Bill James just doesn't get it.

I'm talking about Linear Weights AND Win Shares.

by Tangotiger

Linear Weights is simply the number of runs a player contributed, over and above what an average player would have contributed, given this player's conditions. That's it. +10 does not mean that this player has 10 times more value than a player with +1 runs. Nor does it mean that a player that is -1 runs has no value. Linear Weights is about value over a .500 player.

Therefore, a player that is .490 does have less value than a .500 player. But a .500 player makes 2 million$/ year. So, a .490 player should be expected to make at least 1.5 million$. Even a team like the Expos that average .400 pay out 30 million $ in salaries. There's alot of value on that team.

As well, the question is not whether you'd rather have a .490 player with 100 at bats, or a .450 player with 600 at bats. That is a question about replacement level. From that perspective, you can create LWTSaboveRepl, and show the player's plus/minus against some other baseline, like .350. Linear Weights answers a very specific question, and therefore, should be applied only within the context of that question.

Ask 2 different questions, and you get 2 different answers. And you need 2 different methodologies.

Win Shares and the replacement level of .250. What Win Shares calculates is a player's gross win contribution, GIVEN THAT HE PLAYS ON A .500 TEAM. I flipped through the whole Win Shares book and nowhere does James mention this. A player's wins created is not based on some "marginal" level that is half or whatever the league average. No. It is based on how much from the league average he is. That is, how much from the .500 level is he.

This is how a .250 player gets ZERO win shares. It assumes that a .500 team, without this player will play at .475 or so. That is, an 81-win team will win 77 or so games. And if they had a .500 player instead, they would have won 81 games. So, that's 4 wins short. Doesn't make sense to penalize the other 24 players because we "know" that they are .500 players, always were and always will be. So, you have no choice but to dock the 4 wins from the one variable that changed. So, the average player would have created 80 runs, while this guy created 40 runs. That's a 40 run difference from the .500 level, and -4 wins created from the .500 level. Since the .500 player creates say about 4 wins, then this .250 player is worth 0 wins created.

It's an accident, AN ACCIDENT, that the "marginal level" is half the league average.

Loss Shares is very simple to calculate.

       Wins over .500 = (Wins - Losses) / 2        is a true statement.
Since we know what wins over .500 equals (see Linear Weights), and we know what Wins equals (see Win Shares), then we can derive Loss Shares.

Bill James is a great baseball mind, and is the inspiration for countless baseball analysts, including me. But sometimes, Bill James just doesn't get it.