[158FC79]

GIB plays against his human partner intentionally,
please refer to board number 3 in history or number
6 ( as I played ) on 2025-09-19 7:40

E-W were playing out 3 nt and when N-S were having 4 tricks already and two sure tricks in hands of AI,
north unexpectedly will not take them and letting E-W complete 3 nt! Why BBO let it happen, are they thinking that it is very entertaining and gives peculiar twist to tournaments with AI? I will not recommend anybody to rent AI or pay to use it in any form

[pescetom]

It would have helped to provide a link

But the incriminated play is trick 6 here
https://www.bridgeba...DK%7Cpc%7CDJ%7C

[steve2005]

Gib when it has determined that 2 cards are equally correct will always choose the smaller card
So, either the TEN♠ or 2♠ are equal as GIB has complete confidence that you will find a diamond switch to GIB's ACE♦
so it plays the 2

[158FC79]

how do I know what ace there: diamond or heart? What it gives? The result is the same: two down, north must immediately take two tricks, what is the reason to permit me to take spade trick so after I must guess which ace is there

[158FC79]

I do not know how provide a link

[smerriman]

158FC79, on 2025-September-19, 08:11, said:

Why BBO let it happen, are they thinking that it is very entertaining and gives peculiar twist to tournaments with AI? I will not recommend anybody to rent AI or pay to use it in any form

158FC79, on 2025-September-19, 08:11, said:

how do I know what ace there: diamond or heart? What it gives? The result is the same: two down, north must immediately take two tricks, what is the reason to permit me to take spade trick so after I must guess which ace is there

GIB has zero AI in any form. All it does is look at possible hands that players may have that match the bidding, analyse the results double dummy, and choose one of the cards that gives the best result on average. Double dummy of course assumes you will make the right play, even if you have no way of knowing what that is..

Though in this case it's clearcut that West has the heart ace; with the diamond ace declarer would have crossed to dummy to take a club finesse.

(And it's not playing low thinking its card doesn't matter, it's playing low because it thinks there is a small chance that declarer has really stupidly left you with club winners that you really stupidly forgot to cash and wouldn't be able to if it overtakes. Again, zero AI, just pure double dummy numbers..)

[158FC79]

pescetom, on 2025-September-19, 10:59, said:

It would have helped to provide a link

But the incriminated play is trick 6 here
https://www.bridgeba...DK%7Cpc%7CDJ%7C

What you think, if there was no diamond Ace at north, would GIB the same permit me to have trick number 6 and let them make 3 nt?

[158FC79]

Idiotism, If south have the rest tricks in club suit then how opener intend to make 3 nt? Spades gone, one trick in spades; presume opener have all heart tricks- say 4 tricks, 5 so far; one trick in clubs, where the rest? I do not understand why it so preprogrammed not to take all tricks at once and go to such complications

[johnu]

158FC79, on 2025-September-19, 14:13, said:

how do I know what ace there: diamond or heart? What it gives? The result is the same: two down, north must immediately take two tricks, what is the reason to permit me to take spade trick so after I must guess which ace is there

Why doesn't GIB overtake and cash the winning diamond? Why doesn't GIB do any of a thousand different things. If GIB actually had AI, then it would know that letting partner keep the lead isn't nearly as good as overtaking and cashing ♦A. But GIB doesn't have any AI, and expects partner to play perfectly, which obviously doesn't happen, even when GIB is that partner.

As far as what ace to play partner for, that's a bridge question, and the answer IMO is you have to play partner for ♦A. Why? A very quick analysis shows that declarer discarded ♦9 on a spade trick. Declarer apparently has 6 clubs for his club play. If declarer had ♦A9(x), they would have tried to run diamonds after winning ♠A, and then taken a club finesse instead of trying to drop ♣K or whatever that club play was, since the opponents have at least 4 spade tricks and ♣K once they get the lead, under most circumstances.

So, with that reasoning, partner must have ♦A since declarer doesn't have it.

[158FC79]

How many seconds you have for analysing? 30 sec?
A good lot of boards go banana for lack of time, I can barely count cards when dismissed. What if north has no aces? Why should south consider that maybe north has D ace and gives me lead just to see how smart I am,
This is wrong approach, you must as quickly as pos
to clear the scene so partner must not think and be nervous

[smerriman]

I'm not sure what point you are making. Gib does bad stuff, not because it uses AI (as you claim), or because BBO think it is funny; it's because GIB uses the very simple algorithm outlined above that it has been programmed to use, nothing more. Such is life playing with a robot.

[158FC79]

Well,.maybe, BBO should do smth about it, improve software, I dunno, and I call it AI cause GIB is Ginsberg intellectual bridge player and on many sites it is play vs AI( robot ) or use another AI say latest computer bridge champ

[158FC79]

Idiotism, If south have the rest tricks in club suit then how opener intend to make 3 nt? Spades gone, one trick in spades; presume opener have all heart tricks- say 4 tricks, 5 so far; one trick in clubs, where the rest? I do not understand why it so preprogrammed not to take all tricks at once and go to such complications

[pescetom]

smerriman, on 2025-September-19, 23:56, said:

I'm not sure what point you are making. Gib does bad stuff, not because it uses AI (as you claim), or because BBO think it is funny; it's because GIB uses the very simple algorithm outlined above that it has been programmed to use, nothing more. Such is life playing with a robot.

I'm not sure what point *you* are making, except that the robot's failure to take it's tricks and set the contract has the (near incredible in obtusivity) technical reason described here.
BBO itself made the rather rapid switch from being a basically free for all bridge community to a for profit service renting robots at significant cost.
It can hardly be surprised if new customers are bewildered and resentful when the robot they paid to use makes a silly mistake.

[smerriman]

pescetom, on 2025-September-20, 14:50, said:

I'm not sure what point *you* are making, except that the robot's failure to take it's tricks and set the contract has the (near incredible in obtusivity) technical reason described here.
BBO itself made the rather rapid switch from being a basically free for all bridge community to a for profit service renting robots at significant cost.
It can hardly be surprised if new customers are bewildered and resentful when the robot they paid to use makes a silly mistake.

My point is that the type of logic desired here is completely contrary to the fundamental Monte Carlo algorithm, and is not something that can be simply 'improved', but requires a completely different type of robot. The alternate approach of Ben and AI is not even remotely close to being competitive yet.

Of course, it wouldn't be difficult to "improve" the robot by hardcoding a check for "if you can beat the contract in your hand, cash all of your top tricks". That seems to be what the OP is suggesting, and it would work on this hand, but with a little more consideration it's not hard to see it would make for significantly worse scores in the long run.

Obviously it would be nice to have a better robot, but like the 'false alert' thread, something that appears easy to a human is often extremely difficult for a robot and vice versa..

[johnu]

smerriman, on 2025-September-20, 14:59, said:

My point is that the type of logic desired here is completely contrary to the fundamental Monte Carlo algorithm, and is not something that can be simply 'improved', but requires a completely different type of robot. The alternate approach of Ben and AI is not even remotely close to being competitive yet.

How close is Ben to GIB advanced these days? In Jan 2024 there was a comparison of results for various robots, and Ben was in shouting distance for being a pretty new program.

[158FC79]

All bullshit about robot giving me the lead to take extra trick in clubs!!! What you say! If south has extra trick in clubs- he takes it lately, opener must play out both clubs and hearts, cause there is no switch to dummy, so if south has J and 9 in clubs- he can take nothing, cause opener might have Q and 10, so robot must take all trick, give a lead to opener and see what happens! So I am completely sure it is a preprogrammed trick to harm human partner!!

[pescetom]

smerriman, on 2025-September-20, 14:59, said:

My point is that the type of logic desired here is completely contrary to the fundamental Monte Carlo algorithm, and is not something that can be simply 'improved', but requires a completely different type of robot. The alternate approach of Ben and AI is not even remotely close to being competitive yet.

Of course, it wouldn't be difficult to "improve" the robot by hardcoding a check for "if you can beat the contract in your hand, cash all of your top tricks". That seems to be what the OP is suggesting, and it would work on this hand, but with a little more consideration it's not hard to see it would make for significantly worse scores in the long run.

I think a hardcoded check of the top tricks could already do much to avoid user frustration, with limited downsides (which could be tuned by user configuration of n). Even just "take your top tricks if":
- they represent all the remaining tricks
- they guarantee our game at IMPs
- they guarantee our game plus at least n overtricks
- they set opponent's game at IMPs
- they set opponent's game and guarantee at least n undertricks.

[158FC79]

Then I am right, if north take all the tricks then 3 nt is finished with 1 undertrick