[gwnn]

---

I very rarely post any cardplay hands as I know from first-hand experience how easy it is to muck it up, but why on earth would GIB choose the ♠T over a low one?

Sorry about the "bug" in the title. Maybe we cannot expect GIB to "get these right" but I think it's ok to at least hope.

[gwnn]

Maybe GIB "knows" that S has the ♠J from the bidding, and maybe it is very hard to program GIB to mistrust human bidding. I don't know.

[Stefan_O]

gwnn, on 2016-April-24, 14:35, said:

Maybe GIB "knows" that S has the ♠J from the bidding, and maybe it is very hard to program GIB to mistrust human bidding. I don't know.

I dont see any bid from South promising the Jack of spades.... and all of South's bids seem consistent with the bid-descriptions given.

Technically, of course, it COULD be that in NONE of the monte-carlo samples generated before playing the Spade ten, East happened to be dealt the Jack.
But doesnt Gib generate some 50 sample deals or more? If so, it seems extremely unlikely...

Certainly, this play looks very weird...

[Bbradley62]

Stefan_O, on 2016-April-24, 17:29, said:

But doesnt Gib generate some 50 sample deals or more?

This largely depends on whether it was a basic GIB or an advanced GIB, and only OP can tell us that (if, in fact, even he can) since South's name is Redacted.

[gwnn]

It was a BBO challenge. Those are I guess advanced robots?

I made the jack of spades comment because 3S showed "twice rebiddable spades." AK98xx maybe doesn't qualify but AK987xx maybe does?

I don't like complaining about cardplay, as mentioned above, but at the same time I do wonder sometimes if "well those were the hands it simulated, end of story" is the best we can hope for. There seem to be two big issues with GIB's cardplay currently:

-taking finesses when GIB had a claim. We know for a fact that GIB has the ability to adjudicate human claims so I think it should be able to adjudicate "auto-claims" as well. So if it does know how to test whether the hand is high, it should be able to figure out how to play only high cards in the last few tricks. Of course I know how hard "easy, one-afternoon scripts" can get to implement.

-it unblocks cards randomly, maybe because it doesn't understand "solving a guess" or because it blindly believes the human bidding sequence. This is possibly incredibly hard to program. And maybe it's not as easy as saying "if dummy is known to have only higher cards than I do, I should switch to my lowest card" - or maybe that would just fix one small case and not easy to generalize.

[Stefan_O]

gwnn, on 2016-April-25, 00:19, said:

-taking finesses when GIB had a claim...

-it unblocks cards randomly...

Have you posted (or can u post) some actual hand-records of this?
Would be interesting to have a look.
(haven't really noticed this as significant issues...)

[gwnn]

Stefan_O, on 2016-April-25, 01:43, said:

Have you posted (or can u post) some actual hand-records of this?
Would be interesting to have a look.
(haven't really noticed this as significant issues...)

There have been a few threads on this. I'm too lazy (or want to work?) to look'em up now. I will show you this evening, if nobody before me does.

[Bbradley62]

gwnn, on 2016-April-25, 00:19, said:

I made the jack of spades comment because 3S showed "twice rebiddable spades." AK98xx maybe doesn't qualify but AK987xx maybe does?

According to the documentation, "twice rebiddable" means 7+ cards, or 6 cards + 3 of 5 honors or more than 4 8421 points. So, even KJxxxx would qualify.

[gwnn]

Thanks, I should have known that myself. I was trying to cut GIB some slack but it seems this particular slack is uncuttable.

Any news on whether BBO challenges use basic or advanced robots?

[johnu]

I would bet that because the challenge matches are free, basic robots are being used.