Log In   -   Register?

Sortable Draft Board!            Auction Calculator!            Probables Leaderboard!

new to this whole sabermetrics - war plus 48 does not give me a projection of total wins for a team, Why?

0 votes
Asked another way,  when I go to the War grid and pull up the 2013 cubs add up all war for the 25 players I get 21.  I read somewhere on this site that a team of replacement players should win 48 games.  If I as these numbers together I get 69 this is somewhere close to how many games the cubs will win. However if I do the same for the cardinals, I add all 25 players and get a value of 26 this only added to 48 this only gets you to 75, obviously the cardinals are going to finish in the 90's in wins.  Why does this not project out? if War is accurate, shouldn't this work out.

General observations.

Pitching seems to be undervalued in this war grid

The difference  in the rosters are the cubs have more 1 war players the cardinals have more 3-7 war players.
asked Sep 17, 2013 in Sabermetrics by chifan (10 points)

4 Answers

0 votes

Firstly, By only choosing those 25 players you will be leaving out players who have earned WAR but are no longer on the roster. Using the team WAR totals and adding the 48 replacement Wins I get 74 Wins for the Cubs and 82.5 for the Cards.


But this isn't a 'projection' or even an end of season estimate however, because it doesn't account for the games that are yet to be played. But it does give you a quick sense of whether a team is overachieving or getting an advantage in some way not measured in WAR.

You have also chosen two teams that are in the more outlier-y side of the process, as noted by Blake Murphy at BTB: http://assets.sbnation.com/assets/3118601/actual_v_war_bar.jpg

Fangraphs WAR also attempts to strip out the value of defense from pitching value which may or may not be why it feels low to you.

answered Sep 17, 2013 by James Gentile (2,000 points)
0 votes
Being new to the process as well, two things jump out to me.

First of all, my guess is you are only using Positional WAR for your calculations. Using the custom graphs, it took me a long time to realize that fangraphs & WAR aren't just straight up stupid, but I was.

Secondly, I have found that in the last ten seasons (When defensive stats are taken into account), WAR only accounts for about 78% of the movement in win rate. A variation of about 5 games/year is the expected variance per team per season.
answered Sep 26, 2013 by Pearberr (13 points)
0 votes
Also, I think its worth mentioning that WAR is adjusted for quality of opponent.  As a result, many teams with strong schedules have more WAR than actual wins, while many teams with weak schedules have more wins than WAR.  And this doesn't show a failure of WAR, it just shows that the best teams don't necessarily have the most wins because of schedule differences between teams.  

Evidence of this is that the 3 luckiest teams (PHI, PIT, STL) measured in wins minus WAR, are all members of the weaker national league.  

Most fans of sabermetrics, including myself, would argue that WAR is far more effective than wins in judging a team's talent level.  

I'm going to make a broad generalization here, but let's assume that all AL teams have a stronger strength of schedule than NL teams just for the sake of time.  AL teams should be expected to have more WAR than actual wins, and NL teams vice versa.  

Thus, the true "luckiest" teams in terms of wins vs cumulative WAR are AL teams with more wins than WAR, and the unluckiest teams are NL teams with more WAR than wins.  

From this excersize, one can conclude that the Yankees and Indians are due for fewer wins next season based on luck alone, while the Rockies, Cubs, Giants and Mets are due for more wins next season, assuming similar 2014 strength of schedules.
answered Nov 6, 2013 by Noah Baron (9 points)
0 votes
I've been wondering this as well. For the last 3 years (I missed the last couple weeks of this past season) I took and graphed the total WAR for each team vs the number of wins the got for that year on a scatter plot and ran a linear regression. The equation was y= 1.092x +59.719 with an R^2 value of 0.5723. So there's a pretty clear correlation between Wins and WAR, much more significant than something like batting average,  but it doesn't appear to have a 1:1 relationship, and it definitely looks like the "replacement level" should be closer to 60 than 48.

Edit: I have been informed that I did not include the pitcher WAR in this data. I haven't done so, but I imagine once pitcher WAR is included, it will be much better. On another note, why does there not appear to be a way to see the team's total WAR on this site?
answered Nov 8, 2013 by nonbelieverben (5 points)
edited Nov 13, 2013 by nonbelieverben