(This idea and the initial work is courtesy Solanus, but I have time and access to post this from work and he doesn't.)
Recently I took a look at the standings page on Baseball-Reference and saw the Pythagorean records for each team. With the Pythagorean record being a factor of runs scored and allowed, it gives an indication of our possible record. However, another way that baseball statistics determine "wins" is through the Wins Above Replacement calculations. I was curious as to how close those different methods track with actual wins.
I decided to keep to Baseball-Reference, so I simply pulled the main tables for Team Standings, Team Batting Value, and Team Pitching Value. I then totaled up the Batting WAR (the combined value) and Pitching WAR for each team, and compared the majors Wins total to the majors WAR total, in order to determine a "replacement" number of team wins per team game. I then figured out the number of "WAR Wins" by adding team WAR to the "replacement" wins for each team.
As of this morning, the majors have combined to play 4,722 games (amazingly enough, that's 2,361 wins and 2,361 losses). Similarly, there have been a total of 975.2 BR WAR accumulated so far. The difference between the majors wins and majors WAR, divided by the number of games, comes out to about 0.2935 "replacement" wins per team game. For a team like the Cardinals that has played 158 team games, the replacement level is about 46.4. Adding that to their 39.7 WAR, the Cardinals have 86.1 "WAR Wins".
When comparing actual Wins to Pythagorean Wins and then to "WAR Wins", what I found was that some teams had a fairly close correlation between Pythagorean Wins and "WAR Wins". This made sense to me initially, as both processes try to look at the team's performance (whether as a unit or as individuals) and determine how their record might come out. But others, like the Cardinals, had significant differences between the two - the help I think I need is in understanding why those differences exist.
Here are the raw numbers:
|Team||Wins||Losses||Pythag Wins||Pythag Losses||Total rWAR||rWAR Wins||WoPyth||WoWAR||WAR v Pyth|
(The last three columns compare Actual Wins to Pythagorean Wins or WAR Wins and then WAR Wins to Pythagorean Wins.)
Most of the teams have a number of WAR Wins in line with their Pythagorean Wins, with 19 teams having only three games difference between the two measures. Most of the others have a more significant deviation, but not wildly egregious.
And then there are the Cardinals, who have nearly 11 games of difference between their WAR and Pythagorean Wins counts. Their Pythagorean Wins are four more than their actual, but their WAR Wins are nearly seven below. I think I understand how their Pythag record is so good: they have a tendency to win big and lose close. But their WAR doesn't care that they are scoring their runs, because it doesn't measure how the Cardinals bunch up their hits or how they perform in RISP situations. It simply looks at the raw batting numbers and sort of assumes a more balanced distribution of when those hits, walks, and such happen.
I haven't done so, but I think if you look at Washington and Oakland, they may share some of the same traits, or perhaps their pitching staff becomes more stingy when runners are on. Similarly, the Dodgers and Rockies may be too disperse in their offense and/or too combustible when pitching with runners on.
Let us know what your own takes are on this subject.