[EarlPurple]

---

The Lin posting tool isn't working for me so it's here:

https://www.bridgeba...SA|pc|SJ|pc|S5|


Ok, so having a look at how a couple of players managed MASSIVE percentages in the Daylongs (do they have several accounts so they have already seen all the boards?). In any case I came across that one where a player made 3NT+4 for a huge score and had a look at the play.

It looks clear to me that West who is getting squeezed in 3 suits on the run of clubs will have to yield a trick in hearts but at the end you need to play the final card from dummy (ace of hearts) and so North has to go down to a single diamond which means he'll have to overtake the queen with the ace and partner can hold the suit.

Ok, that may be tricky for an average human in the East position who doesn't realise the value of the 7 of diamonds but we're talking about a robot here who is not expected to miscount things, and should be able to work out that the spades in hand are useless and covered by partner but he'll be squeezed if you don't guard the diamonds.

[pilowsky]

Seems to work ok.

---

[smerriman]

EarlPurple, on 2022-July-21, 03:58, said:

We're talking about a robot here who is not expected to miscount things, and should be able to work out that the spades in hand are useless and covered by partner but he'll be squeezed if you don't guard the diamonds.

Even with the code that GIB has built in to allow fluctuations, when I ran this through an older version of GIB, it generated 290 possible deals and in every single one of them South had AK of spades for its 3NT bid. It therefore makes no difference what card it plays, and chooses one at random.

GIB is well known to trust the bidding too much.

Yes, you'd think it'd be very logical when down to so few cards left to try dealing all distinct hands, even if you give the other ones a very low weight to break ties only. But it doesn't.

[EarlPurple]

Clearly needs improvement then. At that situation there are so few tricks left it is possible for GIB to calculate on all the possible holdings, and none of them will favour the diamond discard, some will break even, and in the case of the actual hand held, will show GIB must discard the spade.

Presumably we are posting these as a constructive way of getting the developers to improve GIB.

[EarlPurple]

Oh I see on the handviewing I just have to use the hv tag and put in the whole URL. I don't need to use any Excel converter.

Does it work if you insert it as media?

[smerriman]

EarlPurple, on 2022-July-21, 06:14, said:

Presumably we are posting these as a constructive way of getting the developers to improve GIB.

The last GIB update was 3.5 years ago, and the last update that involved its play logic was 8 years ago (with that only being forcing it to cash an ace against 7nt). So I wouldn't hold your breath

[thorvald]

smerriman, on 2022-July-21, 05:28, said:

Even with the code that GIB has built in to allow fluctuations, when I ran this through an older version of GIB, it generated 290 possible deals and in every single one of them South had AK of spades for its 3NT bid. It therefore makes no difference what card it plays, and chooses one at random.

GIB is well known to trust the bidding too much.

Yes, you'd think it'd be very logical when down to so few cards left to try dealing all distinct hands, even if you give the other ones a very low weight to break ties only. But it doesn't.

Without ♠A or K, how could West have a T/O double of 1♣?

So is the bot trusting opponents bidding more than partners?

[smerriman]

thorvald, on 2022-August-17, 07:56, said:

Without ♠A or K, how could West have a T/O double of 1♣?

So is the bot trusting opponents bidding more than partners?

It has West holding the ♦A instead of the ♠A. Both are the same from the T/O double perspective, just that the former gives South the stops he promised.