All Forums > Discussions in SIMology
11/13/2013 2:26 AM (edited)
I asked admin if there was a 'player under posession penalty'..the answer was no. The reason I asked was that I play in high cap leagues where you can afford to pay a 30mpg player to play 10mpg. The FG% on these players is often just terrible regardless of RL FG%. A penalty is an intended bias on a player or team. I trust admin not to bold face lie. Therefore why does the player, played significantly under req mpg shoot so poorly wrt RL? IMO it may work like this, as an example, a player that plays 30mpg has a 0.9 bias on RL advance stats for the first say 30% of req MPG, then 40% @ 0.0 bias, followed by 30% @ 1.1 bias wrt FG% & req mpg. In the end you should get near RL FG%, but if you only use the 0.9 bias part of the players game they don't shoot so good. So in fact it's not a penalty but a matter of the sim dynamics..therefore no bold face lie. (lol) When a system runs on a per min basis I had always thought the advance estimators would be constant thoughout the game. Appears I was wrong...again. It may be there to replicate warming-up. Any thoughts on this matter?
11/12/2013 11:44 PM (edited)
btw - now I try to use +80% of my bench players RT. Even in an unlimited cap league I try to stay under 21,500 min's total. I think this increase a teams PER.
11/13/2013 12:57 AM (edited)
Most likely on a per/min basis, say a varible bias from like 0.8 - 1.2.  I also doubt it's a linear function either, more likely S shaped curve. Those first 10 min's of a 30mpg player can be really ugly...
11/13/2013 10:10 AM
the real answer is much simpler: randomness + small sample size
11/13/2013 3:57 PM
If 10 games aren't enough to normalize stats, playing a guy 1/3 of his minutes makes it even harder to see normalization.

Add to that that high cap leagues typically have better defenses as well.

11/15/2013 8:22 PM (edited)
I'd like you to consider this..1) There's "Standardization", this is wrt the norms of the NBA. Say 90% of the teams in the history of the NBA have fallen between 22-34 assist per game, then anything outside this range will be an outlier and a bias is put on that player/team. 2) "Normalization" (bias) is the means by which you achieve Standardization. Standardization is nescessary, or we would look like the Harlem Globetrotters.
Then there's Standardization within range wrt the shape of the curve within the range. In short the league is made into a bell shaped curve wrt league/nba mean. This is the tough part to get right and where our beefs lay. The system will adjust the league mean to nba standard and then adjust the league to a bell shaped curve about the mean. This has the effect of lowering the top end and raising the bottom end thus creating parity. Fair competition is based on 'disparity' wrt the participants. So a system that creates 'parity' seems counter-productive.
11/15/2013 8:16 PM (edited)
Then there's "Correction" to predetermined values. FGA per game @ uptempo is about 95fga/g, if you had a few 105fga/g games..there's a correction to 95fga/g coming your way.  If your player is shoot a high FG% wrt RL..you have a Correction coming your way sooner or later. IMO sparky uses Corrections to generate the "Randomness"
11/15/2013 8:52 PM
Lastly, when the shape of the bell curve approximates a square..parity is king. These are parameters that are controlled by admin, and right now the freak'n bell is a box.
11/16/2013 8:38 AM
Exactly - there is far, far too much game-to-game variant programmed.  That too large variant is what causes huge outliers and corrections. There shouldn't be 60 point swings and certainly not 50 rebound swings, when teams don't touch their settings.  This huge variant also allows many of the other things (too many 3s, too high a 3 %, too many TOs, too many / too few FTA, etc).  Simple fix, but it would mean WIS would need to give 5 seconds to the NBA SIM, and we all know that isn't happening.
11/16/2013 1:05 PM
Posted by smokey57 on 11/15/2013 8:16:00 PM (view original):
Then there's "Correction" to predetermined values. FGA per game @ uptempo is about 95fga/g, if you had a few 105fga/g games..there's a correction to 95fga/g coming your way.  If your player is shoot a high FG% wrt RL..you have a Correction coming your way sooner or later. IMO sparky uses Corrections to generate the "Randomness"
This is not true.
11/16/2013 5:32 PM
The correction thing is kind of a joke, smokey.  It's not programmed to happen... but you expect it to happen because of the high number of outliers that effect the outcome of games.  The site staff would call it normalization... but it takes too long to occur.

11/16/2013 9:56 PM
Watch fga/g, to/g & pf/g, these 3 catagories correct very quickly. The fga correct wrt tempo and to's & pf's to team average. If you plot these 3 stats over the 82 games you will clearly see the correction to mean. The sim corrects almost immediately on these catagories. I had a team go 105fga in the pm today, in the am or tomorrows pm I'll see a sub 90fga game.
11/16/2013 10:06 PM (edited)
Normalization is a nice way to say bias. Usually to mean or a predetermined value as with tempo.
11/16/2013 10:23 PM (edited)
or the stats you draft - admin's stance is that all other considerations aside a 50% shooter will eventually shoot 50% given enough attempts - that's what they mean by 'normalization', flipping a coin enough times, that's all

and has admin ever used the phrase 'correction'? I feel like that's more a user coined term for some reason
11/16/2013 10:20 PM
My gripe is this, normalization is being used to force fit the leagues, and the shape if the bell curve is way too flat (parity). admin need to adjust the fit to a less flat (disparity) curve. Then we can see which owners have better skill sets. Right now it's 3 parts skill and 7 parts luck.
of 4
All Forums > Discussions in SIMology

Popular on WhatIfSports site: Baseball Simulation | College Basketball Game | College Football Game | Online Baseball Game | Hockey Simulation | NFL Picks | College Football Picks | Sports Games