[sacto123]

---

Instant tournament with bots (IMPS).

The explanation said that the 4C bid denied a major suit ace. Thinking that slam might be a bad idea with two aces missing and the opening leader aware of the fact, I chose not to bid slam. Silly me. I suggest that the explanation be revised. I understand that the explanations are sometimes not completely accurate (but close enough), but this one is too egregious (not even close). Only four bidders on this hand used this exact sequence. Three stayed out of slam. One (maybe did not read the explanation) bid it anyway.

[smerriman]

GIB has no idea how to cuebid. Avoid it wherever possible. Long standing problem that could only be resolved by switching to Italian style cues.

[manudude03]

From memory, it seems the system does need changed here. Bidding 3M over 3D shows shortage, while bidding 4M would appear to promise AK.

[johnu]

smerriman, on 2019-June-12, 16:54, said:

GIB has no idea how to cuebid. Avoid it wherever possible.

GIB knows how to cuebid. What GIB doesn't know is what to do with the information the cuebidding provides and shows.

smerriman, on 2019-June-12, 16:54, said:

Long standing problem that could only be resolved by switching to Italian style cues.

There's no reason to expect that GIB will "handle" Italian style cuebids any better. That will take several orders of magnitude more and better programming than just adding Italian cuebids.

[smerriman]

johnu, on 2019-June-12, 18:28, said:

There's no reason to expect that GIB will "handle" Italian style cuebids any better. That will take several orders of magnitude more and better programming than just adding Italian cuebids.

It would work far better; Italian cuebidding is *significantly* simpler logic wise for a robot to understand. The majority of issues with GIB's cuebidding is because it leaves itself with no possible bid; after 1♠ - 3♠ - 4♦ it will bid Blackwood with xx in clubs but extra strength simply because it has no alternative.

Yes, there is still plenty of complexity involved, but even the very basics in combination with Advanced GIB's ability to plan out the auctions would result in a huge improvement.

[johnu]

smerriman, on 2019-June-12, 19:07, said:

It would work far better; Italian cuebidding is *significantly* simpler logic wise for a robot to understand. The majority of issues with GIB's cuebidding is because it leaves itself with no possible bid; after 1♠ - 3♠ - 4♦ it will bid Blackwood with xx in clubs but extra strength simply because it doesn't know what else to do.

Yes, there is still plenty of complexity involved, but even the very basics in combination with Advanced GIB's ability to plan out the auctions would result in a huge improvement.

Get back to me when GIB can 100% successfully make a simple preference bid when holding a singleton or void in one suit and 4+ cards in the other suit. Compared to cuebidding, this should be a cakewalk. Even better, get back to me when the GIB description of a competitive 4 or 5 level game bid doesn't show 25-30+ HCP. How hard can that be??? Answering my own question, apparently too hard to implement after ~20 years of GIB. Presumably I agree with you that it should be simple, but clearly it is not. Wishing it were so doesn't make it so, no matter how hard you wish.

[barmar]

manudude03, on 2019-June-12, 17:06, said:

From memory, it seems the system does need changed here. Bidding 3M over 3D shows shortage, while bidding 4M would appear to promise AK.

Yeah. Part of the problem is that the code that interprets cue bids is not aware of the bidding rule database. So it doesn't realize that it bypassed diamonds because the bidding rule would make 3♦ show something else, and thinks that the non-cuebid denied the ace.

Italian cue bids wouldn't help here, since it would just make it think that North denied both the ace and king, not just the ace.