Gib algorithm bugs Gib fails in both defense and play
#1
Posted 2014-September-29, 09:58
#2
Posted 2014-September-29, 12:59
Note that the version of GIB in the download client is several versions older than the version used by the web client. This mostly affects bidding, but there are side effects on declarer play, because its simulations are based on hands that are consistent with the bidding.
#3
Posted 2014-September-29, 13:03
GIB's play is actually worse than OP states... after the spade lead and clearing the hearts, Declarer should be able to count 4 spades, 5 hearts, 1 diamond and 2 clubs for 12 tricks. No finesses (other than the one given at Trick 1) are necessary. South should not throw winning spades from dummy as he draws trumps.
#4
Posted 2014-September-29, 14:37
Bbradley62, on 2014-September-29, 13:03, said:
GIB's play is actually worse than OP states... after the spade lead and clearing the hearts, Declarer should be able to count 4 spades, 5 hearts, 1 diamond and 2 clubs for 12 tricks. No finesses (other than the one given at Trick 1) are necessary. South should not throw winning spades from dummy as he draws trumps.
this is %percent innocent compared with the mentioned case. if u can sort my hands according to score unfortunately I m not able to , doubled point is 100 or 200 according to vulnerebality.i strongly recommend you to review it . you will laugh a lot.
#5
Posted 2014-September-29, 14:38
0 luca, on 2014-September-29, 14:37, said:
contract is equal or a above 5
#6
Posted 2014-September-29, 14:47
#8
Posted 2014-September-29, 15:05
#9
Posted 2014-September-29, 15:07
0 luca, on 2014-September-29, 15:04, said:
It's probably a language barrier issue, but the word "finesse" does not belong in this conversation at all... simply discard ♦J on your winning ♥Q on Trick 2, or later discard it on ♠Q.
#10
Posted 2014-September-29, 15:08
Bbradley62, on 2014-September-29, 15:05, said:
You found it and I said ok.
#12
Posted 2014-September-29, 15:18
Bbradley62, on 2014-September-29, 15:07, said:
another issue when I rent gib from client and use it in ipad which gib version am I using ?Client or web or mobile?
#13
Posted 2014-September-29, 16:45
In any case, if you want better GIB play you had better use the "advanced" robots, the more expensive $1/day ones. The first hand BBradley pulled to me is the classic limitation of double-dummy based algorithm; from DD point of view you don't have to take 100% spade pitch, since guessing which minor suit finesse to take (or squeeze East if neither finesse works) is also 100% DD, so it has no reason *not* to play a spade on a heart, did so randomly, and eventually traps itself into having to guess the minor suit position. Supposedly the advanced bot should avoid these kind of things, having a true single-dummy engine kicking in after some number of tricks. The cheap ones are lobotomized for speed and taking up less server resources, and don't use the single-dummy algorithm.
The second one appears to me an instance of where the bidding perhaps convinces GIB that East *has* to have DK for the 2h bid for some reason, and thus views the finesse as also 100% compared to just cash the SQ, since maybe all hands in its sample East has DK. This goes to a long standing complaint that GIB should be more flexible about the hands it gives the opps in its sample, not being too strict about the opps having their bids, since maybe the superior play will emerge if you move to a more relaxed constraint on the sample generated.