[MorK]

Hi,

GIB held:
105
AQJ9762
53
107

I opened first hand 1 NT VUL. Opp (GIB) overcalled with 2S (S+m), and GIB now found the 2 NT bid to make a drop to 3 H
7 HCP in a 7 card suit after 1 NT opening from PD is a drop?
Btw: I opened a 14 count with the bare King of Hearts and he still made 4 H (Good, the opp bid 3S after his 3H drop, so he decided to revalue his hand to 4H)

Sincerely Yours,
Felix Zimmermann

[barmar]

7222 is the worst shape when you have a 7-card suit. I'll bet it would have used a Texas transfer immediately if it had a singleton.

GIB also doesn't bid Capp with weak hands, so there's a good chance the defenders may be able to take 4 tricks pretty quickly.

Given all that, the hand may only be worth an invite, but I don't think GIB has an invitational sequence available.

[arigreen]

GIB doesn't have an invitational bid available. GIB simulated among 3H (game forcing) and 2N (planning to bid 3H to play). In my experiments, GIB bid 3H about 60% of the time and 2N 40% of the time. In many of the examples where GIB bid 2N, the opponents bid up to 3S and GIB bid 4♥ anyways.

While I would prefer to force to game with these cards, GIB's decision to sometimes bid 2N is not so unreasonable.

[barmar]

I wouldn't be surprised if you got a similar distribution of bids if you posed the hand to MSC.

[Jlall]

barmar, on Mar 2 2010, 03:21 PM, said:

I wouldn't be surprised if you got a similar distribution of bids if you posed the hand to MSC.

I would lay you odds that there was not a single MSC panelist who didn't force to game heh.

But w/e that is not the standard gib should be held to.

[Bbradley62]

arigreen, on Mar 1 2010, 03:40 PM, said:

GIB doesn't have an invitational bid available. GIB simulated among 3H (game forcing) and 2N (planning to bid 3H to play). In my experiments, GIB bid 3H about 60% of the time and 2N 40% of the time...

So, GIB won't always make the same decision when faced with exactly the same set of circumstances? Is that a desirable feature?

[barmar]

I have to admit that I was surprised the first time I saw a bunch of tables where the auctions started the same and then robots went in different directions. It's even more common in the play: you can have identical auctions, but different opening leads. The natual assumption is that computer programs will behave deterministically.

But bridge is a game of partial information, and guessing is sometimes required. If the robot is going to guess, why should it always guess the same thing?

Assuming GIB uses a decent sized sample, clear-cut decisions should almost always go the same way. It's only the close ones that are likely to have multiple results.

I don't think any computer bridge program works entirely off a rule base, simulations are necessary to fill in the gaps. It might be possible to ensure consistent behavior across all tables by seeding all their random number generators the same, though. That would take out some of the luck factor, I'm not sure if it's desirable or not.