WARP Formula

[Neags23]

Does anyone know the formula used for Wins Above Replacement Player? I can't find it anywhere.

I'm guessing it may be in the depths of Baseball Prospectus, but you have to be a subscriber. I wouldn't mind paying the 5 bucks for all they have to offer, but I don't want to if I don't know for sure they're willing to share their formulas.

[injury log]

The formula isn't really a secret. WARP combines VORP, which measure the number of runs a player contributes above replacement level on offense, and FRAA, which measures the number of runs a player contributes above an average player on defense. Both can be negative, of course. Add these together, you get the number of runs a player contributed on offense and defense combined; divide by 10 to get the number of wins contributed, which is WARP. At least, that's my understanding of how it works.

I'm not sure if the FRAA formula is public (some of Clay Davenport's formulas for fielding metrics aren't published, if I remember correctly), but WARP is built of off VORP, primarily, and you can find a description of the calculation of VORP in wikipedia, among other places.

[Mattymo]

Originally Posted by injury log

The formula isn't really a secret. WARP combines VORP, which measure the number of runs a player contributes above replacement level on offense, and FRAA, which measures the number of runs a player contributes above an average player on defense. Both can be negative, of course. Add these together, you get the number of runs a player contributed on offense and defense combined; divide by 10 to get the number of wins contributed, which is WARP. At least, that's my understanding of how it works.

I'm not sure if the FRAA formula is public (some of Clay Davenport's formulas for fielding metrics aren't published, if I remember correctly), but WARP is built of off VORP, primarily, and you can find a description of the calculation of VORP in wikipedia, among other places.

Yeah but what if their blood type is A-? And how about if they were born in the PM as opposed to the AM?

All these stat formula's dumbfound me...I'm more of a avg/HR/RBI/OBP guy

[Joshv02]

Originally Posted by injury log

I'm not sure if the FRAA formula is public

It is not, which means you can't really put together a WARP stat properly. If you are ambitious, you could do a ZR to runs type stat like this: http://www.baseballthinkfactory.org/...g_zone_rating/ or http://www.baseballthinkfactory.org/..._zone_rating1/

To calculate ZR, you can look here: http://www.baseballthinkfactory.org/...s_zone_rating/

I've no idea if the data in-game is available to calculate ZR. I've never tried. I don't believe Michael Humphries ever published the formula for his DRA, which uses basic stats, so I don't know if there is a good publically available formula you can use other than something like this.

[Neags23]

Originally Posted by Joshv02

It is not, which means you can't really put together a WARP stat properly. If you are ambitious, you could do a ZR to runs type stat like this: http://www.baseballthinkfactory.org/...g_zone_rating/ or http://www.baseballthinkfactory.org/..._zone_rating1/

To calculate ZR, you can look here: http://www.baseballthinkfactory.org/...s_zone_rating/

I've no idea if the data in-game is available to calculate ZR. I've never tried. I don't believe Michael Humphries ever published the formula for his DRA, which uses basic stats, so I don't know if there is a good publically available formula you can use other than something like this.

Hmm, well, there goes my idea for creating a mod that outputs WARPs....

[bababui]

I'm suprised that a mathematical wiz cant reverse-engineer this formula.

[Elendil]

Markus has promised to look into implementing a ZR-type fielding stat in the next full version. That would be sweeet.

I bet you could still come up with a pseudo Zone Rating by adjusting Range Factor at each position for pitching staff strikeouts and groundball percentage. I haven't worked out exactly how, but it seems doable.

[injury log]

Originally Posted by bababui

I'm suprised that a mathematical wiz cant reverse-engineer this formula.

Do you mean the FRAA formula? It would be a monumental mathematical task to reverse engineer, considering all of the factors it takes into account. It combines a player's defensive stats (the normal counting ones- Assists, Putouts, Errors, etc) with adjustments- and who knows what type, mathematically speaking- to take the quality of the fielder's teammates out of the equation, to consider the context in which the fielder played, to incorporate info about the pitchers who were on the mound when the fielder played and their GB/FB/K ratios, the LH/RH balance of the staff, park factors, how many baserunners were on when the player was in the field, etc, etc.

All that said, I've never read anyone argue persuasively that FRAA is much better than other defensive metrics available; sabermetricians still seem to consider defense a difficult thing to measure.

[Joshv02]

Originally Posted by Elendil

Markus has promised to look into implementing a ZR-type fielding stat in the next full version. That would be sweeet.

I bet you could still come up with a pseudo Zone Rating by adjusting Range Factor at each position for pitching staff strikeouts and groundball percentage. I haven't worked out exactly how, but it seems doable.

That would be cool.

Currently, game logs show ball location (hit/out). But, I think they don't use the zone grid, but rather the same grid (or a similar grid) as that on retrosheet (project scoresheet). So, you'd have to translate them into something similar -- that is, if the data were easily available in game exports.

You'd then have to know who was on the field for each play, too, so again, I just never fooled around with the export functions.

[orwell]

Originally Posted by Mattymo

Yeah but what if their blood type is A-? And how about if they were born in the PM as opposed to the AM?

All these stat formula's dumbfound me...I'm more of a avg/HR/RBI/OBP guy

I hear you. In general, I love stats (I figure anyone who plays OOTPB has to) but some are a bit much. But these guys have to create an ongoing need for their commercial existence. And some of them are announcers' dreams (i.e. BA with bases loaded on their son's birthday). "Situational" crap like that is 100% useless yet announcers love it to death.

[Neags23]

Originally Posted by orwell

I hear you. In general, I love stats (I figure anyone who plays OOTPB has to) but some are a bit much. But these guys have to create an ongoing need for their commercial existence. And some of them are announcers' dreams (i.e. BA with bases loaded on their son's birthday). "Situational" crap like that is 100% useless yet announcers love it to death.

Yeah, a lot of the situational stats I think are pointless. But I've found the BP stats to be pretty helpful/interesting, i.e. warp...

[injury log]

Originally Posted by orwell

I hear you. In general, I love stats (I figure anyone who plays OOTPB has to) but some are a bit much. But these guys have to create an ongoing need for their commercial existence. And some of them are announcers' dreams (i.e. BA with bases loaded on their son's birthday). "Situational" crap like that is 100% useless yet announcers love it to death.

Yeah, the value some people seem to find in those kinds of situational stats makes me fear for the American educational system.

I won't argue in favour of stats like VORP and WARP- I'm sure everyone's heard that before- but I would point out that they are measured in simple terms. VORP just counts how many more runs a player produced, on offense, than a rubbish player at his position. WARP does the same, but in wins. So a guy with a WARP of 5 helped his team win 5 more games than they would have with some random player they picked up off the waiver wire. Ok, it's not quite that simple, but pretty close.