BBO Discussion Forums: GIB not taking out takeout doubles - BBO Discussion Forums

Jump to content

Page 1 of 1
  • You cannot start a new topic
  • You cannot reply to this topic

GIB not taking out takeout doubles

#1 User is offline   Stephen Tu 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 4,098
  • Joined: 2003-May-14

Posted 2012-July-03, 14:26

GIB to me passes way too aggressively on takeout doubles of opening preempts and preemptive overcalls, letting opps doubled into game w/ overtricks rather than taking a smaller minus by bidding on. I'm often on the receiving end of this largesse but this time was for the bad guys.



Why pass with one only trump trick, and a reasonable suit to bid? Is it the default to bid after these takeout doubles and being overridden by sim? Shouldn't the takeout double show more than just "3+ clubs", and also show 3+ H, 3+D, and 2- spades? Maybe for sim purposes it should assume 1- spade?
0

#2 User is offline   jamegumb 

  • PipPipPip
  • Group: Full Members
  • Posts: 94
  • Joined: 2010-August-20

Posted 2012-July-03, 15:22

This is a frequent complaint. My main question regarding GIB's simulations:

Does the bot base its decision on which outcome is more frequently correct? Or based on the expected outcome of each outcome?

That is, let's say the bot simulates this 100 times, of which 60 times 3S and 4H go off 1, and 40 times 3S and 4H make. It's right in this case to pass 60% of the time, but it's such a spectacularly wrong decision when incorrect that you'd still want the bot to pull the double.
0

#3 User is offline   Stephen Tu 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 4,098
  • Joined: 2003-May-14

Posted 2012-July-03, 16:30

View Postjamegumb, on 2012-July-03, 15:22, said:

Does the bot base its decision on which outcome is more frequently correct? Or based on the expected outcome of each outcome?

That is, let's say the bot simulates this 100 times, of which 60 times 3S and 4H go off 1, and 40 times 3S and 4H make. It's right in this case to pass 60% of the time, but it's such a spectacularly wrong decision when incorrect that you'd still want the bot to pull the double.


You'd want to weigh by expected outcome at IMPs/total points but by frequency if it's matchpoints. GIB does factor in form of scoring so I'm pretty sure it gets that right. I think it's passes too often probably because it has bad inputs into the simulated hands, not having representative sample of what doubler has and what preemptor has (here it says 3s only promises 6+ spades; shouldn't it expect 7+ spades?). Either that or maybe bad definitions of the continuations?
0

#4 User is offline   Bbradley62 

  • PipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 6,542
  • Joined: 2010-February-01
  • Gender:Male
  • Location:Brooklyn, NY, USA

Posted 2012-August-10, 15:30

Another example. IMPs, btw.
I'm amused that I can have 7, but not 8, diamonds with the re-opening double.
0

#5 User is offline   Quantumcat 

  • PipPipPipPipPip
  • Group: Full Members
  • Posts: 944
  • Joined: 2007-April-11
  • Gender:Female
  • Location:Bathurst, Australia
  • Interests:Archery, classical guitar, piano, watercolour painting, programming, french

Posted 2012-August-11, 05:40

If you have the minimum number of the other suits, 7 is the maximum number of diamonds (0 + 3 + 3 + 7 = 13 cards).
I Transfers
0

Page 1 of 1
  • You cannot start a new topic
  • You cannot reply to this topic

1 User(s) are reading this topic
0 members, 1 guests, 0 anonymous users