10/30/2013 9:25 AM
With regards to the outcomes, this week I've been really able to dive into numbers. The high scores in DI are because the average difference in teams is higher than in DIII.  The way the system has been set up, it uses matchups to determine how a play progresses and the larger discrepancies are given too much power. Right now, I'm adjusting the numbers (chances of tackles, chances of completing a tackle, knocking down a pass, etc) at the higher levels of differences . Those almost exclusively favor the very strong team, and while that should happen more often it shouldn't be the best result every single play.  The good part about the matchup settings is that this does not affect teams that are closely matched.  

In addition, I'm getting a high number of completions in my test games as well. I'm changing the numbers on completions too, to fit with a more realistic model of how completions happen.  I'm also recalculating the base completion percentage using not just the QB rating, but the relative difference between the QB, the Receiver and the defender(s).

There are certainly elements of the engine that can be pulled out and scrutinized and I expect you all to do that as soon as you find something. Having said that, I feel the engine is in a pretty solid state, and with some minor number tweaks this thing will be humming along. 

Also, the reason that these types of changes did not work with my predecessors is because the structures that were used were hard to manipulate. I've worked the last three months to pull all of these numbers out into separate files that can be imported. This has allowed me to look at these numbers and adjust accordingly and revert if necessary. 

10/30/2013 10:42 AM
Oriole:  I did not really follow the development of the beta.  While there are many items that I'm not very happy about (nothing that is a complete non-starter), I will say that I feel like you are doing a pretty good job of addressing items like this as they arise.  I'm sure you are overwhelmed with comments and things to do, but taking a few minutes to communicate to us what you are doing, is worth at least as much as the actual work that you complete.


10/30/2013 12:12 PM
Posted by oriole_fan on 10/30/2013 9:27:00 AM (view original):
With regards to the outcomes, this week I've been really able to dive into numbers. The high scores in DI are because the average difference in teams is higher than in DIII.  The way the system has been set up, it uses matchups to determine how a play progresses and the larger discrepancies are given too much power. Right now, I'm adjusting the numbers (chances of tackles, chances of completing a tackle, knocking down a pass, etc) at the higher levels of differences . Those almost exclusively favor the very strong team, and while that should happen more often it shouldn't be the best result every single play.  The good part about the matchup settings is that this does not affect teams that are closely matched.  

In addition, I'm getting a high number of completions in my test games as well. I'm changing the numbers on completions too, to fit with a more realistic model of how completions happen.  I'm also recalculating the base completion percentage using not just the QB rating, but the relative difference between the QB, the Receiver and the defender(s).

There are certainly elements of the engine that can be pulled out and scrutinized and I expect you all to do that as soon as you find something. Having said that, I feel the engine is in a pretty solid state, and with some minor number tweaks this thing will be humming along. 

Also, the reason that these types of changes did not work with my predecessors is because the structures that were used were hard to manipulate. I've worked the last three months to pull all of these numbers out into separate files that can be imported. This has allowed me to look at these numbers and adjust accordingly and revert if necessary. 

This really scares me.  This is the first I've seen that teams at D1 have a greater difference than teams at D3.  Is this true?  On both an absolute and relative basis?  I routinely see 20 point differentials in D3 cores, which is way bigger on a relative basis compared to D1.  Are D1 teams really 20+ points apart on an absolute basis?  If they are, D1 has changed radically in 2.0.

If teams are 20 points apart in cores, I think the matchups should favor the better team almost exclusively.  I think the same thing should happen at 10 points.  At 5 points, I think the better team should win most of the matchups though not in such a dominant fashion.  It sounds like you are trying to throttle the game again using a less overt method.

IMO, fix passing.  This seems to be the root cause of the high scoring.  Weaker teams are passing better than they should.  Even better teams are completing passes at a higher rate than they should given the matchups.  If this doesn't fix scoring, then look into the next issue.  IMO, that next issue needs to be fatigue and STA and how that doesn't currently affect player performance.

Making gamewide changes to get real world results is not a good strategy.  Fix the engine and it should give you real world results when two relatively equal teams play each other at any level.  And keep in mind that there will always be more blowout games than in real life because there are uncoached and unrecruited teams and teams do not ease up in blowouts.  There are games in GD like Baylor or Oregon going full throttle for 4 quarters against overmatched opponents.   
10/30/2013 2:09 PM
Sorry Oriole I would have thought that these adjustments would have been made in the last 8 months, that's why we tested, not after v3 is rolled out. Obviously v3 IS NOT READY for human participation. But, it's probably working as intended.
10/30/2013 2:20 PM
Maybe Oriole is actually Kathleen Sebelius.
10/30/2013 2:27 PM
Any chance we could just revert to 1.0?
10/30/2013 2:32 PM
Did WIS have anything to do with Obamacare? They are running a close race to disaster.
11/3/2013 2:03 PM
Posted by oriole_fan on 10/29/2013 5:31:00 PM (view original):
Hello guys. I know some of the scores seem a bit artificial at this time. There are still some score influencing modifications that I'm in the process of correcting.  (example, there is a discrepancy in rushing numbers are inflated when attributes hit a certain difference. Also, QB are throwing at a high percentage completion based on the QB rating. I've been running various tests over the past few months, and I've honed in on a better model for how this is calculated. ) Besides that, the exhibitions scores do seem influenced by the punter/kicker issue as well. I'm also looking into creating a more realistic result from a backup kicker. 



 

A bit?  A bit?  Seriously, a bit?  That's like saying the A-bombs did "a bit" of damage to Hiroshima and Nagasaki.  
 

 

11/3/2013 2:06 PM
Posted by cebrake on 10/30/2013 2:09:00 PM (view original):
Sorry Oriole I would have thought that these adjustments would have been made in the last 8 months, that's why we tested, not after v3 is rolled out. Obviously v3 IS NOT READY for human participation. But, it's probably working as intended.
Bingo!
11/3/2013 5:25 PM
Have any regular season 3.0 games been played yet? I agree that the known bugs should have been fixed before this rolled.
11/3/2013 5:38 PM
Posted by jibe on 11/3/2013 5:25:00 PM (view original):
Have any regular season 3.0 games been played yet? I agree that the known bugs should have been fixed before this rolled.
They start on the 5th
11/3/2013 5:38 PM
Posted by jibe on 11/3/2013 5:25:00 PM (view original):
Have any regular season 3.0 games been played yet? I agree that the known bugs should have been fixed before this rolled.
First ones go through on Tuesday. I will say now that the QB percentage rate was fixed that the scores seem more realistic now. There is still a higher running success rate however. 
of 2

Terms of Use Customer Support Privacy Statement

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

© 1999-2014 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.