BBO Discussion Forums: Testing GIB's response algorithm to opening 2C - BBO Discussion Forums

Jump to content

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

Testing GIB's response algorithm to opening 2C

#1 User is offline   bobade 

  • PipPip
  • Group: Members
  • Posts: 27
  • Joined: 2008-November-26

Posted 2021-September-05, 22:01

There is a GIB bug which has never been fixed and nobody seems to care - the "cheaper minor" bid after 2C - 2D - 2M. I raised this as a topic on the GIB forum and was greeted with a big ho-hum. Because I find myself playing with a robot partner pretty often (Daylong tournaments), I'm hoping for some partnership understandings. Not just the expectation that GIB will never be perfect. Sheesh! . . . Neither will I.

My plan was to test GIB on Robot Challenges, opening 2C on every hand, rebidding 2M, and compiling a series of hands where GIB bids weaker minor. Trouble it, that auction sequence did not turn up in 32 challenge hands. There must be a better way.

Is there a way on Bridge Base to set up a testing platform where I can set criteria an have GIB play a bunch of hands? I might specify opening human as having 22-25 HCP with a 6 card major, and his/her robot partner having less than 11 HCP and no suit longer than 5C. After 100 hands like that, I think we would have an idea of what algorithm GIB is using in order to choose a "weaker minor" rebid.
0

#2 User is offline   smerriman 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,720
  • Joined: 2014-March-15
  • Gender:Male

Posted 2021-September-05, 22:49

You can use the bidding table found in the Practice section. GIB can be placed in as many seats as you like (no cost), and if you tap the menu icon and go to Deal source, you can input constraints on the deals (usually via the advanced tab).

But as I described in the other thread, what it's doing is pretty well known, and I've already done this analysis other times this bug has been commented on in the past; it bids cheaper minor both when it's appropriate, and when it has a natural bid in that suit.

The latter is a bug, but bugs haven't been worked on for many years now.

Example code for the advanced tab that shows this bug in action:

condition hcp(south)>=23 and spades(south)>=6 and spades(north)<3 and clubs(north)>=5 and hcp(north)>=5


It'll throw in 2NT for some hands, and 3 for others; you can figure out the exact point range and shapes if you like, but there's not much you can do about it as its partner.
0

#3 User is offline   bobade 

  • PipPip
  • Group: Members
  • Posts: 27
  • Joined: 2008-November-26

Posted 2021-September-08, 18:38

smerriman, thank you so much for the great information about testing GIB's performance using the Practice bidding table. Your example code was particularly helpful.

After playing with this for a while (starting with the auction 2C - 2D - 2S - ?), I have made the following observations (not statistically tested):

1) When GIB makes a "cheaper minor" 3C bid, the bid explanation may just say "cheaper minor, forcing to 3S" or "cheaper minor 11- HCP, 12- Total Points, forcing to 3S." This happens with weak hands and with strong hands, and is unpredictable.

2) As you mentioned, when GIB has a hand with 5 or more clubs and 5+ HCP, it will bid 3C and alert it as cheaper minor (using one of the two descriptions.) This is a bug.

3) GIB is very aggressive counting points, counting a hand with 4HCP, no 5 card suit, and 1 -2 cards in openers suit as 5+ points.

4) Therefore, after 2C-2D-2S it will bid 3D or 3H with a 4 card suit (headed by an honor) and 4 HCP.

5) The one good thing I can say about GIB: with 3 card support, it will raise openers suit after 2C-2D-2S rather than bid its own suit.

My conclusions:

1) Though I didn't test GIB after 2C-2D-2N, I think (based on playing experience) that sequence is probably safe, as would be 2C-2D-3N. So opening 2C with 22+ HCP and a balanced hand is constructive. (But what to do when GIB responds 2N to 2C? I didn't evaluate whether GIB responds logically to 4N quantitative or plays 4C as Gerber.)

2) If you have a long suit and want to open 2C, the suit had better be excellent, able to play opposite a stiff, and you better be near game in hand, because there is a good likelihood that you will never be able to tell what partner is bidding on until you see the dummy.
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