[tompeer]

Playing in 4♠ - makes easily but GIB went 1 down at every table...except mine where it threw the winning j♠ and went two down.

Was it something I did? How can I make it up to GIB

{comments}

---

[cloa513]

Fred gives GIB this big quota of bad play it has to do every month and you drew the short straw to help it fill that quota.

[ArtK78]

cloa513, on 2012-May-31, 07:47, said:

Fred gives GIB this big quota of bad play it has to do every month and you drew the short straw to help it fill that quota.

That is the only explanation I can come up with. Truly a remarkably poor play.

But I have seen quite a few remarkably bad things GIB has done in some robot tournaments lately. But these plays and bids tend to even out as you are playing against 2 GIBs and only with 1. So that means you should get a 67% game routinely.

[nathan2008]

yes... SJ is a real bug. Let us say "GIB miscliks SJ" lol. but for the play. GIB is playing west has DK+HA. So he wants to give an endplay to west.

[Stephen Tu]

There are two things going on here, that can perhaps help you understand why GIB screws up:

1. From its perspective, there's nothing "weird" about robot west playing K from Kt. Double dummy, playing the K from KT first trick doesn't really ever cost anything if north's trump suit is known. So it still sees trumps 2-2 as a significant common possibility,

2. If east had QTx, at this point it can't make the hand. Even had it played low, not J, East will drive out the club stopper. Now South can draw the last trump and exit a club, but then gets endplayed in diamonds and has to lead away from the HK. So GIB is catering to spades 2-2 only, and giving itself transport back to hand in trumps so it can take both red hooks. The Sj is an unblocking play, trying to make if spades 2-2, at the expense of an extra undertrick if they were 3-1. Maybe it felt the field was only in 3 and it had to make? I have no idea how it weights making contract vs. undertrick considerations at matchpoints.

I don't know if this should be described as an "easy make". One needs the foresight to either take first club trick in dummy, or take the diamond hook at trick 2 before touching trumps. I could see decent human declarers getting sloppy and going down.

[ArtK78]

Stephen Tu, on 2012-May-31, 12:29, said:

One needs the foresight to either take first club trick in dummy, or take the diamond hook at trick 2 before touching trumps. I could see decent human declarers getting sloppy and going down.

Really? Is it so difficult to see that, outside of the trump suit, the only entry to hand is the ♣K?

I expect better (Maybe that is why I am often disappointed in GIB's play).

[ArtK78]

tompeer, on 2012-May-31, 06:31, said:

Was it something I did? How can I make it up to GIB

Well, after you get ***** by GIB, you should send flowers. Perhaps some cuddling would be good.

[Stephen Tu]

ArtK78, on 2012-May-31, 13:33, said:

Really? Is it so difficult to see that, outside of the trump suit, the only entry to hand is the ♣K?

I expect better (Maybe that is why I am often disappointed in GIB's play).

Maybe given how fast GIB plays, playing hands at lightning speed well under a minute, we should give it a break when it occasionally gets sloppy.

I gave this hand to my home GIB set to "par mode", long thinking time/large sample analysis, and it plays it perfectly to make. Same with some other declarer play hands posted in this forum recently. I think generally GIB plays quite well as declarer (not well as defender because of signalling ineptitude), when given plenty of time to think. The question is do we really want the BBO GIBs to take 60-90 seconds on the first couple tricks (it gets faster deeper into the hand where the number of possibilities drop fast and it has already analyzed possibilities), or will humans get frustrated waiting for it to play. Also how much computer time per GIB process given the hundreds of people playing robots simultaneously does BBO server farm have to spare?

The download client at least has control for thinking time, but tourneys you are at the mercy of the online server defaults.

Personally, I'd rather get through a lot of hands fast, and have my partner robo-declarer occasionally throw a hand, than sit twiddling my thumbs waiting. If I wanted to wait that long, I'd go play live bridge.

[cloa513]

Stephen Tu, on 2012-May-31, 15:07, said:

Maybe given how fast GIB plays, playing hands at lightning speed well under a minute, we should give it a break when it occasionally gets sloppy.

I gave this hand to my home GIB set to "par mode", long thinking time/large sample analysis, and it plays it perfectly to make. Same with some other declarer play hands posted in this forum recently. I think generally GIB plays quite well as declarer (not well as defender because of signalling ineptitude), when given plenty of time to think. The question is do we really want the BBO GIBs to take 60-90 seconds on the first couple tricks (it gets faster deeper into the hand where the number of possibilities drop fast and it has already analyzed possibilities), or will humans get frustrated waiting for it to play. Also how much computer time per GIB process given the hundreds of people playing robots simultaneously does BBO server farm have to spare?

The download client at least has control for thinking time, but tourneys you are at the mercy of the online server defaults.

Personally, I'd rather get through a lot of hands fast, and have my partner robo-declarer occasionally throw a hand, than sit twiddling my thumbs waiting. If I wanted to wait that long, I'd go play live bridge.

All irrelevant, it is a lousy contract requiring nice break in spades 2/2 or K/Q singleton onside- any 3 card set of spade onside or KQx offside are death combinations and all finesses to work but order of cards to play is irrelevant (except play spades as soon as possible to avoid being ruffed against) (North always ruffs at some point in all side suits). Whether West might hold K10 is irrelevant as East gives it away by playing the QS (Defence of the contract we don't care about and anyway he'd have to play 10S otherwise).
All GIB declarer has to do to make the contract is not assume/simulate is the unlikely singleton A♥ offside and singleton K♦ offside and any play other than A♠ and low one (there is no combination of EW hands where any other play is right- a smoother play running the J♠ doesn't work if West has KQ8 of ♠ since West just covers the J♠ and declarer loses two spades.

All Souths who opened 1H deserve the result they got, its a crummy 1H and when your partner is poor at declarer play and opponents poor at defence- you want the best chance to play the hand and 1NT with transfers gives that.

This post has been edited by cloa513: 2012-June-01, 05:41

[barmar]

Was this a basic or advanced robot? If it's a basic bot, I'm not surprised at the ♠J play, it seems to do silly things like this a lot, perhaps because of the limited amount of thinking time we allow the bot so it ends up playing almost randomly, or maybe because of a bug in the double dummy solver.

I'd be much more concerned if you saw this with the advanced bots. Basically, if you're looking for something approaching intelligent play, you should rent the advanced bots (or play in robot tourneys, which use advanced bots).

[Bbradley62]

barmar, on 2012-June-01, 09:22, said:

Was this a basic or advanced robot? If it's a basic bot, I'm not surprised at the ♠J play, it seems to do silly things like this a lot, perhaps because of the limited amount of thinking time we allow the bot so it ends up playing almost randomly, or maybe because of a bug in the double dummy solver.

I'd be much more concerned if you saw this with the advanced bots. Basically, if you're looking for something approaching intelligent play, you should rent the advanced bots (or play in robot tourneys, which use advanced bots).

It was in Robot Duplicate (MP), #5583.

[barmar]

Sorry about that. I thought Uday once told me that hand IDs beginning with "M-" were from the MBC, I must have been mistaken.

[Bbradley62]

cloa513, on 2012-May-31, 15:42, said:

All Souths who opened 1H deserve the result they got, its a crummy 1H and when your partner is poor at declarer play and opponents poor at defence- you want the best chance to play the hand and 1NT with transfers gives that.

Did this come completely out of left field? How does any South open 1H when all Norths open 2S in first seat?

[Bbradley62]

barmar, on 2012-June-01, 10:05, said:

Sorry about that. I thought Uday once told me that hand IDs beginning with "M-" were from the MBC, I must have been mistaken.

Travellers beginning with "M" come from MBC, but this traveller starts with the tournament number: http://online.bridge...sername=tompeer

[barmar]

Ahh, I was confusing the traveler ID with the hand ID.