[heart76]

barmar, on 2018-December-19, 09:47, said:

The robot probably assumed there was a reason Nigel played high, and this affected the types of hands that were used in the defense simulations.

It's a common quirk of GIB defense that the choice among irrelevant spot cards can have effects like this. OTOH, it rarely falls for falsecards that try to keep it from giving partner a ruff (but not never, about once a week I've been able to trick it).

There is no falsecarding involved here.
And this "common" quirk shifts 10 IMPs out of nowhere...
What GIB knows before playing the cards is that dummy does not have a spade stop.
So what you're saying is that there is a possible scenario, based on a random card played from dummy at trick 1 in spades, where it is MORE LIKELY to win by giving away your chances in that suit and give away your only entry, the HA.
What is this scenario at IMPs? It has to involve his partner holding QJxxx in H, but then shouldn't you play the H10 from A108 and retain your entry as a plan B?

It would be nice if you could compare with the basic GIB and anyway try to fix it, because this does look like a bug

[cyrenman]

Group B

cyrenman 70 - 13 sep123

https://webutil.brid...7331&u=cyrenman

[barmar]

heart76, on 2018-December-19, 10:33, said:

There is no falsecarding involved here.

I didn't say there was any falsecarding here, I was talking about playing from declarer's hand and trying to make it look like declarer is short in the suit when it's actually the other defender.

The simple fact is that GIB's defense is based almost entirely on simulations. It deals out random hands consistent with the auction and play so far and performs double dummy analysis. Playing different cards results in different hands being chosen in the simulations.

Because of time constraints, it can only simulate a few dozen different hands at each turn to play. So a small difference in the hands that are chosen can easily result in a different decision being made.

What I've seen when analyzing things like this is that it determines that in most of the hands, it doesn't matter what it does at that trick, they all lead to the same result. But there will be one or two outlier hands where it makes a difference, and that determines the choice. And those outliers may be different depending on the card that was played.

If anything, this is a flaw in the basic design of GIB's defense algorithm. We'd have to do significant redesign to fix it, and we don't have anyone competent to do this.

[mlbridge]

Mlbridge 20-37 versus Robot

https://webutil.brid...3150&u=mlbridge

[mlbridge]

I have a question about this hand:

https://www.bridgeba...S8%7Cpc%7CHQ%7C

How does one ask for aces after the 2 hearts response?

[mlbridge]

Regarding GIB defense and play, this was an interesting declarer play by GIB from my match:

https://www.bridgeba...h8%7Cpc%7Cck%7C

I was in 3 spades going down and GIB was in 4 spades making. The opening lead was a 4 of diamonds and GIB played the 2 from dummy!

Not sure what kind of logic or simulation the GIB was using. But I am not sure any human would have play the 2 of diamonds.

[smerriman]

mlbridge, on 2018-December-19, 12:59, said:

I have a question about this hand:

https://www.bridgeba...S8%7Cpc%7CHQ%7C

How does one ask for aces after the 2 hearts response?

3♥ then 4NT next time, I guess.

[mlbridge]

smerriman, on 2018-December-19, 13:25, said:

3♥ then 4NT next time, I guess.

Thanks.

I can never figure out this bidding sequence.

I also keep forgetting the GIB works on total points and not HCP. I keep thinking GIB have 17+ hcp. :-)

[Elyk25]

Elyk25-Gordontd

55-1

https://webutil.brid...135039&u=elyk25

[heart76]

barmar, on 2018-December-19, 11:55, said:

Because of time constraints, it can only simulate a few dozen different hands at each turn to play. So a small difference in the hands that are chosen can easily result in a different decision being made.

What I've seen when analyzing things like this is that it determines that in most of the hands, it doesn't matter what it does at that trick, they all lead to the same result. But there will be one or two outlier hands where it makes a difference, and that determines the choice. And those outliers may be different depending on the card that was played.

I think you realize that this cannot be the case here, as with one side entry only with a singleton in dummy it does always make it worse for you if you let it go.

Quote

If anything, this is a flaw in the basic design of GIB's defense algorithm. We'd have to do significant redesign to fix it, and we don't have anyone competent to do this.

You have my full respect and appreciation for the work you do. However, FWIW if I were the TD I'd take the consequence of it and have a policy to amend the results of those few horrendous outliers - not to be misunderstood, this of course applies to any player in tournaments where the blunder counts.

[Elyk25]

Elyk25-GIB

23-23

https://webutil.brid...255979&u=Elyk25

[shuba]

Shuba GIB -12
Shuba Natalie +21

Links will be laiter

[cyrenman]

Group B

cyrenman 28 - 12 Gerardo

http://webutil.bridg...aeb4-1545197345

[xenoquatz]

Group C
http://webutil.bridg...noquatz&v3b=app

Xenoquatz 46-26 Povratnik

[cyrenman]

Group B
cyrenman 40 - 29 olegru
http://webutil.bridg...aeb4-1545197368

[sfi]

sfi 21 - 18 heart76

https://webutil.brid...545217155&u=sfi

[gordontd]

gordontd - xbabarx
37 - 23
https://webutil.brid...1717&u=gordontd

[derppp]

derppp 42-25 mlbridge

http://webutil.bridg...pp&v3b=app#row-

[Icycookie]

icycookie - barmar 33 - 11
https://webutil.brid...168&u=Icycookie

[cyrenman]

Group B
cyrenman.15 - 17 ranmit
http://webutil.bridg...nman&v3b=appHas