Team WAR

[Shoeless Jeff]

I apologize if this topic has already been posted, but I did a search and was unable to find anything. I was trying to explain to a co-worker today what WAR was (kudos to Edwin Starr) and my co-worker asked me a couple of great questions I was unable to answer.

First, if a team is filled entirely with players who are truly replacement (0.0 WAR), how many games would they win?

Second, if you determine that baseline, shouldn't the total WAR of all of its players for the year equal the total number of wins over that baseline?

Last year, the standings and cumulative WAR (per baseball reference) is below

Cards 91 - 71 40.5 WAR
Brewers 89 - 73 33.1
Cubs 84 - 78 40.6
Reds 75 - 87 29.5
Pirates 69 - 93 15.4

For the most part overall WAR is somewhat directionally close, but does not add up in total. So it it doesn't add up in total, does it make sense individually or are there other factors that need to be considered as I don't believe WAR takes into account timing of the various components, the interplay between individuals or just plain luck.

[TheBigLebowski]

I don't believe WAR starts with a true zero. By that I mean it wouldn't be calculable to take a team of 0.0 WAR players and project their wins. I believe WAR should only express what one player can or cannot add to an otherwise normally constructed ballclub.

[Griffey012]

I want to say the baseline is around 40-50 wins, as in a team full of 0.0 WAR players would still win that many games.

As far as the team thing, I would guess it evens out in the long run, but there will be plenty of short term variance.

[WrongVerb]

I think baseline number of wins would be 54. As Tommy Lasorda used to say: A team will win 1/3 of its games, lose 1/3 of its games. It's the remaining games that make all the difference.

So 1/3 of 162 is 54. Consequently, a team with 40 WAR should win somewhere around 94 games. However, there are some squishy parts to the WAR formula (pitching, defense) which may significantly add or subtract from that number.

[Kinsm]

I apologize if this topic has already been posted, but I did a search and was unable to find anything. I was trying to explain to a co-worker today what WAR was (kudos to Edwin Starr) and my co-worker asked me a couple of great questions I was unable to answer.

First, if a team is filled entirely with players who are truly replacement (0.0 WAR), how many games would they win?

Second, if you determine that baseline, shouldn't the total WAR of all of its players for the year equal the total number of wins over that baseline?

Last year, the standings and cumulative WAR (per baseball reference) is below

Cards 91 - 71 40.5 WAR
Brewers 89 - 73 33.1
Cubs 84 - 78 40.6
Reds 75 - 87 29.5
Pirates 69 - 93 15.4

For the most part overall WAR is somewhat directionally close, but does not add up in total. So it it doesn't add up in total, does it make sense individually or are there other factors that need to be considered as I don't believe WAR takes into account timing of the various components, the interplay between individuals or just plain luck.

https://www.sports-reference.com/blo...and-team-wins/

[PTjvs]

WAR makes no attempt to factor in leverage or variance (luck). WAR is trying to tell you how many games a player (or collectively a team) would win based on the performance levels they provide without attempting to integrate any of the actual performance events into their respective game context. If you have a team that collectively has 28 WAR for a season and have them play a million seasons, they should average to winning 80 games a year (replacement level team wins 52, +28 WAR).

If you want a metric that tries to credit each player with how much they contributed towards the actual wins and losses over the course of a season, what you want is WPA (win probability added).

The main reason WAR is much more often used than WPA is that it is more predictive of future behavior.

jvs

[RedsManRick]

The WAR replacement baseline is a .294 team winning percentage or just shy of 48 wins. A simpler way to consider it is that there are 1000 WAR to be distributed among the players every year. Though a nice round number, this number was arrived at empirically, looking at player performance relative to those types of players who were freely available. This specific figure represents a triangulation of sorts from a number of the sabermetric outlets that were refining WAR.

Read more about that here: https://blogs.fangraphs.com/unifying-replacement-level/

A few things it's helpful to keep in mind about how WAR works:

• WAR is calculated at the player level and all its components (e.g. batting, baserunning, defense) are calculated as runs above average
• Runs are almost always produced or prevented through the combined efforts of teammates. To isolate the contributions of an individual player, WAR uses a run value for each event based on how much it is worth on average. A single corresponds to a fraction of a run, an out to a negative fraction of a run, a homer for nearly 2 runs, etc., etc.
• A player's WAR is essentially the sum total of those events. The same goes for a team's WAR. Over time, "luck" (mostly timing) tends to balance out -- but it never does fully. Thus, a team that has especially good or bad timing over the course of the year is going to have a team WAR that over or under-estimates their actual run production & prevention.
• A team's actual W-L record is going to reflect both their underlying performance as estimated by WAR and the good or bad "luck" they experienced. So you'll find team WAR aligns much more closely to a team's pythag or "expected" W-L total than their actual W-L total.
• The replacement part of WAR is just the adding in of a "bonus" amount based on playing time to account or the difference between average and replacement level.
• The wins part is just turning runs to wins with a multiplier (e.g. ~10 runs = 1 win).

[bellhead]

WAR makes no attempt to factor in leverage or variance (luck). WAR is trying to tell you how many games a player (or collectively a team) would win based on the performance levels they provide without attempting to integrate any of the actual performance events into their respective game context. If you have a team that collectively has 28 WAR for a season and have them play a million seasons, they should average to winning 80 games a year (replacement level team wins 52, +28 WAR).

If you want a metric that tries to credit each player with how much they contributed towards the actual wins and losses over the course of a season, what you want is WPA (win probability added).

The main reason WAR is much more often used than WPA is that it is more predictive of future behavior.

jvs

That's interesting about the WPA, I'd looking now to see if I find the Reds 2020 with the recent additions.

[SteelSD]

That's interesting about the WPA, I'd looking now to see if I find the Reds 2020 with the recent additions.

Please don't waste your time. WPA is only marginally useful when evaluating the relative value of situational and state-based game events only. It's completely useless when applied as a player performance evaluation metric. Neither WPA or WAR are predictive measurements, but at least WAR will give you a starting point in determining whether or not a player might have performed well.

[bellhead]

Please don't waste your time. WPA is only marginally useful when evaluating the relative value of situational and state-based game events only. It's completely useless when applied as a player performance evaluation metric. Neither WPA or WAR are predictive measurements, but at least WAR will give you a starting point in determining whether or not a player might have performed well.

I did see where the sharps have moved our line after the last two signings.