Like us on Facebook


Follow us on Twitter





Results 1 to 11 of 11

Hybrid View

  1. #1
    Join Date
    Sep 2012
    Posts
    2,528
    vCash
    1500

    Question about WAR

    I don't know too much about SABR stats and all of that, but I've always been interested in learning more about them. Now I have a question about WAR: is there a defined value for a replacement player? Like are there certain stats that correlate to what a replacement player would contribute, and therefore make it plausible to define how much more valuable someone is than one?

    My issue is that the term "replacement player" means different things to different teams and therefore players. For example, Tampa Bay consistently loses top players and doesn't miss a beat. This is because of how well run the organization is and how deep their farm system is. A "replacement" player can come in and replace someone like Carl Crawford or Evan Longoria and the team might be only a game or two worse. Obviously those guys aren't 1-2 WAR guys. Tampa's replacement players are much better than average. Meanwhile take a team with a horrible farm system. If they called someone up that player might be far worse than an average replacement player. That means different players have different standards, and while that actual stat of WAR might not be affected (as it is calculated through individual stats) it may lose some of its meaning.

    So basically my question is what statistically, if anything, defines a "replacement player", and other than ruining the acronym, wouldn't the stat be more useful if it was something like "wins above average player"? Is that what it already is and I'm just confusing myself? Again, I know very little about SABR, just trying to clarify this.

  2. #2
    Join Date
    Dec 2011
    Location
    Wethersfield/Storrs, CT
    Posts
    8,793
    vCash
    1500
    WAR is based off of everyone in the league.

    It's based on an average that defines "replacement players" which happens to be 0-2 rWAR.

    WAR isnt just based on the team the given player is on. It's a pretty advanced stat that takes almost everything into consideration. So Longoria's WAR isnt affected by the quality of Tampa's bench. Another tip, don't think of "replacement player" as a bench player, because there are a ton of "replacement level" guys starting every day.

  3. #3
    Join Date
    Jul 2010
    Location
    San Diego, California, United States
    Posts
    7,511
    vCash
    1500
    This is one of the biggest hangups with SABRmetrics and is the #1 reason that proves -- to me -- that most who get upset with the stat are just ignorant about it. (Not you, you don't seem to be upset).

    In any statistical measure which seeks to compare multiple parts (in this case, about 600 different parts), a baseline needs to be set from which everyone can be compared. This is what "Replacement Player" is in WAR. It has nothing to do with any real backup on a team, or anyone's specific team.

    All stats that go into calculating WAR, be it offensive, defensive, or baserunning, all compare the player to "league average." League average is an objective, scientific measure. There is no question or doubt as to what "league-average" is.

    Therefore, the first step is always calculating a Wins Above Average. If you find the math on either site, you could do the work yourself to find it, and I believe BR even has it as a listed statistic (FG doesn't). Therefore, everyone is graded against the same standard, and therefore if one player is better than another, that means they were better than the standard by a higher margin.

    However, if you place league-average as your baseline, you get a value of "0" for a league-average player. However, by default, a league average player is better than about 50% of the league's players.... which means a league average player is quite valuable.

    Therefore, in order to express the fact that a league-average player contributes value, a new baseline -- which is lower -- must be set. This is wehre "replacement level player" comes in. This is considered a AAA player who they'd have to call up to replace a major league player. This, then expresses that league-average players are (insert amount) better than replacement level players, and therefore more valuable than below-average players (who still have enough value to be on a major league roster).

    For Fangraphs it is very easy to calculate. A league-average player is worth 20 net runs above replacement every 600 plate appearances.

    he equation for Runs Above Replacement would be:

    RAA + [(20/600)*PA]

    So, if a player is worth 50 Runs Above Average after their positional and park adjustments, and had 700 plate appearances, the equation would run as follows:

    50 + [(20/600) * 700] = 73.3 Runs Above Replacement

    You would then need to convert runs to wins. Again, this isn't doing anything "subjective", it is basically making the statistic communicate better information. That is, runs are nice, but wins are more telling than runs. How many runs equal a win?

    FanGraphs uses 10 runs per win, while BR has a very complex formula to adjust for the yearly run-scoring environment. Therefore, BR is communicating better information to the fan, while FG is really just making a cleaner number which communicates better information than runs, but not necessarily information translating to wins.

    And for anyone who wants to contend that 20 runs per 600 PA is just a "subjective measure", who cares? It doesn't make the statistic itself subjective because, again, we're just creating a baseline to judge everyone off of. If an average level player is worth 23 runs per 600, that doesn't suddenly change the fact that Trout was worth more runs than Cabrera. It may change the total amount a little bit, but not much.

    We do it with statistics all the time. We always create baselines to judge players, and they're more subjective with traditional stats than with advanced stats. At least with this research has been done to come up with the 20R/600PA number. With home runs we create arbitrary endpoints like 30, 40, 50, etc.

    In this situation, we always -- without fail -- group 39 HR with 30 HR, and not with 40, even though the player was much closer in value via HR to 40 than 30.
    Last edited by WadeKobe; 10-09-2012 at 01:56 PM.

  4. #4
    Join Date
    Nov 2006
    Location
    A Phillies Phan in Ocean Twp & BKLYN
    Posts
    5,406
    vCash
    1500
    Quote Originally Posted by Wade>Kobe View Post
    This is one of the biggest hangups with SABRmetrics and is the #1 reason that proves -- to me -- that most who get upset with the stat are just ignorant about it. (Not you, you don't seem to be upset).

    In any statistical measure which seeks to compare multiple parts (in this case, about 600 different parts), a baseline needs to be set from which everyone can be compared. This is what "Replacement Player" is in WAR. It has nothing to do with any real backup on a team, or anyone's specific team.

    All stats that go into calculating WAR, be it offensive, defensive, or baserunning, all compare the player to "league average." League average is an objective, scientific measure. There is no question or doubt as to what "league-average" is.

    Therefore, the first step is always calculating a Wins Above Average. If you find the math on either site, you could do the work yourself to find it, and I believe BR even has it as a listed statistic (FG doesn't). Therefore, everyone is graded against the same standard, and therefore if one player is better than another, that means they were better than the standard by a higher margin.

    However, if you place league-average as your baseline, you get a value of "0" for a league-average player. However, by default, a league average player is better than about 50% of the league's players.... which means a league average player is quite valuable.

    Therefore, in order to express the fact that a league-average player contributes value, a new baseline -- which is lower -- must be set. This is wehre "replacement level player" comes in. This is considered a AAA player who they'd have to call up to replace a major league player. This, then expresses that league-average players are (insert amount) better than replacement level players, and therefore more valuable than below-average players (who still have enough value to be on a major league roster).

    For Fangraphs it is very easy to calculate. A league-average player is worth 20 net runs above replacement every 600 plate appearances.

    he equation for Runs Above Replacement would be:

    RAA + [(20/600)*PA]

    So, if a player is worth 50 Runs Above Average after their positional and park adjustments, and had 700 plate appearances, the equation would run as follows:

    50 + [(20/600) * 700] = 73.3 Runs Above Replacement

    You would then need to convert runs to wins. Again, this isn't doing anything "subjective", it is basically making the statistic communicate better information. That is, runs are nice, but wins are more telling than runs. How many runs equal a win?

    FanGraphs uses 10 runs per win, while BR has a very complex formula to adjust for the yearly run-scoring environment. Therefore, BR is communicating better information to the fan, while FG is really just making a cleaner number which communicates better information than runs, but not necessarily information translating to wins.


    And for anyone who wants to contend that 20 runs per 600 PA is just a "subjective measure", who cares? It doesn't make the statistic itself subjective because, again, we're just creating a baseline to judge everyone off of. If an average level player is worth 23 runs per 600, that doesn't suddenly change the fact that Trout was worth more runs than Cabrera. It may change the total amount a little bit, but not much.

    We do it with statistics all the time. We always create baselines to judge players, and they're more subjective with traditional stats than with advanced stats. At least with this research has been done to come up with the 20R/600PA number. With home runs we create arbitrary endpoints like 30, 40, 50, etc.

    In this situation, we always -- without fail -- group 39 HR with 30 HR, and not with 40, even though the player was much closer in value via HR to 40 than 30.
    It does appear they pull these "runs" and the fact that 10 of them equal a win, out of the air.
    Quote Originally Posted by Crucis View Post
    Parity is about equality of opportunity, not equality of results.

  5. #5
    Join Date
    Jul 2010
    Location
    San Diego, California, United States
    Posts
    7,511
    vCash
    1500
    Quote Originally Posted by thefeckcampaign View Post
    It does appear they pull these "runs"
    No, the runs don't come from thin air at all, but I appreciate your cynicism.


    and the fact that 10 of them equal a win, out of the air.
    Yes and no. Last year there were about 8.6 runs scored per win. 10 seems to be a decent reference point without being exact. I think the 10 number makes it easier to compare players across eras, since their run production is already adjusted to the run-scoring environment, it seems unhelpful to era adjust twice.

    So, in those terms, it seems, to me, that BR does a better job of communicating an actual piece of data (Wins added above a replacement level player), FG does a better job of creating a number which is uniform across eras as a comparison tool.

    73 runs above average is 73 runs above average, no matter what era you do it in. Just because it translates to more wins in one era than another doesn't mean that player was more productive. So I like both and think they both have their advantages.

  6. #6
    Join Date
    Nov 2006
    Location
    A Phillies Phan in Ocean Twp & BKLYN
    Posts
    5,406
    vCash
    1500
    Quote Originally Posted by Wade>Kobe View Post
    No, the runs don't come from thin air at all, but I appreciate your cynicism.




    Yes and no. Last year there were about 8.6 runs scored per win. 10 seems to be a decent reference point without being exact. I think the 10 number makes it easier to compare players across eras, since their run production is already adjusted to the run-scoring environment, it seems unhelpful to era adjust twice.

    So, in those terms, it seems, to me, that BR does a better job of communicating an actual piece of data (Wins added above a replacement level player), FG does a better job of creating a number which is uniform across eras as a comparison tool.

    73 runs above average is 73 runs above average, no matter what era you do it in. Just because it translates to more wins in one era than another doesn't mean that player was more productive. So I like both and think they both have their advantages.
    He is not scoring these runs. The formula is just saying we think he would score these runs.

    Listen, I am not here to piss on SABR stats. I just feel this particular one has a lot of holes in it. The more I research it, I do not feel I can embrace it. The fact that it is 10 runs and not 8.6, if that is the average, is one within itself. I have a feeling it will go the way of the QB rating.

    I am sorry you feel like it is I am being cynical. I just do not feel this is just as the formula needs work.
    Last edited by thefeckcampaign; 10-10-2012 at 10:58 AM.
    Quote Originally Posted by Crucis View Post
    Parity is about equality of opportunity, not equality of results.

  7. #7
    Join Date
    Jul 2010
    Location
    San Diego, California, United States
    Posts
    7,511
    vCash
    1500
    Quote Originally Posted by thefeckcampaign View Post
    He is not scoring these runs. The formula is just saying we think he would score these runs.

    Listen, I am not here to piss on SABR stats. I just feel this particular one has a lot of holes in it. The more I research it, I do not feel I can embrace it. The fact that it is 10 runs and not 8.6, if that is the average, is one within itself. I have a feeling it will go the way of the QB rating.

    I am sorry you feel like it is I am being cynical. I just do not feel this is just as the formula needs work.
    nothing you've said is a valid criticism so far...

  8. #8
    Join Date
    Aug 2008
    Posts
    52,937
    vCash
    1500
    30 Team Stadium Checklist: 10 to go

    1) Yankees 2) Orioles 3) Rays 4) Red Sox 5) Mets 6) Braves 7) Phillies 8) Nationals 9) Marlins 10) Pirates 11) Padres 12) Astros 13) Mariners 14) Twins 15) Cubs 16) White Sox 17) Cardinals 18) Indians 19) Tigers 20) Royals

  9. #9
    Join Date
    Dec 2011
    Location
    Wethersfield/Storrs, CT
    Posts
    8,793
    vCash
    1500
    Quote Originally Posted by Pinstripe power View Post
    Yeah WAR just shows how good of a marksman each player is.

  10. #10
    Join Date
    Jul 2010
    Posts
    38,794
    vCash
    1000
    Moving to sabr forum.

    We have a few threads there like this, and I welcome you to read through those as well.


  11. #11
    Join Date
    Jul 2010
    Posts
    38,794
    vCash
    1000
    ^ perfectly said

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •