
And his job is. . .?
His job is to make an on-the-scene evaluation of each player’s contribution to victory, by an organized system of rules designed to ensure fair treatment for every game, every player, and every type of contribution to victory. For each game—for each win—the team that wins has 10 “value points” to be distributed among the players who participated in the game. The scorer’s job is to decide, consistent with the policies laid out by the organization, how many of those points are credited to each player.
Could one player get all 10 points?
In theory. In the National League, if you had a 1-0 game that was won by a solo home run by the pitcher, perhaps that player—the pitcher—could be given all 10 points. Even then it would be unusual to exclude the rest of the team, because there would usually be fielding plays that contributed to the shutout, and I would suspect the pitcher would get 8 or 9 points, rather than 10.
In the American League, I suppose that it is theoretically possible that one player could get all 10 points, in exactly the right circumstances. Suppose that a game was won 10-9, with one player driving in all 10 runs, and suppose that that player also made a series of brilliant defensive plays to keep the other team from scoring even more runs, sort of like that famous game that Ryne Sandberg had in ’84, when Herzog said that Ryne Sandberg was the greatest player he had ever seen. If a player had a game like that, only more so, I suppose that that player might get all 10 points awarded for the victory. But it would certainly be an extremely unusual event.
What if a pitcher pitches a no-hitter?
A no-hitter is a glamour event, a thrilling event. This isn’t a measure of glamorous contributions; it’s a measure of win contributions.
So a pitcher might pitch a no-hitter, and get no points?
Well, I can’t imagine how that could happen, given the Fay Vincent definition of a no-hitter, which requires that you win the game. Who was the guy. . .Ken Johnson. .who pitched a no-hitter but lost? If you don’t win the game, there are no points for anybody. But using the Fay Vincent definition of a no-hitter, I can imagine how, at the extreme, in the same sort of I-suppose-its-possible world in which a position player might get all ten points, a pitcher pitching a no-hitter might get as few as 2 points.
Suppose, for example, that the game is 12-0. Then the pitching and defense has been outstanding, but the offense has been outstanding, too, and they have really contributed as much to the win as the pitching has, so you’ve got 5 points to pay the hitters, 5 points for pitching and defense. Suppose that the pitcher has struck out 1 batter and walked 6, but survived it in part because of four double plays. Then you’ve got to credit the shortstop and the second baseman for keeping the other team off the scoreboard. Suppose that the center fielder has made three shoestring catches and pulled a ball back into the park. You’ve got to give him credit for that. So in that situation, a pitcher could pitch a no-hitter and get as few as 2 Win Points for it.
How many points does the starting pitcher usually get? What’s the normal split between the pitcher and the hitters?
Three ways—pitchers, hitters and fielders. The goal would be that on average, over time, half the points would go to the hitters (and baserunners), and half would go to the pitchers and fielders. But it depends on the game. If you have a game that is won 1-0, obviously the pitching is mostly responsible for that. The offense hasn’t done much; you got a win because the pitcher was great. Or if the game is won 14-11, there might not be any credit given to the pitcher(s), because what has really happened is that the offense has overcome the bad performance of the pitchers, not that the pitchers have really contributed anything to the win.
My notion is that there would be a scale, something like this:
| |
0 |
1 |
2 |
3 |
4 |
5 |
6 |
7 |
8 |
9 |
10 |
| 0 |
|
1 |
1 |
2 |
2 |
3 |
3 |
4 |
4 |
5 |
5 |
| 1 |
|
|
2 |
2 |
3 |
3 |
4 |
4 |
5 |
5 |
6 |
| 2 |
|
|
|
3 |
3 |
4 |
4 |
5 |
5 |
6 |
6 |
| 3 |
|
|
|
|
4 |
4 |
5 |
5 |
6 |
6 |
7 |
| 4 |
|
|
|
|
|
5 |
5 |
6 |
6 |
7 |
7 |
| 5 |
|
|
|
|
|
|
6 |
6 |
7 |
7 |
8 |
| 6 |
|
|
|
|
|
|
|
7 |
7 |
8 |
8 |
| 7 |
|
|
|
|
|
|
|
|
8 |
8 |
9 |
| 8 |
|
|
|
|
|
|
|
|
|
9 |
9 |
| 9 |
|
|
|
|
|
|
|
|
|
|
10 |
| 10 |
|
|
|
|
|
|
|
|
|
|
|
Winner’s score along the top, loser’s score along the side, and the number given is the number of points assigned to the offense in the game, based on that score. If the score of the game is 1-0, there is one point for the offense, 9 for the pitching and defense, since it is obviously the pitching and defense which is primarily responsible for the win. If the score of the game is 10-0, that would be five points for the offense, five for the pitching and defense, since both were outstanding. If the score of the game was 4-2, that would be 3 points for the offense, 7 for the defense, since the offense has just been so-so, but the game was won due to good pitching and defense.
But this is a guideline, not an absolute rule. Suppose that the score of the game is 8-5. The chart above says that in an 8-5 game, you give 7 points to the offense, three to the pitching and defense. But suppose that you’re playing in Texas and the wind is howling to right and the umpire won’t call anything a strike. Then an 8-5 game might be considered a draw from the standpoint of pitching versus hitting, and you might credit 5 points to the pitcher.
Or suppose that the game is 7-1, the manager brings in the last guy in the bullpen, the guy with a 6.42 ERA and a three-year contract, to pitch the ninth. Let’s say he gives up a three-run homer with two out in the ninth, 7-4 final. The chart says that a 7-4 game is 6 points to the offense, 4 to the pitchers, but it’s really a 7-1 game with 3 runs on the scoreboard that didn’t mean anything. Then the evaluator might justifiably give 6 points to pitching and defense—obviously, none of them to the guy who pitched the ninth—and 4 to the offense. There’s an element of discretion involved, as long as the values are reasonable.
Can the evaluator split points, give one player 5 ½ points and another 2 ½?
No.
Why not?
Because at some point you have to decide. If you can give halves you can give quarters, if you can give quarters you can give tenths, if you can give tenths you can give hundredths. At some point you have to face the issue. A ten-point system is senstive and articulate enough. If you go beyond that you’ll very soon be going beyond the point at which human judgment can accurately perceive meaningful distinctions.
How can it be an accurate system if the scorer can do whatever he wants?
Well, to begin with, the scorer can’t do whatever he wants. There is a system, there is a process, and there are rules. Our goal will be to make the results as close as possible to objectively true.
My view is this. All systems have flaws. The system of Win Shares and Loss Shares, for example, is flawed in that it lacks sensitivity to the timeliness of a player’s production, and is flawed in that it’s evaluations of a player’s defense are on some occasions at odds with the observations of people on the scene.
Allowing space for the bias of the observer certainly introduces a risk of error, but weigh that honestly against the errors in other kinds of analysis. First, the individual scorer has very limited space for discretion within the guidelines. Second, only some small portion of that discretion is error. Most of it is insight. We go to the game, we see fantastic plays, we see baserunning blunders, we see pressure pitching and timely hitting. These perceptions are not essentially bias. They merely contain some small portion of bias.
If the system is 90% objective and 10% subjective, and if the subjective input is 90% insight and 10% bias, that’s an error rate of 1%. We can live with a 1% error input in order to get the advantage of a system that knows who got the big hits, who made the bonehead baserunning moves and who came through with the critical defensive plays, because the traditional sabermetric analysis is essentially blind to those contributions.
Isn’t this really just a more complicated system of wins and losses, like crediting the pitcher with the win or the loss?
In a manner of speaking it is, yes.
And hasn’t it always been the position of sabermetrics that won-lost records were a highly unreliable guage of how well a pitcher has actually pitched?
Well, not trying to speak for sabermetrics, but. . .no, that’s never exactly been my position.
I would argue that, in the universe of baseball statistics, it is difficult to find any more successful competitor than pitcher’s wins and losses. They’re universally understood and universally available. A pitcher’s won-lost record, in the traditional world of baseball statistics, is his defining measure. . .he’s a 20-game winner, or he’s a 14-game winner. They’re in the newspaper every day. Even serious statistical analysts, when not focused on the issue, will define a pitcher by his won-lost record.
The won-lost record is imperfect, certainly, and we’ve spent a lot of time attacking it, certainly. But we attack won-lost records essentially for two reasons. First, people treat won-lost records as if they were the be-all and end-all for starting pitchers, which is wrong. You do get the occasional case in which a pitcher has pitched well but finished 11-14, while another pitcher has pitched poorly but finished 17-8. Some baseball traditionalists try or tried to pretend that this never happened, and we needed to pull the rug out from under that practice, which has largely been accomplished.
And second, won-lost records are faulty because they didn’t evolve as the game evolved. Well, that’s not 100% true. . .Saves, Blown Saves and Holds are direct descendants of Wins and Losses, an evolution of them. But whereas batting average evolved into slugging percentage, on-base percentage, then into OPS and a thousand other variations, the rules which govern wins and losses have never changed although the game has changed, and these rules are tremendously flawed in the modern game.
What I am suggesting here is not an extension of the reactionary 19th-century statistic pitcher wins, but rather a modern variation of the concept, recognizing that the subject is much more complicated than can be accommodated by simply assigning the win to some pitcher.
Is there really room in the system for fielding?
If you’re talking about measuring fielding skill well enough to say whether Grady Sizemore is a better defensive center fielder or Coco Crisp. . .no, there probably is not enough space in the system to make reliable estimates on that level. But if you’re talking about defensive contributions becoming more significant than they are in an MVP discussion, there is plenty of room for that.
In a typical victory you’re going to have 5 points for pitching and defense, so you’re probably going to have 1 or 2 points for fielding. What will happen a lot, based on my experience in evaluating Red Sox games in this way, will be that multiple players have claims on limited points, and key defensive plays play a huge role in resolving those. Let’s say that there’s a sequence that goes single, single, home run, and you win the game 3-2. You’ve got three points to give to the offense, and you don’t really want to give them all three to the guy who hit the home run, because the three runs are not all his doing. It’s only a three-run homer because the other two guys got on base ahead of him. On the other hand, you don’t want to give each of them one point, because a single is not the equal of a three-run homer. If you give two points to the guy who hit the bomb, there’s one point for the other two guys.
Very often one of those two guys will have made some key defensive play somewhere in the game, so he becomes the guy who gets the point. It’s kind of a half-point for offense and a half-point for fielding. But those points will add up so that, over the course of the year, the better fielder will have a real advantage.
Does the evaluator have to explain his report?
He does, yes. . .not at great length, but something like this:
- August 10, 2006
- Cubs vs. Milwaukee at Miller Park
- Milwaukee 8, Chicago 6
- Milwaukee 7 points for offense, 3 for pitching and defense as per chart.
- 2 points to Prince Fielder (walked 3 times and HBP, scored 3 runs)
- 1 point to Kevin Mench (two-run single in first)
- 1 point to Gabe Gross (walked and scored, drove in run with sac fly)
- 2 points to Graffanino (HBP in 4-run first, fly advanced runner in 3rd, 2-run single in 4th, also turned key double play in 9th)
- 1 point to David Bell (Sac fly in 1st, key single in 2-run third)
- 1 point to Mike Rivera (RBI single in third, leadoff single in 5th, threw out runner stealing in 8th)
- 2 points to Doug Davis (Pitched first 7 innings, allowing 4 runs.)
Graffanino and Rivera have split one point for their defensive play, which, with the 2 points to Davis makes 3 for pitching and defense. Francisco Cordero could have been given a point for pitching a scoreless ninth inning and earning a save, but Cordero struck out no one, walked one and got the help of a 4-6-3 DP, so I felt that his contributions were less significant than those of Rivera and Graffanino.
How would anyone know what the accounting was?
We could append it to the box scores in BJOL, or include it in BJOL as a separate report. One of the great things about having our own box scores is that we can put anything into them that we think is helpful for understanding the game. I’ll score some Red Sox games and post those as a starting point.
What if the scorer awards points in a way that is just totally unreasonable? What if the scorer decides that Francisco Cordero should get 4 points for pitching out of trouble in the ninth inning, for example?
There would have to be some sort of appeal process. Speaking of this as a mature system, we might have 15 or 20 people in the Milwaukee “cell” who were interested in getting accurate summaries for the Milwaukee games. If two or three of those people felt that the values awarded for the game were unreasonable, they could appeal the decision, and the system administrator would appoint somebody who was not in the Milwaukee cell to review the evidence and decide whether the award was reasonable.
In theory, if the system was healthy enough, we could get multiple people filing values for every game, and a computer program that looked for the consensus of those filings.
Four points for Cordero for a one-inning 8-6 save would obviously be unreasonable because the system calls for 3 points as the normal ration for pitchers and fielders in an 8-6 victory. 3 points would be unreasonable; 2 points would be unreasonable. But 1 point would be reasonable, and the values given would probably stand whether Cordero was given one point or whether he wasn’t.
Couldn’t that make the whole thing politicized and contentious?
Well, yes, but the system does not belong to that evaluator. The system belongs to everybody who is working it. We couldn’t let that evaluator pollute the data by not staying within the guidelines.
OK, but in the big picture. . .why do you want to do this anyway? Why not stick with the traditional sabermetric analysis?
The traditional sabermetric analysis of the MVP debate makes a very long circuit toward the truth. What I am trying to do here is to organize an effort to cut toward the truth by a more direct route.
We can figure out, at the end of the season, who has created the most runs, who has made the most outs, what the park factors were, what the defensive contributions were, and based on those things we can make a probabilistic estimate of value. We can say that this player created 120 runs, that player created 110 runs, therefore this player was more valuable than that one, probably. But what if he wasn’t really more valuable? What if the guy who created 110 runs had had more hits when it counted? Would we know that?
What if there was a player who created 70 runs, but who made massive contributions to his team’s success by his glove work. Would we know that?
I think of a couple of examples. One was 1973, when Dick Green hit .063 in the World Series—and may well have been the A’s most valuable player for the series, because of a series of brilliant defensive plays. I always think about that. If a player had that kind of value, over the course of a season, would we know it?
And, on the other hand, David Ortiz in 2005. In 2005 the Yankees and the Red Sox won 95 games each, both teams finishing 95-67. David Ortiz hit .300 with 47 homers, 148 RBI. A-Rod hit .321 with 48 homers, 124 RBI, and won the MVP award.
By a tradional sabermetric analysis, A-Rod appears to be more valuable. A-Rod created as many runs as Ortiz did, and Ortiz of course was a DH, while Rodriguez played third base, albeit not very well. The team performance was even, A-Rod was a better baserunner or at least a faster baserunner. …Ortiz is actually a very good baserunner for a guy who runs like a snow plow. But by a conventional sabermetric analysis, A-Rod has to win.
The only thing is. . ..what if it’s not true? I watched almost every inning of every game that Ortiz played that year. I’m sure I saw all or almost all of the 19 contests between the Yankees and the Red Sox. I could never convince myself that A-Rod was more valuable that season than Ortiz—or even that he might have been. Time and time and time again that season, Ortiz got the game-breaking hits that separated the Red Sox from defeat. A-Rod, at season’s end, was being booed by Yankee fans. I don’t believe that A-Rod was more valuable. I think the well-established sabermetric analysis is just wrong.
I have been trying to pick my way through MVP debates with careful logic for 30 years, and obviously that analysis is not the same now as it would have been in 1978, when I wrote the Guidry/Rice article. Over the years I have learned to to measure more things, to consider more things, and I have realized that some of the ways that I structured the analysis, back in the 1970s, were just not well thought out.
When A-Rod won the MVP Award I thought, “well, OK, back to the drawing board one more time. We have to start measuring who got the big hits, who made the plays that really counted.” How do we do that?
We could do it, of course, by some sort of situation-value analysis. A team’s chance of winning is .342 before this at bat, .421 after the at bat. . .total up the changes in the value of the state, and you have the value of the player.
That works real well in theory, but practically not at all in fact. To deal with defense, in that scheme, is either
- impossible, or
- extremely complicated and not very successful.
A fielder goes to the wall and pulls a grand slam home run back in the park, the system is going to want to treat that the same as a routine fly to center. You can maneuver around that by treating the fly ball as two events—the ball in play before it is fielded and the state of the game after the play is made—and good luck with that, but nobody has made it work yet.
The essential problem with that model is that it requires that we attribute to individuals the complex interactions of players. Suppose that a runner is on first, two out, and the batter grounds a single to right, only the batter from first heads for third and is thrown out. A simple analysis of the change in the status of the game will charge the batter with a negative contribution, since the outcome of the play is negative for the batter’s team.
A more sophisticated analysis will distinguish between the positive contribution of the batter—hitting a single—and the negative contribution of the baserunner, who bites off more than he can shoe. But even then, what do you do with the fielder? How much of the responsibility for the fact that the runner is out going first to third do you give to the right fielder, who makes the throw, and how much do you give to the baserunner, who thought he could make it but couldn’t?
Resolving that issue—arbitrarily—there is still the issue of the third baseman making the tag. That’s a tough play, catching a 140-foot throw and applying a tag with a man sliding at you. That’s not an automatic. How do you account for that?
What if the batter is out because he has been successfully decoyed by the second baseman? How do you give credit to the second baseman?
What if the throw to third is there in time, but skips away from the third baseman, allowing the run to score? How do you account for that?
What if the throw to third gets away but there is no advance because the pitcher backs up the play?
What if the pitcher doesn’t back up the play? Now you’ve got:
- The pitcher, who gives up the ground ball single,
- The batter, who grounds the ball into right field,
- The baserunner, who tries to go first-to-third and should be out, but actually scores the run because the third baseman doesn’t catch the ball and the pitcher doesn’t back up the play,
- The third baseman, who doesn’t catch the ball and allows the baserunner to score, and
- The pitcher again, who failed to back up the play.
Not to mention the second baseman, who
- Got a bad jump on the ground ball, allowing it to roll into right field, and/or
- Slowed the baserunner down by 3/10ths of a second by decoying him near second.
How can you decide how much credit goes to each of those players?
Real plays in baseball are more complicated than that, involving more players and more branches of the tree of possibilities. The reality is that the only way to “evaluate” the contributions of each player in every play is just to pick some option and go with it, right or wrong. The reality is also that these weird plays often come at game-changing moments, and contain more “value”—more change in one team’s chance of winning—than 20 other plays. It may seem like it should be more “objective” than the system I have outlined here, but the reality is, it isn’t.
You’re ranting.
Sorry. My basic point is that to make a “game state accounting” system work in practice, you have to make arbitrary choices. If you have to make arbitrary choices anyway, isn’t it better to acknowledge them than to hide them behind a smokescreen of mathematical objectivity?
You’re still ranting. Let me ask a couple of little questions first. What about contributions to losses? If you’re going to account for all the wins, why not account for all the losses as well?
Sure. I’m all in favor of it, in favor of doing a “Loss contributions accounting”. But this system, if it were to exist, would require the participation of many people. The first thing we need to do is to establish the network. Those people, once the network is established, can decide where to go with it from there. If they want to do a “Loss values accounting”, I’m all for it.
You’ve assumed that all wins are equal. That’s an arguable assumption, but not an inevitable assumption. If the Braves and Mets go down into the last weekend of the season and play head to head, shouldn’t those games count more than ten points like a Pittsburgh/Cincinnati game in the middle of June?
Again, the people manning the system can decide that. I suggested 10 points per game as a bedrock of the system. If the people who run the system vote on it and decide that a big game ought to be 20 points, 30 points, that’s fine with me.
I guess the big question is. . .well, two parts. First, why is this system any better than the MVP voting system that we have had since 1931?
It is better than the MVP vote in three ways. First, the MVP vote is based upon the notion that the voter, having seen one team play all season, can hold the entire season in his head, digest everything that has happened, and weigh the value of every player’s contribution on his mental scales—the players on the one team that he has seen, and the players on the 13 or 15 teams that he has seen in only a few games. In reality, this is entirely impossible. Our minds are not that complicated, that we can hold 100,000 events in our mind and weigh and measure each one to figure out who has contributed the most. Even to hold in your head a clear image of all the players and all the plays in one game, to balance all of that in your mind and summarize it fairly on the tally sheet. ..even that is very difficult. But it is a task much more in proportion to our real abilities than to form a mental accounting of the entire season.
Second, the MVP vote is an after-the-fact review. This would be there every day. It would develop with the season. June 23, we could say who the Most Valuable Player has been, through games of June 22. Heading into that critical Mets-Braves showdown in the last week of the season, we would be in position to say, based on the 159 games already played, that the Most Valuable so far has been Jose Reyes or Chipper Jones or whoever it has been. This is how the MVP race stacks up to this point.
And third, it is better because it invites the participation of people who care enough to try to get the answer right, rather than excluding the public, excluding the fans, and falling back on the elitist assumption that only the sportswriters are capable of evaluating the players. Not that I’m knocking the sportswriters or the MVP vote; I’m not. The sportswriters, I think, have always taken their votes seriously, and have always done their best to identify the most valuable players. They set up a system that was the best they could do in 1931. This is not 1931. We can do better.
But why is this system better than a traditional sabermetric model? Aren’t you, in proposing this “scoring network” evaluation, aren’t you standing sabermetrics on its head?
It has been the main thrust of sabermetrics, it seems to me, to avoid subjective judgment. This system provides space for subjective input.
It has been the foundation of sabermetrics, it seems to me, to show that wins are a natural function of accomplishments, that there are predictable relationships of batting events to runs and of runs to wins, and that therefore we can derive a clear statement of the player’s value from his accomplishments. Why are you throwing that out the window, to create a system that may very well wind up arguing, based on who got some timely hits and on the scoring decisions of anonymous fans, that some player who hits .275 with 18 homers has been more valuable than Albert Pujols?
First, the concept of a “traditional sabermetric model” is gibberish. Sabermetrics is not based on tradition. Sabermetrics is founded on the concept of questioning traditional beliefs, not on a moral imperative to embrace them.
Second, the main thrust of sabermetrics has been to pursue the truth.
Third, this system provides very limited space for subjective input. The error introduced by this may be and should be smaller than the error inherent in pursuing the question of who has been the Most Valuable Player along the long arc of probability.
Fourth, we are not throwing anything out the window. The things that we know, because of sabermetrics, we will still know.
As to the question of why this is better than a traditional sabermetric approach in terms of the answer that it delivers at the end of the day. . ..I don’t know that it will be better. I am only arguing that it might be better.
I have always believed in pursuing the truth along as many avenues as we were able to open up. If we have one method of analysis that suggests an answer, that’s interesting, but if we have two methods, three methods, four methods, independent of one another and as different from one another as possible, and if they all suggest the same answer, then we begin to see that that answer is unavoidable.
Or, on the other hand, suppose that we don’t find the answer we expect. Suppose that we find that the Most Valuable Player in the American League is actually Placido Polanco or Chone Figgins or Casey Blake. Remember the guy in ’67 who thought that Cesar Tovar had contributed more to the success of his team than Yastrzemski? What if he was right? What if some sort of anonymous player, because of his timely hits and crucial defensive plays, really has contributed more to the success of his team than A-Rod or Albert? Wouldn’t you want to know that?
It’s not as if, if we were to find that the Most Valuable Player in the National League has actually been Jimmy Rollins, it’s not as if we would be unable to look behind that evaluation. We would have a record of what it was, specifically, that Rollins had done that made him valuable. We would have box scores and videotapes of every game in which he was credited with value.
Saying that because you have statistics you don’t need to observe value is like saying that as long as you have coins you don’t need bank accounts. Just keep your coins in a jar and, if you want to know how many you have, count them.
I have been wrestling with the question of how you identify value in baseball for 30 years. If it had been possible to do something like this in 1977, I would have proposed it in 1977. It wasn’t possible then. We didn’t have VCRs then, we didn’t have spreadsheets, we didn’t have the ability to communicate immediately across the nation the way we can now, and we didn’t have the organized networks of people who care about the subject that we have now because of Bob Davids and Dave Smith. It’s a different world; it’s a better world. We have opportunities now that we never dreamed of in 1977.
August 4, 2007
- Red Sox vs. Mariners at Safeco
- Boston 4, Seattle 3
- Boston 4 points for offense, 6 for pitching and defense as per chart.
- 1 point to Youkilis (Single and Double, scored a run)
- 1 point to Ortiz (Singled and scored in each of 2-run rallies)
- 1 point to Ramirez (Participated in both rallies with double and a walk)
- 1 point to Varitek (2-run double was the key hit of the game)
- 5 points to Matsuzaka (Repeatedly got strikeouts to pitch out of trouble)
- 1 point to Papelbon (Scoreless 9th for the save)
August 5, 2007
- Red Sox vs. Mariners at Safeco
- Boston 9, Seattle 2
- Boston 6 points for offense, 4 for pitching and defense as per chart
- 1 point to Coco Crisp (2 doubles, 2 runs scored, RBI)
- 1 point to Ortiz (1 and 3 walks, 2 runs scored, RBI)
- 1 point to Ramirez (Single and Home Run, 2 RBI)
- 1 point to Drew (Triple and 2 walks, run scored and RBI)
- 1 point to Lowell (2 hits and an RBI)
- 1 point to Lugo (2 hits and a run scored)
- 4 points to Beckett (pitched until game was under control.)
August 6, 2007
- Red Sox vs. Los Angeles Angels in Anaheim
- Los Angeles 4, Boston 2
- Los Angeles 3 points for offense, 7 for pitching and defense as per chart
- 1 point to Figgins (Single, Triple and Sac Fly)
- 1 point to Kotchman (2-run single)
- 1 point to Izturis (Solo Home Run)
- 4 points to Weaver (Pitched first six innings, allowing 2 runs)
- 1 point to Speier (Scoreless 7th inning)
- 1 point to Shields (Scoreless 8th inning)
- 1 point to Rodriguez (Scoreless 9th inning)
Bill James
Brookline, Massachusetts |