Ping: Cato

[Eckstein 4 Prez]

I'm sorry to be the one to point this out, but there is something wrong with the way Catobase calculates Win Shares.

After I posted that message about having one offensive-oriented league and one defensive-oriented league, whose Win Share numbers led me to believe the leagues were being valuated together, I decided to add up the Win Share numbers for both leagues and test them. What I found was an even deeper problem than the leagues being calculated together.

The first league I added up had 1530.1 hitting Win Shares, 654.1 fielding Win Shares, and 1220.5 pitching Win Shares. This added up to 3404.7 total Win Shares. The league had 1135 wins (162 games/team for 14 teams, plus one playoff game), which should add up to 3405.0 Win Shares. I'm not going to quibble over the 0.3, although I know Bill James is vehemently against reducing Win Shares below integers.

The problem is that this league's Win Share percentages, which should be at 48% for offense and 52% for defense, are at 44.94% for offense and 55.06% for defense (19.21% fielding and 35.85% for hitting). Now, this could be a manifestation of the problem I suggested in the other thread: both leagues are being calculated together. Or it could be that one team had something anomalous that causes this result (although if so, it wasn't apparent from perusing the team pages).

But here's the problem: that wasn't my pitching-heavy league, it was my hitting-heavy league. The other league had 3401.8 Win Shares (should be 3402) which broke down like this: 932.1 hitting, 851.2 fielding and 1618.5 pitching.

That breaks down to 27.40% offense and 72.60% defense (25.03% fielding and 47.58% pitching).

For both leagues calculated together, I have 36.17% offense and 63.83% defense (22.12% fielding and 41.71% pitching).

Now, I'm sure you wouldn't have let totals like this get through testing, so my guess is that you've something in there that's a variable being calculated as though it were a constant. Either that, or some small math error somewhere that's bollixing up the total. If there is anything I can do to help you in fixing these results, feel free to PM me or email me at whrimfunis@yahoo.com. Thanks, and again, sorry to be the bearer of bad news.

[Eckstein 4 Prez]

Bump.

[jcato]

I'll look trace through the win shares again and make notes about what is happening. There are some assumptions that have to be made because stats are not available. I'll identify those areas and why it is being calculated the way it is.

As you know, Win Shares is a very complicated calculation and there are many places where something can go awry. It may take a while to find the time to do this, but I will get to it.

[Eckstein 4 Prez]

Thanks. I just wanted to make you aware of the problem.