is war underrating catchers?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • filihok
    Team Veteran
    • Nov 2012
    • 3193

    #16
    Originally posted by Matthew C. View Post
    I thought one of the versions did consider pitch framing - can't remember which one. I guess neither attempts to quantify staff handling, if that is even quantifiable.
    This is, of course, the biggest thing that WAR is missing. There's a lot of work being done on pitch framing and pitch sequencing. Once those things can be quantified to any reliable degree they will be added to WAR and this graphs


    WARvs.Wins_.jpg
    *http://www.hardballtimes.com/main/ar...-war-good-for/




    will tighten up even more

    Comment

    • drstrangelove
      Registered User
      • Apr 2011
      • 3765

      #17
      Originally posted by filihok View Post
      You misunderstand WAR.

      WAR is the measure of the number of runs that a player produces on the field through his offense, defense, and baserunning.

      WAR does factor in that catcher play less games, by giving them no credit for games they don't play. A player does not produce on the field when he's not playing.

      Ivan Rodriguez is the only catcher with more than 10,000 career plate appearances. He ranks 62nd all time in plate appearances. His 74 WAR ranks 66th all time. (remember I typically use FanGraphs' WAR, not Baseball-Reference's WAR)

      If catchers caught more they would provide more value on the field to their team.

      You bring up pitchers. You mention 10 WAR in 40 games. I don't know if you had a specific pitcher in mind but in 1970 and 1971 Ferguson Jenkins put up 10.4 WAR in 39 and 40 games. In those 39 and 40 games he faced 1265 and 1299 hitters. In those two seasons the league leader in PA's for hitters were 745 and 739. He faced a lot more hitters than any hitter faced a pitcher. It's totally reasonable that he'd have more (or a similar) WAR even though he only played in 1/4 of the games.

      WAR measures the quality and quantity of what is done on the field.

      Catchers, because of the demands of the position, typically do not produce as much as other players.

      It's as simple as that.

      Where do catchers rank all time in things like runs, hits, doubles, triples, home runs, rbi, games, plate appearances, etc? Are catchers not treated 'fairly' by those stats?

      If you want to make WAR a rate stat (to normalize for playing time) you can take all players' WAR and divide by PA's. Then multiply by 500 or 600 or 700 or whatever. I imagine that you'll find that catchers' WAR per whatever number of PA's compares to that of other positions. Just as their H/PA, 2B/PA, 3B/PA, HR/PA, RBI/PA etc would.


      WAR is very accurate in measuring what it was designed to measure.

      [ATTACH=CONFIG]120075[/ATTACH]


      If you want to use WAR to measure Hall of Fame worthiness, then instead of adjusting WAR you can adjust the standards for catchers.

      The top 10 catcher have between 86 and 56 career WAR.

      Those totals rank between 36th and 177th all time among non-pitchers.

      Grade catchers on their own scale when discussing Hall of Fame worthiness.
      Good summary and spot on. Catchers are not like other players on the field because the demands of their task either deprives them of playing full time, lessens their effectiveness offensively (through sore knees, tired legs, brusied knuckles, etc.) or both.
      "It's better to look good, than be good."

      Comment

      • dominik
        Registered User
        • Oct 2008
        • 16944

        #18
        I agree with your points filihok (that WAR measures actual value and catchers by that definition fo produce less).

        I also now get that my pitcher example was wrong because a pitcher faces 1000 guys a season ( did not get that there is no normalization for pitchers-some 19th century guys have seasons with like 18 WAR).

        However doesn't that mean that WAR is of limited value when comparing different position (why have a positional adustment at all then?) and eras (no one would say old hoss radbourn is better than pedro martinez)?

        I don't believe that talent is distributed equally among positions (there is not a babe ruth or ty cobb for every position) however if the best catchers have about half the WAR than most other position leaders (bench is the best at 72 at a 45th rank, then the next catchers are like 80th or so) I think we can safely say that WAR is not fair in comparing catchers to other players.

        or does this just mean catching is not as important as people think?
        Last edited by dominik; 02-12-2013, 01:10 PM.
        I now have my own non commercial blog about training for batspeed and power using my training experience in baseball and track and field.

        Comment

        • brett
          Registered User
          • Jul 2006
          • 13923

          #19
          Originally posted by dominik View Post
          I agree with your points filihok (that WAR measures actual value and catchers by that definition fo produce less).

          I also now get that my pitcher example was wrong because a pitcher faces 1000 guys a season ( did not get that there is no normalization for pitchers-some 19th century guys have seasons with like 18 WAR).

          However doesn't that mean that WAR is of limited value when comparing different position (why have a positional adustment at all then?) and eras (no one would say old hoss radbourn is better than pedro martinez)?

          I don't believe that talent is distributed equally among positions (there is not a babe ruth or ty cobb for every position) however if the best catchers have about half the WAR than most other position leaders (bench is the best at 72 at a 45th rank, then the next catchers are like 80th or so) I think we can safely say that WAR is not fair in comparing catchers to other players.

          or does this just mean catching is not as important as people think?
          2 things: 1 Radbourn has significantly fewer war than Pedro, like 69 to 84. And pitchers being able to throw 678 innings is a direct sign of lower league quality and depth. It is both possible because of less league depth, and it is a strategic CAUSE of lower depth of performance level on the field. Innings pitched is a result and cause of league quality and so using war to rank players across eras requires some kind of adjustment for LQ. (In fact Radbourn tops out at 26 war when prorated to 162 games, but the guy pitched 65% of his team's innings, and equivalent of about 75 games in a 112 game schedule so he was more valuable than Pedro ever was in his time.

          2) When we look at catchers on a per game basis they look pretty good. (In fact the average catcher is equal to an average player at another position) and Gary Carter for example would work out to over 105 WAR given Hank Aaron's playing time. However I have seen true replacement level catchers at times and I think that a replacement level catcher is about half or so a win worse than a replacement level player at another position. After all, almost any player could fill in as a replacement at first base, regardless of his position, but virtually no one at another position could fill in at catcher so you have a lot smaller pool to get someone from if your starter goes down, AND your backup catcher still can't catch more than about 3/4 of the time.

          Oh and I guess there is a point 3. I think catchers should get some credit for ERA. Whether catchers vary much in affecting their staffs' ERAs I think that a replacement level catcher can certainly directly hurt a staff's ERA. Some of pitching value up to the average level should be taken from pitchers and given to catchers. That solves another issue with WAR that we get some huge seasons, such as Neikro with 9.6 WAR with only a 111 ERA+ and like a 10-16 record.
          Last edited by brett; 02-12-2013, 02:09 PM.

          Comment

          • Ben Grimm
            Semi-lucid User
            • Jun 2012
            • 6139

            #20
            Originally posted by brett View Post
            Oh and I guess there is a point 3. I think catchers should get some credit for ERA. Whether catchers vary much in affecting their staffs' ERAs I think that a replacement level catcher can certainly directly hurt a staff's ERA. Some of pitching value up to the average level should be taken from pitchers and given to catchers. That solves another issue with WAR that we get some huge seasons, such as Neikro with 9.6 WAR with only a 111 ERA+ and like a 10-16 record.
            Interesting topic and I do believe catchers get the shaft where WAR is concerned - but I also understand the outcome of it. IIRC, somebody (maybe it was even Bill James) did study some years back on CERA and found it to be totally unreliable. It also affects a player such as Tim McCarver. Tim pretty much spent his Phillies years as Carlton's personal battery mate - while Boone caught the rest. Is it really fair to add credit to McCarver, who was catching a HOFer, and leave Boone hanging in the wind while backstopping some very mediocre pitchers?
            "Chuckie doesn't take on 2-0. Chuckie's hackin'." - Chuck Carr two days prior to being released by the Milwaukee Brewers

            Comment

            • filihok
              Team Veteran
              • Nov 2012
              • 3193

              #21
              Originally posted by dominik View Post
              However doesn't that mean that WAR is of limited value when comparing different position
              Nope. Not in the slightest.
              WAR tells you how much value a player produced on the field; regardless of what position a player plays.
              Catchers don't produce as much as other players because they don't play as much. It's that simple.

              (why have a positional adustment at all then?)
              Do you think the average defensive first baseman would play as well at SS as the average fielding SS?

              If we put first basemen all over the field, overall fielding would go down. That's why we have positional adjustments.

              and eras ?
              I'm not really sure about this.
              Since WAR values are first calculated as runs above average, then every player in every ERA is compared to the average player in that era.

              I don't believe that talent is distributed equally among positions (there is not a babe ruth or ty cobb for every position) however if the best catchers have about half the WAR than most other position leaders (bench is the best at 72 at a 45th rank, then the next catchers are like 80th or so) I think we can safely say that WAR is not fair in comparing catchers to other players.
              It appears that you've missed everything that was said upthread.
              WAR treats every player the same when measuring how much value they contributed on the field.
              It is totally fair.
              Catchers do not provide as much value as other players because they play less.

              This just means that catchers should be measured against other catchers and not the general pool of other players.

              or does this just mean catching is not as important as people think?
              Doesn't mean that at all.

              Comment

              • filihok
                Team Veteran
                • Nov 2012
                • 3193

                #22
                Originally posted by brett View Post
                2) When we look at catchers on a per game basis they look pretty good. (In fact the average catcher is equal to an average player at another position) and Gary Carter for example would work out to over 105 WAR given Hank Aaron's playing time. However I have seen true replacement level catchers at times and I think that a replacement level catcher is about half or so a win worse than a replacement level player at another position.
                Show your work. If you've calculated that the replacement level for catchers is 1.5 wins less than the current replacement level and you can show it, then the 'community' would be happy to make that change.

                However, what happens to this line

                WARvs.Wins_.jpg


                if catchers are given 1.5 more WAR to adjust for that replacement level?
                Does the new model better fit real world observations?


                Here's a recent article by Dave Cameron at FanGraphs that discusses replacement level. He looks at players who have been 'freely available' (available on waivers, signed minor league deals, NRI, etc) in recent seasons.
                One of the most often cited criticisms of WAR is that replacement level is essentially an arbitrary construct, making the entire model just an act of theoretical speculation. The idea of replacement…


                There's 3 catchers on the list that Cameron identified as replacement level. Miguel Olivo, Chris Snyder and Eli Whiteside. Combined they've provided .8 WAR in 1457 PA's. That's about .4 WAR in 600 PA's. Nothing to suggest the -1.5 replacement level that you suggest.




                Oh and I guess there is a point 3. I think catchers should get some credit for ERA. Whether catchers vary much in affecting their staffs' ERAs I think that a replacement level catcher can certainly directly hurt a staff's ERA. Some of pitching value up to the average level should be taken from pitchers and given to catchers. That solves another issue with WAR that we get some huge seasons, such as Neikro with 9.6 WAR with only a 111 ERA+ and like a 10-16 record.
                Every defensive player on the field deserves some credit (or debit) for a pitcher's ERA. Not just the catchers.
                No one has been able to figure out how to distribute that credit so far.

                Comment

                • brett
                  Registered User
                  • Jul 2006
                  • 13923

                  #23
                  There is an assumption, and I mean an assumption in the logical sense, that the next available player at every position that a team would probably have if they lost someone would be 64% as good (32% winning percentage player) as an average player at that position. I am challenging this assumption. I do not believe it is the logical default. I think the logical default position should be that the next available likely player at different positions would have a different value relative to an average player at that position. In the development of WAR the 32% replacement player at all positions was assumed or "defined" for pragmatic reasons. What WAR says then is "IF the next available player at all positions was a 32% player, then the WAR values of all players would be the ones that we get.

                  In addition to this, I am also saying that the replacement catcher might not even be necessarily lower, but that if a team had to replace its starter they would historically have to use a replacement catcher for some of the games, AND another sub-sub replacement level catcher for some of the games because no catcher can catch 162 games, and a replacement catcher is going to tend to be capable of catching fewer games than an MLB starter.

                  Anyway, assigning different replacement levels to different positions would not affect the graph. If we gave catchers .4 war for a lower replacement level, then we could give other positions a net of -0.4 or -.05 per position which would keep the intercept at 52 wins.

                  By the way, how were the teams selected for the graph? It looks like there were 2 outlier teams, one with 50 some wins and another with 110 or so wins that act as stabilizing points that have a big impact on the slope of the graph. A WAR should be a win, meaning that the slope should be 1.00, but it is 0.97 meaning that a WAR is only equivalent to .97 wins. That leaves .03 wins per war unaccounted for. Given that an average team produces 81 wins versus 52 for a replacement team, the average team is getting 29 WAR. .03 x 29 is 0.87 wins unaccounted for so the estimate that a replacement catcher may be 0.4 to 0.7 war lower than other positions would be within the .87 wins missed by war in the equation.

                  Edit, actually teams are winning .97 per WAR, not getting 1 win per .97 war so the average team is .87 wins below what they would have with a perfect correlation. Still, I'd like to see the equation with the two outliers removed.
                  Last edited by brett; 02-12-2013, 08:26 PM.

                  Comment

                  • brett
                    Registered User
                    • Jul 2006
                    • 13923

                    #24
                    Originally posted by filihok View Post

                    Every defensive player on the field deserves some credit (or debit) for a pitcher's ERA. Not just the catchers.
                    No one has been able to figure out how to distribute that credit so far.
                    I am not taking about the defense effect on ERA, but the catcher's effect on (defense independent) pitching WAR.

                    Comment

                    • filihok
                      Team Veteran
                      • Nov 2012
                      • 3193

                      #25
                      Originally posted by brett View Post
                      There is an assumption, and I mean an assumption in the logical sense, that the next available player at every position that a team would probably have if they lost someone would be 64% as good (32% winning percentage player) as an average player at that position. I am challenging this assumption. I do not believe it is the logical default. I think the logical default position should be that the next available likely player at different positions would have a different value relative to an average player at that position. In the development of WAR the 32% replacement player at all positions was assumed or "defined" for pragmatic reasons. What WAR says then is "IF the next available player at all positions was a 32% player, then the WAR values of all players would be the ones that we get.
                      WAR is defined at that level for the reasons stated in the article I posted above and I'm posting again.

                      One of the most often cited criticisms of WAR is that replacement level is essentially an arbitrary construct, making the entire model just an act of theoretical speculation. The idea of replacement…


                      In addition to this, I am also saying that the replacement catcher might not even be necessarily lower, but that if a team had to replace its starter they would historically have to use a replacement catcher for some of the games, AND another sub-sub replacement level catcher for some of the games because no catcher can catch 162 games, and a replacement catcher is going to tend to be capable of catching fewer games than an MLB starter.
                      Again, I think you're misunderstanding replacement level.

                      Anyway, assigning different replacement levels to different positions would not affect the graph. If we gave catchers .4 war for a lower replacement level, then we could give other positions a net of -0.4 or -.05 per position which would keep the intercept at 52 wins.
                      I still see no evidence that this is necessary.

                      By the way, how were the teams selected for the graph? It looks like there were 2 outlier teams, one with 50 some wins and another with 110 or so wins that act as stabilizing points that have a big impact on the slope of the graph. A WAR should be a win, meaning that the slope should be 1.00, but it is 0.97 meaning that a WAR is only equivalent to .97 wins. That leaves .03 wins per war unaccounted for. Given that an average team produces 81 wins versus 52 for a replacement team, the average team is getting 29 WAR. .03 x 29 is 0.87 wins unaccounted for so the estimate that a replacement catcher may be 0.4 to 0.7 war lower than other positions would be within the .87 wins missed by war in the equation.

                      Edit, actually teams are winning .97 per WAR, not getting 1 win per .97 war so the average team is .87 wins below what they would have with a perfect correlation. Still, I'd like to see the equation with the two outliers removed.
                      The teams were selected 'randomly'. I don't know how.

                      Here's a link to FanGraphs WAR for every player.
                      Here's a link to Baseball-Reference WAR for every player.
                      Here's a link to W-L records for every team.

                      Feel free to re-create the graph using your own randomly selected teams.

                      Comment

                      • filihok
                        Team Veteran
                        • Nov 2012
                        • 3193

                        #26
                        Originally posted by brett View Post
                        I am not taking about the defense effect on ERA, but the catcher's effect on (defense independent) pitching WAR.
                        Ok.
                        That doesn't negate that every player who dons a glove for a team has an effect on that team's pitchers' ERAs.

                        If you can find a way to quantify the catcher's effect on pitchers' performance you'll be a hero in stat nerd world.

                        Comment

                        • filihok
                          Team Veteran
                          • Nov 2012
                          • 3193

                          #27
                          I quickly plotted the numbers for 2012

                          WARGraph.jpg

                          This is using FanGraphs' numbers

                          A couple of things to remember:
                          FanGraphs' replacement level is about 43 wins
                          FanGraphs' pitcher war is based on FIP not runs allowed
                          for a single season, one should expect more error as wins and losses will compound as one team beats another

                          The model still matches real life observations very well:
                          The expected intercept is at 43. The observed intercept is at 44.9
                          The expected slope is 1. The observed slope is .958

                          Comment

                          • filihok
                            Team Veteran
                            • Nov 2012
                            • 3193

                            #28
                            Just did 2011 and 2010

                            2011
                            WAR2.jpg


                            2010
                            WAR3.jpg
                            Last edited by filihok; 02-12-2013, 10:28 PM.

                            Comment

                            • filihok
                              Team Veteran
                              • Nov 2012
                              • 3193

                              #29
                              Then, I randomly selected 30 teams.

                              I did this by using the Random Integer Generator at Random.org

                              I got the following numbers

                              Random1.jpg

                              Random2.jpg

                              Random3.jpg



                              I used these numbers to select the teams.

                              The first group of numbers was for 2012
                              The second group of numbers was for 2011
                              The third group of numbers was for 2010

                              So, for example, the first random number was 26. This was the 26th team in the 2012 set which was the Colorado Rockies with 30 WAR and 64 wins. I continued like this for each group.

                              Unfortunately, I found the random number generator sometimes produced the same number within a run. When this happened I skipped that number and moved onto the next team, leaving that slot temporarily blank. I then ran the generator for a forth time and got the following numbers

                              Random4.jpg

                              I used these to fill in the blanks. Using the first two numbers to fill in the missing slots in 2012, the second two for 2011 and the third two for 2010.


                              That gave me this data set.

                              WARCORR.jpg


                              So, for the sample set the equation is: y=1.0367+40.731 with an R2 of .71972
                              compared to an expected equation of: y=1+43


                              Not as exact as the data from the article, but, still enough to show that WAR is measuring what it claims to measure.



                              *probably could have done this much more smoothly...but...it's late...and once I started I didn't want to start over.
                              Last edited by filihok; 02-12-2013, 10:41 PM.

                              Comment

                              • brett
                                Registered User
                                • Jul 2006
                                • 13923

                                #30
                                Thanks for all the work.

                                Comment

                                Ad Widget

                                Collapse
                                Working...
                                X