rockaction
Footballguy
ESTABLISHING CROSS-POSITIONAL MEASUREMENTS OF VALUE THAT ASSIGN EACH POSITION A CORRESPONDING VALUE BASED ON HOW MANY RUNS EACH POSITION IS WORTH BECAUSE THAT POSITION, ON AVERAGE, SAVES THAT MANY RUNS. THE SAME NUMBER IN EACH OF THE PAST TWENTY YEARS, I SAY. YEAH, IT'S A CONSTANT. SMILE, YOU LOVE CROSS-POSITIONAL WAR!
Oh wow. Spent my day not watching baseball, but in a complete tizzy. I was doing my fantasy football stuff on Twitter and the Cal Raleigh /Aaron Judge MVP debate was in full swing And they are flipping out and going back and forth and all that. Then they start getting all advanced and they're talking about WAR and they're tweeting this stat called fWAR out and comparing Judge and Raleigh like these numbers show what their values are at the particular position they play and only that position? No? Wait, you're really telling me that I can compare an SS, C, and RF? Yes, I can?!
Awesome! How do we got about doing that?
"Well, we have a chart where we figured out the value of one position versus another. It's measured in how many runs are saved or prevented by each position during the year."
Oh! Interesting! How did you come up with each position's value. Wait, are these like absolute values or constants or . . .
"Yes, they're constants year in and year out." Here check out our website at fanGraphs. It says that "The actual numbers we use are based on some calculations that were done about a decade ago that used the performance of players who moved positions."
Oh. What did those calcuations tell you?
How were you able to assign a value to the position?
Oh, is that not public or something? Is it private?
OKAY ENOUGH OF THE CHARADE, THIS IS ME TALKING TO YOU, THE FBG
I'm serious. This is how they measured the difference in runs saved from one position to the next. The following links and description are their methodology
library.fangraphs.com
"We compare players to their positional averages because we want to base player evaluations on a similar set of plays. It would be silly to compare the conversion rate of a third baseman against that of a right fielder because not only do they field a different number of balls, they’re just a totally different kind of batted ball. As a result, when you look at the initial defensive statistic, you’re looking at only how that player compares to their fellow position-mates.
And when it comes to overall player evaluation and WAR, we need some way to balance out those differences in position. We need to find a way to account for the difference between the positions. The positional adjustment does just that.
In general, we want to add runs for players who play tough positions and subtract runs for players who play easier positions to account for the fact that average at one does not equal average at the other in terms of total run prevention. For example, for a shortstop, the adjustment is +7.5 runs per full season. For a left fielder, it’s -7.5 runs per full season.
That means the difference between an average left fielder and an average shortstop is about 15 runs per season. That’s a sizable gap. The positional adjustment is a method for putting a bunch of different positions onto one uniform scale so that when you add the adjustment to a player’s runs saved to get what we call “DEF,” meaning Defensive Runs Above Average, you can compare players no matter their position.
The actual numbers we use are based on some calculations that were done about a decade ago that used the performance of players who moved positions. It’s certainly reasonable to suggest that those numbers have changed as the game has changed, so use the adjustments as guides more than as firm rules. The DH adjustment might be too negative because it’s harder to hit when you’re not playing in the field, and the catcher adjustment might be a bit too large. There’s lots of room to disagree on the precise decimals and if you’re so inclined, I’d invite you to come up with a more accurate rendering of the numbers. Here is one recent example of such work.
One of the biggest challenges in this effort is that players aren’t assigned positions at random and don’t switch positions at random, meaning we have to get around the likely bias. For example, left-handed players don’t play shortstop, so the very best left-handed defenders wind up at other spots even if a right-hander that talented would play short. We all recognize that these are estimates and you could certainly slide them around a little if you think we’re wrong.
But the important thing is that we need a positional adjustment of some kind. We calculate defense based on position average but the position averages need to be corrected so that an average catcher and an average first baseman don’t look the same. The positional adjustment does this for us. It might make sense to come up with a better version of the numbers we use, but adjusting for position is a vital piece of overall player analysis.
Positional adjustments are calculated based on a full 162 games, which equates to 1,458 defensive innings. So if a first baseman plays 1,214 innings with -12.5 positional adjustment for a full season, his adjustment for that period will be -10.4 runs. For players who play multiple positions, it’s simply the adjustment at each position added together.
C +12.5
1B -12.5
2B +2.5
SS +7.5
3B +2.5
LF -7.5
CF +2.5
RF -7.5
DH -17.5
eta* 10/22
here is the link to the first calculations "done about a decade ago"
here is their recent example
www.hardballtimes.com
eta2* The problems are threefold with this and it's right from the jump
1) They're studying only players who switched positions to get an idea of how many runs each position saves comparatively. There are a few glaring and obvious problems with that
2) They needed a position adjustment, but those formulas and these coefficients listed about are neither public nor extant and are not available for scrutiny. In addition, they're two decades old and based on what? What data?
3) The third problem is that the "runs saved" that drives the comparative engine is a model based on top a model and not actuality. It's wild.
Anyone over here from the football thread that is interested in a more fleshed-out technical argument can PM me. Peace and Godspeed.
Oh wow. Spent my day not watching baseball, but in a complete tizzy. I was doing my fantasy football stuff on Twitter and the Cal Raleigh /Aaron Judge MVP debate was in full swing And they are flipping out and going back and forth and all that. Then they start getting all advanced and they're talking about WAR and they're tweeting this stat called fWAR out and comparing Judge and Raleigh like these numbers show what their values are at the particular position they play and only that position? No? Wait, you're really telling me that I can compare an SS, C, and RF? Yes, I can?!
Awesome! How do we got about doing that?
"Well, we have a chart where we figured out the value of one position versus another. It's measured in how many runs are saved or prevented by each position during the year."
Oh! Interesting! How did you come up with each position's value. Wait, are these like absolute values or constants or . . .
"Yes, they're constants year in and year out." Here check out our website at fanGraphs. It says that "The actual numbers we use are based on some calculations that were done about a decade ago that used the performance of players who moved positions."
Oh. What did those calcuations tell you?
How were you able to assign a value to the position?
Oh, is that not public or something? Is it private?
OKAY ENOUGH OF THE CHARADE, THIS IS ME TALKING TO YOU, THE FBG
I'm serious. This is how they measured the difference in runs saved from one position to the next. The following links and description are their methodology
Positional Adjustment
In essence, the positional adjustment is a correction to account for the fact that different positions are more challenging than others, which is a pretty easy thing to accept.
"We compare players to their positional averages because we want to base player evaluations on a similar set of plays. It would be silly to compare the conversion rate of a third baseman against that of a right fielder because not only do they field a different number of balls, they’re just a totally different kind of batted ball. As a result, when you look at the initial defensive statistic, you’re looking at only how that player compares to their fellow position-mates.
And when it comes to overall player evaluation and WAR, we need some way to balance out those differences in position. We need to find a way to account for the difference between the positions. The positional adjustment does just that.
In general, we want to add runs for players who play tough positions and subtract runs for players who play easier positions to account for the fact that average at one does not equal average at the other in terms of total run prevention. For example, for a shortstop, the adjustment is +7.5 runs per full season. For a left fielder, it’s -7.5 runs per full season.
That means the difference between an average left fielder and an average shortstop is about 15 runs per season. That’s a sizable gap. The positional adjustment is a method for putting a bunch of different positions onto one uniform scale so that when you add the adjustment to a player’s runs saved to get what we call “DEF,” meaning Defensive Runs Above Average, you can compare players no matter their position.
The actual numbers we use are based on some calculations that were done about a decade ago that used the performance of players who moved positions. It’s certainly reasonable to suggest that those numbers have changed as the game has changed, so use the adjustments as guides more than as firm rules. The DH adjustment might be too negative because it’s harder to hit when you’re not playing in the field, and the catcher adjustment might be a bit too large. There’s lots of room to disagree on the precise decimals and if you’re so inclined, I’d invite you to come up with a more accurate rendering of the numbers. Here is one recent example of such work.
One of the biggest challenges in this effort is that players aren’t assigned positions at random and don’t switch positions at random, meaning we have to get around the likely bias. For example, left-handed players don’t play shortstop, so the very best left-handed defenders wind up at other spots even if a right-hander that talented would play short. We all recognize that these are estimates and you could certainly slide them around a little if you think we’re wrong.
But the important thing is that we need a positional adjustment of some kind. We calculate defense based on position average but the position averages need to be corrected so that an average catcher and an average first baseman don’t look the same. The positional adjustment does this for us. It might make sense to come up with a better version of the numbers we use, but adjusting for position is a vital piece of overall player analysis.
Positional adjustments are calculated based on a full 162 games, which equates to 1,458 defensive innings. So if a first baseman plays 1,214 innings with -12.5 positional adjustment for a full season, his adjustment for that period will be -10.4 runs. For players who play multiple positions, it’s simply the adjustment at each position added together.
C +12.5
1B -12.5
2B +2.5
SS +7.5
3B +2.5
LF -7.5
CF +2.5
RF -7.5
DH -17.5
eta* 10/22
here is the link to the first calculations "done about a decade ago"
here is their recent example
Re-Examining WAR’s Defensive Spectrum
Now that we have more information, is it time to update our values?
eta2* The problems are threefold with this and it's right from the jump
1) They're studying only players who switched positions to get an idea of how many runs each position saves comparatively. There are a few glaring and obvious problems with that
2) They needed a position adjustment, but those formulas and these coefficients listed about are neither public nor extant and are not available for scrutiny. In addition, they're two decades old and based on what? What data?
3) The third problem is that the "runs saved" that drives the comparative engine is a model based on top a model and not actuality. It's wild.
Anyone over here from the football thread that is interested in a more fleshed-out technical argument can PM me. Peace and Godspeed.
Last edited:
