BBO Discussion Forums: An actually simple improvement to GIB? - BBO Discussion Forums

Jump to content

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

An actually simple improvement to GIB?

#1 User is offline   smerriman 

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

Posted 2022-May-04, 19:39

The following hand was posted in BridgeWinners recently:


Upon testing, Advanced GIB pauses to run some simulations over 4.. before either passing or doubling (and double is clearly not expecting partner to take it out, since it'll pass as South too). It will never bid anything else.

With a 13-card solid spade suit, advanced GIB.. opens a weak 4. OK, we can probably set that one aside, since there is a flag in the bidding database where GIB is only allowed to run simulations for certain later bids, to prevent it from being super slow every bid.

But have South bid like this:


It pauses after the 3N bid to run simulations.. and bids 6?

It appears as if GIB is only considering bids that are *defined* in the database - ie ones that basic GIB may make with some hand.

Couldn't this be improved by allowing GIB to make 'undefined' bids? Assume everyone else will pass (which is how undefined bids work anyway - could even alert the bid as 'to play') and see what works best.

The *only* downside is that it will take some more processing time - you need to test each simulated deal in all 5 strains, whereas the original simulation may have only involved a couple.

But isn't getting to the right contract worth one bid taking a little bit more time than normal? (And aren't some double dummy analyzers optimized so that calculating all 5 results is simpler than doing 5 unrelated calculations? And if not, restricting it to suits you have some combined length in is probably fast enough already..)
1

#2 User is offline   pilowsky 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,620
  • Joined: 2019-October-04
  • Gender:Male
  • Location:Israel

Posted 2022-May-05, 02:47

I'm not sure what you mean by "either passes or doubles".
On the bidding table (basic GIB?) it only doubles - I only tried 3-4 times.
Also there's no detectable pausing - I suspect that the interface adds some pausing to make humans feel better.
If I put 4 (advanced or basic) robots alone at a table they'll bang out dozens of hands - bid and play - in the blink of an eye (milliseconds).



It doesn't seem to cope with aggressive bidding either:



On a regular bidding table this happened.
With a K-R score of 29.4



But never doubles this. With a K-R score of 29.5


K-R from http://www.rpbridge....cgi-bin/xhe1.pl
0

#3 User is offline   smerriman 

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

Posted 2022-May-05, 03:02

View Postpilowsky, on 2022-May-05, 02:47, said:

I'm not sure what you mean by "either passes or doubles".
On the bidding table (basic GIB?) it only doubles - I only tried 3-4 times.
Also there's no detectable pausing - I suspect that the interface adds some pausing to make humans feel better.

Basic GIB doubles unless favorable, when it passes.

Advanced GIB simulates hands at specific times during the auction, so can come up with different results each time (based on the seed as discussed previously). That's why it passed in the originally posted case, but doubles at other times. There is always a noticable pause when GIB is simulating (as it takes time to calculate double dummy results, vs the database lookup being near instant).

What I'm talking about here is that it isn't analyzing the right alternatives when a simulation is taking place.
0

#4 User is offline   pigpenz 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 2,553
  • Joined: 2005-April-25

Posted 2022-May-08, 15:26

Oh boy GIB
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