[Thranduil]

Hand: https://tinyurl.com/259cf54t

I know that auctions with off-shape takeout doubles often are problematic, but - so I thought - it should at least get the very basic strong overcall, X followed by my suit with opps always silent, right. And... it dsoes not.

First, the 3NT are described as showing a stop in ♥ - but QJ blank is not a stop. Then when I want to go for a slam opposite the "13 HCP" that actually were just 10 - or more like 7 for us humans who know that we cannot count a blank quack in opp's suit - it insists in ♦ being trump. I never confirmed a ♦ fit, I wanted to play 6♠ or 6NT. It even goes so far and bids 7♦ with a missing ace over my 6♠.

[Huibertus]

Not sure on other devices, but on a PC, when you hover the mouse pointer over a bid it shows you the meaning. Use this before deciding to click on it or not. As crazy as it is, it would have warned you 3♠ promises ♦ fit. It would have made clear 4NT therefor is RKCB for ♦ and thus forcing, 5♥ Queen ask for ♦ thus forcing, the feature prevents a lot of frustration.

[Thranduil]

Huibertus, on 2025-June-15, 10:57, said:

Not sure on other devices, but on a PC, when you hover the mouse pointer over a bid it shows you the meaning. Use this before deciding to click on it or not. As crazy as it is, it would have warned you 3♠ promises ♦ fit. It would have made clear 4NT therefor is RKCB for ♦ and thus forcing, 5♥ Queen ask for ♦ thus forcing, the feature prevents a lot of frustration.

I am on PC, yes, and I know that. But it did not cross my mind that 3♠ confirms a ♦ fit - that's just stupid, it should show what I had, a ♠ one-suiter too strong for a direct 1♠ overcall. Therefore I did not bother checking the description of 3♠. And, aside from that, as this very hand demonstrates, these descriptions are wrong quite often.

[pescetom]

Thranduil, on 2025-June-15, 11:11, said:

I am on PC, yes, and I know that. But it did not cross my mind that 3♠ confirms a ♦ fit - that's just stupid, it should show what I had, a ♠ one-suiter too strong for a direct 1♠ overcall. Therefore I did not bother checking the description of 3♠. And, aside from that, as this very hand demonstrates, these descriptions are wrong quite often.

A human partner will only punish you if you panic and bid spades without sufficient strength: but GiB will punish you simply for not having his diamonds, whatever calls you now make and however feeble his diamonds are.

I'm not sure why you say this hand demonstrates that the descriptions are often wrong: here I fear that the description is correct and the "agreement" is wrong
But yes I agree that they are sometimes wrong and often simply the name of a convention (which is meaningless if you do not know the convention or share GiB's idea of it).

[smerriman]

Thranduil, on 2025-June-15, 07:46, said:

I know that auctions with off-shape takeout doubles often are problematic, but - so I thought - it should at least get the very basic strong overcall, X followed by my suit with opps always silent, right.

Why would you think this? Surely you've read the countless threads in the past that say never, ever, ever make an off-shape double with GIB. Virtually the only sequence it can handle is pass, lowest level response, pass, new suit. Any interference or non-minimum bid by partner and you're doomed forever to promise cards that you don't have.

As for the descriptions being wrong, no, GIB plays that 3NT shows a stopper as described, and its book bid is 4♦. Why it overrides this and bids 3NT I don't know; I'm guessing it thought with the expected 5-0-4-4 shape you had, it "knew" you would never pass and so it probably works just as well to get to you to the right diamond contract.

[Thranduil]

smerriman, on 2025-June-15, 14:19, said:

Why would you think this? Surely you've read the countless threads in the past that say never, ever, ever make an off-shape double with GIB.

Because that is standard bidding everyone who has learned about takeout doubles knows. As I wrote in my OP, I know that GIB has issues with off-shape doubles, but did not expect it to mess up a textbook example like this with opps silent. That is why I did not double-check the description of 3♠. Had an opponent bid something in-between, I would certainly have checked the description though.

smerriman, on 2025-June-15, 14:19, said:

As for the descriptions being wrong, no, GIB plays that 3NT shows a stopper as described, and its book bid is 4♦. Why it overrides this and bids 3NT I don't know; I'm guessing it thought with the expected 5-0-4-4 shape you had, it "knew" you would never pass and so it probably works just as well to get to you to the right diamond contract.

Quote

I'm not sure why you say this hand demonstrates that the descriptions are often wrong:

Did you both read my OP? It had a bad 10 count with no stop in hearts, but its bid was described as 13+ HCP and a stop in hearts (yes, it can have 13 HCP if opener has 11, so it's not describing something impossible here). That's an incorrect description right here. A description that does not match the bid. With a bot, I do not know about any "agreements", all I know is that bots usually - but not always - follow the basic bidding principles and what the bid descriptions tell me.
I am a TD myself and I know that in bridge with humans you need to differentiate between a wrong explanation (of a correct bid) and a wrong bid - but that does not apply to a bot from the perspective of the user.

[smerriman]

Thranduil, on 2025-June-15, 15:53, said:

Because that is standard bidding everyone who has learned about takeout doubles knows. As I wrote in my OP, I know that GIB has issues with off-shape doubles, but did not expect it to mess up a textbook example like this with opps silent.

Well, if you hadn't picked it up from all past threads, now you know not to expect that in future.

Thranduil, on 2025-June-15, 15:53, said:

Did you both read my OP? It had a bad 10 count with no stop in hearts, but its bid was described as 13+ HCP and a stop in hearts (yes, it can have 13 HCP here if opener has 11). That's an incorrect description right here. A description that does not match the bid. With a bot, I do not know about any "agreements", all I know is that bots usually - but not always - follow the basic bidding principles and what the bid descriptions tell me.
I am a TD myself and I know that in bridge with humans you need to differentiate between a wrong explanation (of a correct bid) and a wrong bid - but that does not apply to a bot from the perspective of the user.

Yes, I read your post. With GIB a description is not what it holds; it would be completely illegal / impossible for it to always describe what it holds. It's based on generic rules; when playing with a basic robot it often doesn't match because it was stuck for a bid or there is a bug with the description, and when playing with an advanced robot it often doesn't match because it's allowed to lie / deviate from its rules under the assumption that you will believe what it says and bid accordingly.

In this case, it is fully aware that 3NT promises a stopper, as opposed to it being a bug in the description, and is bidding based on the latter.

[Thranduil]

I think you are confusing something. I am not just talking about the lack of a stopper, I am talking about the stark deviation in HCP range too. There is no benefit to "bluffling" like that, all it does is make me go for an impossible slam.
Moreover, with a human partner I can talk before the game about our exact agreements and I know that they might make mistakes or intentionally deviate from our agreements if it is for our benefits - such as opening a good 14-count with 1NT when we agreed to a 15-17 1NT opening. When playing with a computer, I expect it to make no mistakes (unless there's a bug or glitch, which happens even with the best programs) and to always do the same thing in the same situation - therefore, when a bid is agreed to show a certain point range I expect a computer to make that bid only with hands that fall into that point range.

[smerriman]

Thranduil, on 2025-June-15, 16:36, said:

There is no benefit to "bluffling" like that, all it does is make me go for an impossible slam.

Advanced GIB is bad, but still light years ahead of basic GIB, precisely because of its "bluffing", or in better terms, allowing it to deviate from its rules if it believes this will lead to a better result. In this case, obviously, it's silly, but that doesn't mean the fundamental principle is wrong.

Thranduil, on 2025-June-15, 16:36, said:

When playing with a computer, I expect it to make no mistakes (unless there's a bug or glitch, which happens even with the best programs) and to always do the same thing in the same situation - therefore, when a bid is agreed to show a certain point range I expect a computer to make that bid only with hands that fall into that point range.

Well, you shouldn't with GIB. Making it stick to its prewritten descriptions would make it far worse than even the basic robot.