[AyunuS]

Advanced GIB, IMPs.

http://tinyurl.com/qh64t7c

Pretty much speaks for itself. And I could understand with some hands, but here, I don't see any good reason why not 1♠.

[lycier]

---

I don't understand why advanced Gib looks like basic in this point, also can't confirm whether it is a bug.

[Bbradley62]

Why do you think this is a situation where basic vs advanced GIB has anything to do with the conversation?

[steve2005]

Gib needs to take a beginner's bridge lesson. lol

[iandayre]

One of GIB's most basic problems is that it does not plan ahead in the bidding. It figures out the bid most fitting its programming on one round (apparently this hand is too strong for a balancing suit call at the one level), and starts the same process again every round. Here it does not consider what to do over the 1H advance if it makes a takeout double, and paints itself into a corner (not good enough to double and bid a suit). I would estimate that in over half of all GIB auctions, the descriptions of bids are not sensible given the meaning of previous calls at the table).

[steve2005]

Yes, your right that Gib only thinks about current round of bidding. One of the first things I was taught about bidding is to consider what your rebid is going to be.
But thus hand is just silly. Is Gibs system to balance with a double as an unpassed hand on most 9+pts hands in balance? Surly you should be able to bid a suit with as much as 16 or 17 points or whatever range you use which is beneath a double followed by bidding a suit.

[Stephen Tu]

In theory GIB only needs better rules for the current round. If it makes the best call at every opportunity it should be well placed for future rounds and fall into the best bidding sequence. It's easier to code this way. For this one it just has to bump up the priority for balance overcalling 5 cd major above double unless too strong to overcall.

[lycier]

Stephen Tu, on 2016-January-09, 11:46, said:

In theory GIB only needs better rules for the current round. If it makes the best call at every opportunity it should be well placed for future rounds and fall into the best bidding sequence. It's easier to code this way. For this one it just has to bump up the priority for balance overcalling 5 cd major above double unless too strong to overcall.

Great.
Generally speaking, your comments are very instructive, I think you would be our teacher in the Gib robot forum.

[1eyedjack]

Stephen Tu, on 2016-January-09, 11:46, said:

In theory GIB only needs better rules for the current round. If it makes the best call at every opportunity it should be well placed for future rounds and fall into the best bidding sequence. It's easier to code this way. For this one it just has to bump up the priority for balance overcalling 5 cd major above double unless too strong to overcall.

The fundamental problem is that there are three basic hand types for the initial double:
1) A 3-suited hand short in opponent's suit (with values but potentially modest strength, depending on position).
2) A balanced hand too strong for an immediate call in NT
3) A strong jump overcall hand type where an immediate jump overcall would be pre-emptive and it is too strong for a simple overcall

GIB is perfectly happy to apply these rules when it makes its own double, but then later in the auction will, opposite such a double, assume that partner must have hand type (1) without catering for any possibility of the other hand types.

Your "solution", if I read you right, is to increase yet further the already wide range of a simple overcall. In so doing, you reduce the frequency of the initial double and therefore reduce in proportion the frequency of the potential problems that emanate from GIB's fixation on hand type (1) in the later continuations.

Well, that is true as far as it goes, but in my book it does not qualify as a "solution".

If it is too difficult for the programmers to cater for the 3 hand types later in the auction (and I would sympathise with them if that were the case) then they would be better off ditching weak jump overcalls and perhaps Michaels, use the cue as a stone-age general force, and restrict the double to a 3-suited (or perhaps balanced) hand. And perhaps up the minimum strength for the balancing 1N a bit.

ie, surrender to the inevitable conclusion that we are decades away from getting computers to play a system that expert humans regard as standard (when partnering other humans), and instead dumb the system down to something that the computer can play *well*.

[cloa513]

1eyedjack, on 2016-January-09, 15:50, said:

The fundamental problem is that there are three basic hand types for the initial double:
1) A 3-suited hand short in opponent's suit (with values but potentially modest strength, depending on position).
2) A balanced hand too strong for an immediate call in NT
3) A strong jump overcall hand type where an immediate jump overcall would be pre-emptive and it is too strong for a simple overcall

GIB is perfectly happy to apply these rules when it makes its own double, but then later in the auction will, opposite such a double, assume that partner must have hand type (1) without catering for any possibility of the other hand types.

Your "solution", if I read you right, is to increase yet further the already wide range of a simple overcall. In so doing, you reduce the frequency of the initial double and therefore reduce in proportion the frequency of the potential problems that emanate from GIB's fixation on hand type (1) in the later continuations.

Well, that is true as far as it goes, but in my book it does not qualify as a "solution".

If it is too difficult for the programmers to cater for the 3 hand types later in the auction (and I would sympathise with them if that were the case) then they would be better off ditching weak jump overcalls and perhaps Michaels, use the cue as a stone-age general force, and restrict the double to a 3-suited (or perhaps balanced) hand. And perhaps up the minimum strength for the balancing 1N a bit.

ie, surrender to the inevitable conclusion that we are decades away from getting computers to play a system that expert humans regard as standard (when partnering other humans), and instead dumb the system down to something that the computer can play *well*.

You mean something GIB can play well- I'd makers Bridge Baron 13 which is miles better than GIB would disagree with you. Their limits are more like can only play a limited system- limited ability to handle psyches.

[1eyedjack]

cloa513, on 2016-January-09, 18:32, said:

You mean something GIB can play well- I'd makers Bridge Baron 13 which is miles better than GIB would disagree with you. Their limits are more like can only play a limited system- limited ability to handle psyches.

Care to have another go? Usually I can decipher the true intent behind typos, but this was so grammatically flawed that it beat me. Late on Saturday night, I do realise.

[Stephen Tu]

1eyedjack, on 2016-January-09, 15:50, said:

The fundamental problem is that there are three basic hand types for the initial double:
1) A 3-suited hand short in opponent's suit (with values but potentially modest strength, depending on position).
2) A balanced hand too strong for an immediate call in NT
3) A strong jump overcall hand type where an immediate jump overcall would be pre-emptive and it is too strong for a simple overcall

GIB is perfectly happy to apply these rules when it makes its own double, but then later in the auction will, opposite such a double, assume that partner must have hand type (1) without catering for any possibility of the other hand types.

Your "solution", if I read you right, is to increase yet further the already wide range of a simple overcall. In so doing, you reduce the frequency of the initial double and therefore reduce in proportion the frequency of the potential problems that emanate from GIB's fixation on hand type (1) in the later continuations.

Well, that is true as far as it goes, but in my book it does not qualify as a "solution".

I don't really see the problem. Most good human players play a fairly wide range for overcalls and deal with it fine for the most part. 7-17 or some approximation of that. Surely the 14 count in the original problem with 3 pts in a stiff isn't nearly strong enough to double and bid again. What's wrong with forcing GIB to prioritize 1♠ over double when holding weaker holdings?

Later in the auction, it is true that responder is supposed to cater to the "support for all outside suits" more common double rather than the powerhouses. But it's relatively easy to reveal the powerhouses IMO. If it is "good one suited hand", and partner doesn't bid your suit, you just bid the suit. If it's hand too strong for NT, then you bid the appropriate amount of NT. If it's powerful, no stopper, no 4 cd support for partner, no suit to bid, then you fall back on double then cue. It's just a matter of defining the bids for GIB and at the right priority in relationship to each other. I don't see why it would be impossible to program. It's just that right now there are lots of bugs in this area and they haven't addressed them.

1. Advancer of the doubler really needs to be able to prioritize bidding 4 cd unshown majors ahead of longer minors. It's just a ton easier to reach the best most frequent games this way if doubler doesn't have to figure out how to find 4-4 major fit after responder bids a minor in response. Minor should basically deny 4 cd outside major IMO.

2. If advancer shows a decent amount of strength by jumping or bidding NT, then if opener raises NT it shouldn't show quite as much as if advancer didn't show any strength by just bidding at lowest level. No more of this 1S-dbl-p-3m-p-3nt = 25+ baloney

3. If one started with a double intending to show a good one suited hand, on the next round, highest priority should go to bidding that suit! We've seen lots of examples posted where on the next round GIB instead of bidding the suit, tries double again, or cue bid, and then ends up somewhere ridiculous never having shown the suit why it doubled in the first place.

If they fixed these things I don't see a problem with including powerhouse hands in the double.

[Siegmund]

Many humans play "7-17" in direct seat, and "a king lighter when balancing than in direct seat" -- if GIB devalues its HK, sure, it has a maximum balancing 1S. But Axxxx x Kxxx AQJ is strong enough that I would be doubling intending to rebid a spade too.

I find it odd that GIB would double and then pass with this hand. If GIB had doubled then rebid 1S I would have said "nothing to see here, move along, just some bad hand evaluation."

[Stephen Tu]

I don't think that you should drop your standards so much just because you are balancing. Make lighter takeout doubles, by 2.5 pts, fine, direct doubles are stronger than overcalls, and there are 9 cts you need to be protecting on. But if you overcall light to begin with, geared towards obstructing the opponents as often as possible and getting in at a safe level before the opponents figure out they were supposed to double you, now you want to balance overcall on 4 counts?? That's taking things too far. Partner failed to take direct action, to me you still want around 7-8 to act, you are hoping partner had some 10-14 unsuitable for a takeout double and that you have close to half the deck and can buy contract or at least push them up a level or two to set them.

As for the top end of the range, the only reason to double then bid a suit rather than just bid the suit, is because you are afraid of missing game when partner doesn't bid, and that it goes 1s-all pass when game is on. If you have a 16-17 count, and partner doesn't raise you, or bid something, why should you fear missing game? Why should partner with 8 points refuse to raise you? The probability of opps both letting you play 1s and that 1s makes but 2s doesn't is awfully small.

The idea of having to double first just because I have 14 high in balancing sickens me. I don't see any advantage of it. And besides, if you are doubling first on 14, how are you going to now bid when you actually do have 18+?? Now you are going to cue bid then bid your suit? You are going to force the auction really high, and overloading the cue bid.

I'd find it even more ridiculous if the majors were reversed and you were proposing balancing double with singleton spade because 14 is too much to overcall.

[lycier]

delete