smerriman

Odd one. For reference, basic GIB doesn't double (in fact, it doesn't bid 4♥ either, it bids 4♦, but 4♥ based on a simulation seems reasonable). I agree basic GIB doesn't understand what a penalty double is, but it must be relying on a simulation for a double here.. which you'd think it would get right. It does play 2NT as showing a good hand (by a passed hand, a maximum of exactly 12 points it seems), so you are weaker than it expects, but surely even if you had that a simulation wouldn't show 4♠ going down most of the time.. would it?

Yeah, that's pretty bad. Basic GIB passes 3NT, so maybe when simulating it was just trying to decide between 4♥ and 4♠ and didn't think that the former could be corrected to 4♠.. or maybe all of its simulated hands just had you holding spades. Who knows.

In those formats, like in normal bridge, you play when you are declarer, and your partner (the robot) plays when you are dummy.

I was solely commenting on your point about being worth a bid if game was possible opposite a maximum. The HCP are therefore a given.

What would you call a magic fit? I did a quick sim, and any 4 spades along with 17-21 HCP is making 4♠ about 60% of the time. So it's not out of the question hoping for 4 card support.

We know what it used to look like a long long time ago; I doubt the format has changed much since then.

GIB doesn't work like that. To GIB, this hand has 13 total points. So yes, it should probably have the lower limit increased, but it was at least consistent with the description.

Out of interest, what are you saying the bug is here? It may not be your choice of bid but GIB does, after all, have exactly what it promised in the description.

Following the beginner rule of not asking for keycards when you won't know what to do with the answer, could you also argue that South can't have two quick losers in diamonds? (Or is there no other option here - not sure on the bidding system.) ([edit] nevermind, regardless of system obviously there was no chance of doing anything else to stop in 5♣ anyway.)

Without the double, 2♥ is described as 5+♣, 4+♥, 18-22 total points and GIB correctly gives preference. With the double, 2♥ is described as 3+♣, 4+♥, 15-22 total points. So that's where the bug is.

Just to clarify one thing, I think GIB is excellent, and I've learnt far more playing with GIB than humans online. But if straightforward bugs reported here will go unfixed or unseen for years (was only looking at my own; there are unresolved bugs reported by others dating even further back), then BBO should really admit as much and close this forum, rather than promising things are being looked into when they're not.

Apologies if you thought I was putting you down in any way - not my intention in the slightest. I though the main point of this thread was that you thought the description implied a lack of stopper, and thus the description needed fixing - was just pointing out that if you follow this through to a logical conclusion, you'd get something absurd, so your assumption couldn't have been correct. But I guess I must have misunderstood :)

Over the past year and a half I have posted 24 cases of 100% replicable bidding issues with GIB, all with straightforward bidding sequences which will likely reoccur (and in some cases have even popped up in my own hands more than once) - as opposed to highly competitive auctions or freak hands. (I'm not counting a 25th where GIB held a strong 7-6 hand). One of these where GIB passed a cuebid was apparently reported numerous times during a survivor tournament and was quickly fixed. One was recently one of the major changes mentioned in the release notes, regarding responses to cuebid raises - this just affected bidding descriptions and not the actual bidding, and the fix wasn't tested as it quickly resulted in another, even worse, bug. Three more I can no longer replicate, though received no official replies and no mentioned of the fixes in release notes, so it's hard to know whether they were specifically fixed due to the report. At least 2 occurred in daylongs (can't remember about the third), so may have been reported en masse there. 19 still occur today. These include my first, from June 2016, where GIB stops short of game with 12 points opposite a natural 2NT overcall, where Fred himself commented and said he'd make sure the programmers take a close look at it. The fact that GIB passes 1♦ - 2♣ - 3♣ but describes it as forcing has gone ignored since August 2016. jdonn mentioned a change needed to be made in this auction in October last year. It still needs to be made. jdonn commented on GIB mishandling bids after takeout doubles, promising a 6 card suit when only having 4, saying the programmers were working on it. I've seen someone else post on it since; they've obviously still working on it a year later. jdonn confirmed they were aware of issues with balancing 1NT overcalls last January, but nothing was done. jdonn also said he'd report a bug where GIB stopped short of game with 4 points opposite a balanced 22-24 back in January, without result. Without even realising, the same thing happened to me again a few months later. Also in January there was the odd case of GIB promising 4 spades with 3, showing 18 points when it held 13, cuebidding an Ace it didn't have, and then stopping short of game. jdonn said he'd hope fixing an early bug would resolve the rest. None were fixed. GIB's cheaper minor bug has been reported by several people without response. We've got GIB bidding a side suit at the 5 level rather than show a 10 card fit; a previously fixed bug with bar bids resurface, a completely undescribed cuebid, a passout with a standard 1♣ opener, a passed cuebid, a passing forcing redouble resulting in a 0-1 trump fit, broken descriptions after a takeout double, a passed game try, passing a takeout double with nothing (twice), and finally stopping at game despite getting a perfect blackwood response. Granted, the last few occurred recently, and maybe haven't been read yet. Given others were read and not sorted more than a year ago, I'm not holding my breath. So, given BBO clearly don't have resources to read or fix these bug reports, or want any help in getting such resources, is there any purpose in this forum (other than amusement value)?

Yeah, this one has been reported on several occasions already.

Basic GIB passes, doesn't bid 3NT. Advanced GIB was gambling that 3NT would work out better in the long run. Nothing wrong with that.

Look in your browser title bar.

You did. You said yourself that you thought "promises a partial stopper" also means "denies a full stopper". GIB's description promised a "partial stop in ♣, partial stop in ♦, partial stop in ♥". Therefore, if your assumption was correct, you are assuming GIB is denying a full stopper in all three suits by bidding 3NT. This is of course nonsense, so your original assumption was also nonsensical.

If the description of 3NT was posted to Bridgewinners, I agree it would be unanimous - in favour of Stephen Tu's explanation. I'm afraid it's pretty obvious that such a 3NT bid describes a minimum amount of stoppers in each suit; why would you ever want a 3NT bid to deny a full stopper in every suit?

You can't have checked all of the options - 3♠ says 5+ ♥, 13+ total points, forcing to 4♥. Not sure I would have predicted that in advance, but at least it seems to do what you want and let you use Blackwood next. As least the roles weren't reversed - when GIB bids with itself, it passes out 1♠ :D Not too surprising though, these types of hands can be hard for human partnerships to bid, let alone a robot. Given the number of much more common bugs, I don't think crazy hands like this are really something that should be looked into.

Ah, OK. I was thinking of auctions like 1♣ - 1♥, where a balanced 18-19 points always rebids 2NT rather than jump shifting in spades (and responder then bids 3♠ to accept the invite and check for a 4-4 fit). I guess 5422 isn't quite the same, but it feels like a lot of different ways of showing 2 or 3 spades, when you have an initial 2 spade raise, or a later pass, 3nt, 3 hearts, or 3 spades.. They're the descriptions that appear when hovering over the bid when playing with GIB. You can recreate the auction and see them at a bidding table. I wouldn't pass, but in this forum I'm not really interested in what I would/wouldn't bid, just that whatever I choose, GIB would understand it. If I had a couple of extra points I wouldn't know whether to bid 3♥ or not given the definition.

Count me in as always.

Definitely looks like something is wrong with GIB's definitions here: 3♣: Transfer - 6+ ♦, 2- ♠, 6 HCP 3♦: Transfer - 5+ ♥, 2- ♠, 6-11 HCP, 12- total points 3♥: 2 ♠, 11- HCP, 10-12 total points 3♠: 2 ♠, 6-11 HCP, 12- total points, likely stop in ♠ 3NT: 5- ♥, 2- ♠, 8-11 HCP, 12- total points 3♥ should surely be aiming for a 4-4 fit, but the description doesn't even promise a heart.

4NT is undefined, as can be seen by the fact it doesn't provide any new information from your previous bid. Making an undefined bid will always get you a pass, but that was predictable in advance by looking at the definitions.

My non-expert thoughts: Leading the A can't be right (unless you're worried about ruffs elsewhere). Not sure what 'know next' means; at that point you have a singleton so it doesn't matter what you know.. The only time the other two options make a difference is if the 2nd hand to play holds Kxx or Jxx - they may go up with the K on some occasions, so I'd play the Q T (fixed typo, thanks).

Taking the diamond finesse only works if you opened the bidding 1♥ (in second seat no less) with 8 AJ7654 9xxx A8. Perhaps you might do this on the rarest of occasions, but that would be pretty wild - to GIB, both are 0% chances. Making assumptions based on the bidding is an important part of the GIB logic; I agree that it often overassumes too much, but it's tough to call this one a bug.