Normalization Question Topic

slotterhodge
Posts: 107 (0)
Veteran
29
Prospect
Pardon the newbie question, but I'm not understanding something about normalization.  I've reads tons on here about it - and I'm amazed at how much people are willing to help each other!  I get the concept that a pitcher with X stats in 1968 won't perform as well because that was a pitcher's year, and a pitcher with the same stats in 1935 may perform better because that was a hitter's year.  What I don't get is why certain years don't 'normalize well' as someone in one of my leagues said.  Shouldn't a pitcher with an ERC# of 2.50 perform the same, whether it was from 1968 and 1935?  I know more goes into individual performance, but I guess my basic question is, if things are normalized why do certain years not 'normalize well'? 
12/8/2014 3:04 PM
frazzman80
Posts: 6356 (3)
Hall of Famer
238 18
All Star
The aspect of weather a season normalizes "well" depends on what you're comparing it to. For example, a pitcher from 1968 doesn't "normalize well" because his real life stats will look much better than his normalized or # stats. When a 1968 pitcher has a 1.00 real life WHIP and a 1.10 WHIP#, people will say he "doesn't normalize well" based on the fact that his normalized stats will be worse than his real life stats. That's just semantics of some of the users around here.

You are correct that if you look at the # stats, then a 2.50 ERC# pitcher from 1968 should be the same as a 2.50 ERC# pitcher from 1935. But the 1935 pitcher "normalizes well" or normalizes better than his real life stats and vice versa for the 1968 pitcher.

Where this really matters is in the salary though, because salaries vary for players based on the normalized stats. Here's a random example:
1963 Art Fowler - 90 IP 1.00 RL WHIP 1.03 WHIP# (normalizes worse than real life) 2.04 ERC# $2.82 million.
1991 John Hayban - 90 IP 1.03 RL WHIP 1.03 WHIP# (normalizes about the same) 2.06 ERC# $3.18 million.

So one "Normalizes poorly" and the other "normalizes well (or at least OK), but the price difference is substantial (and granted, Hayban gives up fewer HRs, so I imagine that might make up for the difference). So while both have similar numbers, Fowler is cheaper, but Hayban's normalized stats are closer to his real life numbers and will be perceived as "normalizing better" than Fowler. 

Also, the sim uses the actual OAV#, BB#, K# and HR# stats in the log5 formula for ABs, so just looking at ERC# will not tell you the whole story. In the example above, Fowler has a RL .219 OAV and a .227 OAV# (again, "normalizes poorly") while Hayban has a .225 RL OAV and a .227 OAV# ("normalized less poorly"). While both wind up .227 OAV#, again, there is a large salary discrepancy due to how their stats normalize. Since the league average for each player is used in the AB (both the pitcher and the hitter), Fowler is more likely to "help the hitter" (as his OAV normalizes poorly) than Hayban (whose OAV normalizes closer to his real life number).
12/8/2014 10:37 PM
biglenr
Posts: 1801 (0)
Hall of Famer
431 1
Pro

Here's a few other things to keep in mind...
 

  • LHP do worse in the sim than in real life.  Signficantly worse.  You can always find examples where this doesn't occur, but I avoid LHP like the plague.
  • The effect of the + and # stats is really to give you a baseline in comparing two players (and their costs).  I would not base team building decisions solely on them
  • The ballpark in which you play has a big impact.... 1994 Greg Maddux can not be expected to put up a sub 1.0 WHIP in Coors Field.  
  • The OAV and BB/9 rate of a pitcher is more important than the ERC rating.  In general, low OAV ratings also suppress HR's and extra base hits.  If choosing two pitchers with similar ERC and WHIP, choose the one wiht the lower OAV.  
  
ERC+ or ERC# is a good general tool to start your search... But don't end it there. 

FWIW... I usually build good pitching teams that end up in the top 4-5 in a league in WHIP and ERA. 
12/9/2014 1:35 PM
rbow923
Posts: 2246 (0)
Hall of Famer
1068
Hall of Famer
len, does low OAV really suppress HR? Since HR happens later in the decision tree I'd think HR are then based on actual hr/hit so no suppression. In fact my math says they should actually increase by a larger percentage since the result is somewhere between the pitchers stats and the hitters stats,
12/10/2014 1:08 AM
biglenr
Posts: 1801 (0)
Hall of Famer
431 1
Pro
Rbow -

In my opinion, it is first determined whether a plate appearance is a walk, out, or hit.

Once it is a hit, then it is determined what type of hit (single, hr, etc). 

So if you have a lower OAV, you allow fewer hits. 

Let's say a hitter/pitcher matchup goes 1000 times.... And the combination of pitcher/hitter has a 10% chance of the hit being a HR.

If the pitcher has a .280 OAV (assuming average hitter for simplification), then he allows 280 hits to that batter... And 28 of them are HR's.

If he has a .230 OAV, he allows 230 hits to that batter, and 23 of them are HR's. 

In my opinion, that's how it works.  I've walked through the posted logic trees several times and came up with that opinion (years ago), and it's pretty much matched my experience.  You see this a bit skewed sometimes because usage in the sim tends to favor deadball pitchers with ridiculously low HR/9 rates, and many pitchers in the modern era with lower OAV are fastball pitchers with higher than average HR/9 rates. 

12/10/2014 9:13 AM
rbow923
Posts: 2246 (0)
Hall of Famer
1068
Hall of Famer
the chance of a hit being a HR = (HR)/(HITS). If 2 pitchers give up the same number of HR but one of them gives up fewer hits then his HR/HIT number will be higher, resulting in just as many HR against an average hitter, more against big hitters. that is, if they use HR/HIT to determine the outcome.
I'm not disputing your result, just trying to figure out which formula they use.
12/10/2014 11:33 AM
biglenr
Posts: 1801 (0)
Hall of Famer
431 1
Pro

Lots of ways to figure it out... But I'll stick by my opinion unless someone proves it wrong.  My reading of the tables, and my experience experimenting with this, leads me to believe that I have described it correctly, but I'm more than willing to admit that I could be wrong. 

 

12/11/2014 8:43 AM
zubinsum
Posts: 1731 (0)
Hall of Famer
72
Hall of Famer
Posted by biglenr on 12/10/2014 9:15:00 AM (view original):
Rbow -

In my opinion, it is first determined whether a plate appearance is a walk, out, or hit.

Once it is a hit, then it is determined what type of hit (single, hr, etc). 

So if you have a lower OAV, you allow fewer hits. 

Let's say a hitter/pitcher matchup goes 1000 times.... And the combination of pitcher/hitter has a 10% chance of the hit being a HR.

If the pitcher has a .280 OAV (assuming average hitter for simplification), then he allows 280 hits to that batter... And 28 of them are HR's.

If he has a .230 OAV, he allows 230 hits to that batter, and 23 of them are HR's. 

In my opinion, that's how it works.  I've walked through the posted logic trees several times and came up with that opinion (years ago), and it's pretty much matched my experience.  You see this a bit skewed sometimes because usage in the sim tends to favor deadball pitchers with ridiculously low HR/9 rates, and many pitchers in the modern era with lower OAV are fastball pitchers with higher than average HR/9 rates. 

If you guys recall, I published a chart showing a pitcher's HR/100AB# before WIS started publishing HR/9# numbers.  Researching that chart and the normalization equation behind it gave me pretty intimate knowledge of how the sim behaves.  So while technically because of normalization this isn't quite correct, in practice this is exactly how the sim will behave over large samples.

Here is my old chart
12/12/2014 11:51 PM (edited)
slotterhodge
Posts: 107 (0)
Veteran
29
Prospect
Geez I wonder what would happen if I asked a really complex question!  lol   Thanx for all the info guys.  Still trying to understand it all.
12/13/2014 12:49 PM
Normalization Question Topic
△Top△Bottom

Search Criteria

Terms of Use Customer Support Privacy Statement

© 1999-2024 WhatIfSports.com, Inc. All rights reserved. WhatIfSports is a trademark of WhatIfSports.com, Inc. SimLeague, SimMatchup and iSimNow are trademarks or registered trademarks of Electronic Arts, Inc. Used under license. The names of actual companies and products mentioned herein may be the trademarks of their respective owners.