Loading ...
Sorry, an error occurred while loading the content.

5106Re: [hockeydisk] update on new disk

Expand Messages
  • D. Atkinson
    Sep 28, 2013
    • 0 Attachment
      The sim only uses disk info and does not in any way consider the stats of what's been run. The injury calculation each game is independent of all past games.
      The inj rating brings injury length into account, as well as frequency. And in my opinion relies on length too much. So for Shore, instead of getting hurt 4 or 5 times for 5 or so games each time, he will likely get hurt once or twice in a season at most, with one of those being big (30+ games). I would prefer the sim put in more frequent but shorter injuries on the high rate guys.

      On 9/28/2013 9:37 AM, Joe DeAngelis wrote:
       
      Hi Dave,

         I think I understand what you mean about how the simulator works, but I'd like to make sure I have it right.

         For example, let's say Eddie Shore only played 6 out of 82 games last season.  His injury rating is ridiculously high.  My league doesn't have usage limits, and he's played in all 50 of my team's games so far this season.  When we sim game 51, the simulator has no knowledge that Shore has played in 50 games already, and therefore only uses the injury rating to determine whether he will be injured.
         This same idea I would presume is correct for how the simulator handles game events in general...

         (I know the simulator's algorithms are far, far more complex than this...but I want to make sure I have this one basic point right...)

         To wrap those into a single question, then: Is it safe to say the simulator only uses "the disk" information to simulate a game, and does not in any way consider the cumulative totals of the season being run? 

      Thanks,
      Joe


      From: D. Atkinson <boomhound@...>
      To: Bill Corfield <corfieldb@...>
      Cc: hockeydisk@yahoogroups.com
      Sent: Friday, September 27, 2013 4:14 PM
      Subject: Re: [hockeydisk] update on new disk



      No. The skaters should have a 1.7x chance of getting injured over a season. Which should scale, since the injury rating is against the season GP (48), not the extrapolated GP. Should be good.

      On 9/27/2013 4:04 PM, Bill Corfield wrote:
      For example, our league doesn't track any usage except goalie usage...

      So, we would do nothing to the disk with regard to skaters whatsoever? Wouldn't every skater have a massive injury rating/likelihood? 

      Sorry for all the questions...

      Bill Corfield


      -----OriginalMessage-----
      From: D. Atkinson <boomhound@...>
      To: Bill Corfield <corfieldb@...>
      Cc: hockeydisk <hockeydisk@yahoogroups.com>
      Sent: Fri, Sep 27, 2013 7:02 pm
      Subject: Re: [hockeydisk] update on new disk

      This means to run the disk as is, with 48 games played, but calculate usage away from the disk.
      So, use the disk as is, but, for example, for Vinnie Lecavalier, who played 39 games, handle his usage (GP limits) as 39x1.7, or 66 GP. Do usage limits offline, not on the disk in the GP field.


      On 9/27/2013 3:56 PM, Bill Corfield wrote:
      What does this mean exactly?
      What I would recommend is to scale usage offline, using the 1.7x GP guideline, but don't change it on the disk


      Bill Corfield


      -----OriginalMessage-----
      From: D. Atkinson <boomhound@...>
      To: Bill Corfield <corfieldb@...>
      Cc: hockeydisk <hockeydisk@yahoogroups.com>
      Sent: Fri, Sep 27, 2013 5:56 pm
      Subject: Re: [hockeydisk] update on new disk

      I'm debating whether to issue a "pro-rated" disk or not. Not all ratings scale linearly in the game, so a simple 1.7x won't solve all issues. Definitely need to take all relevant stats up the 1.7x factor, just moving GP up will give crazy results. I'd like to test a 1.7x translation to ensure no wild behavior, but that would take time, and I know people want to get on with their seasons. What I would recommend is to scale usage offline, using the 1.7x GP guideline, but don't change it on the disk. Changing it on the disk will effect results, as many are proportional to "per game" calculations.
       

      On 9/27/2013 2:52 PM, Bill Corfield wrote:
      Thanks for the update Dave...

      A question many of us are pondering is how to best use the new disk, which is based on the 48 NHL season, for a full 82 game apba schedule. 

      Some ideas I've heard include multiplying all of the counting stats and GP by 1.7 to bring it up to a full 82 game season. Another idea is to just multiply the GP stat, and leave everything else alone. 

      Would you please comment in detail on what your best recommendation is for a league wanting to use this new disk for a full 82 game schedule? 

      Your efforts are very appreciated. Thank you!

      Bill Corfield/CCHL


      -----OriginalMessage-----
      From: D. Atkinson <boomhound@...>
      To: hockeydisk <hockeydisk@yahoogroups.com>
      Sent: Fri, Sep 27, 2013 5:11 pm
      Subject: [hockeydisk] update on new disk

       
      Currently at the very end of the testing period. Got to a conversion
      point that was acceptable, now just cleaning up some loose ends (like
      goalie assists). Hope to migrate to combined disk over the next 4-5
      days, and get the draft disk out to the team for error checking. If all
      goes well. Hope to release final disk by Oct 8, barring any major error
      finds.
      I really wanted to get this year's disk out earlier than in the past,
      and with Ken's great new rating vote system, I had the rating info much
      earlier than past years. However, you can blame Bettman and the lockout,
      as the short season made the testing difficult. The variabilities on the
      numbers are much higher than past seasons, so getting convergence on the
      disk has been a headache. I can't say that I'm thrilled with the disk
      this year, it is reasonable, but not spectacular, mainly because the
      variability numbers on a short season are just too large. Extra testing
      was needed just to ensure that changes weren't random variations of a
      small sample size (read: short season) or real effects. I did the best I
      could, and it's a good disk (but probably not a great disk). Sorry.
      Blame Bettman. We won't have this problem again now for a few years, so
      we'll let it go and move on. I'll you all know how progress is going and
      give you a solid fix on the release date in a few days when I finish the
      migration. Hang tight, we're getting close.
      Thanks!

      d








    • Show all 10 messages in this topic