Normalization

Just curious how SOM normalizes it's stats (and apologizing in advance for a long post )

For a season it's simple, use the current season to normalize stats. So if the league average is .260 and someone hits .280, The average on their SOM card will relect a .300 average to compensate for only influencing the result 50% of the time.

If we look at a deadball period, use 1906 National league for example would have a "norm" of one HR per 320 at bats. Take Tim Jordan, one of the leading HR hitters that season (1906 Brooklyn), and he would be a HR per 44 at bats. He would then hit a HR per 23.6 at bats on "his" SOM card. Put him against modern era pitching of about one HR per 24 AB and he will hit about 20 HR in a 550 AB season.

One of the biggest problems is normalizing someone like Babe Ruth, in particular in the early 20's. With a league average of a HR per 89.7 AB's in 1921 AL, and Ruth with a HR per 9.2 AB's, his "card' for SOM would have to be 1 HR every 4.8 AB's to normalize for that season. Against closer to historical 1 HR per 36 B's, he would hit 1 HR per 8.5 AB's. In a 540 AB season like he had in 1921, it would be 64 Homeruns.

It's not that bad, I've always felt though that for normalization it would make sense to use the mean of MLB historical numbers averaged with that particular season. I think it works better when putting together players of different eras.

Anyone know what SOM use for it's normalization?

I don't know, but what would be the numbers for '21 Ruth if you use plate appearance instead?

one HR every 6.23 PA's to normalize to the 1921 AL season, which would yield the 59 HR's in 693 PA's as he did in 1921.

This card against a fairly average to historical MLB stats of all years of 1 HR per 39 PA's would yield 64 HR in 693 PA's.

The other questions I have about normalization are as follows - are the 1923 and on Ruth cards factored as hitting in Yankee stadium, where the park effects were favorable to LH power hitters? Hopefully yes, otherwise his Homerun numbers would see a large increase if he indeed did bat in 50% of his games at Yankee stadium.

Ok, so 59 Hrs look just right.

For stadium adjustment, you can have a fair guess about how Strat does it by looking at the "direct" results (the advanced mode, as if the diamonds would not exist). Babe 1921 and 1923 cards have 3.2 direct homerun chances in 8 BP chances, so SOM didn't adjust, they simply apply the 1-8 (neutral) stadium rates for those years. The other two years, Babe has much higher direct homeruns per 8 BP chances, so clearly, SOM adjusted the homerun chances in accordance to a hitter-friendly stadium.

so clearly, SOM adjusted the homerun chances in accordance to a hitter-friendly stadium.

That's good to see. Otherwise you put Maris 61 in a lefty friendly park like Yankee stadium, and it might result in 80 HR for the '61 Maris

Normalizing properly for stadium is very similar to the sabermetric type formulas for ballpark adjusted ERA, or Ballpark adjusted offensive WAR. The only thing is while the Sabermetric stats try to factor it in for a player by player comparison, SOM needs to use the game mechanics and game numbers for the stadium to make it work out right.

Of course, pitcher numbers to to be adjusted for the home stadium they played in as well. For instance Ferguson Jenkins allowed a lot of Homeruns in the slightly HR friendly Wrigley 66 field, but allowing hardly any homeruns, But Rick Reuschel allowing 16 in 242 innings is a pretty good accomplishment in the home run derby park of '78 Wrigley.

"Normalization" or statistical accuracy algorithms in ATG make zero sense, especially in high cap leagues. There never was a season where Ruth, Mantle, Hornsby and Bonds all played against Maddux, Pete, Pedro and Walter. To try and manipulate the results with these lineups is a fools game and does not work.

All this does is create frustration and wonder on the part of the manager.

But they are playing with it lately. Watch you key players, if they are over performing their card after 120 games (especially HRs), good luck for the rest of the season.

sschu

"Normalization" or statistical accuracy algorithms in ATG make zero sense, especially in high cap leagues. There never was a season where Ruth, Mantle, Hornsby and Bonds all played against Maddux, Pete, Pedro and Walter.

Yes, it indeed never happened, but that is the idea - to work out a framework where "realistic" results can be hopefully achieved. The "Filed of Dreams" idea is common to baseball sims.

Hi caliber lineups cause problems in many ways - there is never a real "rest" for the pitcher. One problem I do see that is extremely difficult to counteract is this - many of the top hitters are also high walk players. I large part this was due to pitching around them, just doing your best not to give them something good to hit. Unfortunately I think this almost impossible to verify by statistics, but reducing walks for hi-end hitters would make sense. One thing that should be done is to not include IBB's in stats at all, as this was a decision by the opposing manger, but just doing this won't fix the issue.

To me, one of the hardest things to replicate properly is to normalize for different eras. The swings were different with most deadball players for instance. They are not going to hit more homeruns merely because they face modern pitching and ballparks - it would take a different swing as well.