Win Shares and WAR differences.

[willshad]

I was just browsing through some players' win shares totals, and was pretty surprised by how different WIn Shares totals are from WAR totals. Which is more accurate? For instance, Rogers Hornsby's 1924 season was easily his best, according to WAR, and yet according wto Win Shares, he only scores 38.7. This is a really good score, but only his fifth best..far behind his 1922 totals. I'm not saying I think one or the other is wrong, but asking which, according to the stat people, is more accurate.

[stevebogus]

My understanding is that the two systems begin from entirely different premises and arrive at their values in different ways. It should not be surprising that rankings based on the two should differ.

WAR begins with runs and eventually estimates how many wins a particular number of runs is worth, above a theoretical replacement level. Win Shares begins with wins (1 win = 3 Win Shares), then uses a complicated set of equations to divide up the credit for those wins.

My guess is a team which is very efficient with their runs (wins more games that expected for their run differential) would see their players "scored" better according to Win Shares in that season, since their runs resulted in more actual wins. And a team which underperforms will have fewer wins to spread around.

I suppose your opinion of which is "more accurate" depends on how you feel about luck vs. skills. If a team misses their pythagorean projection (over or underperforms) does that reflect the real skill of the team or is it just luck?