[goffster]

http://tinyurl.com/j48eyys

Obviously when south is willing to pass out 2C, later forcing
partner to game is ridiculous. Yet another example of allowing
the hand evaluator contradict simple common sense rules.

[Stephen Tu]

I'd describe it as more a problem of rules not existing to better define the NS actions. Like double of 1nt is probably better than 2♣. And 3♣ given the definition of 2♣ probably shouldn't be "14+HCP, twice rebiddable clubs", to avoid North thinking it has 25+ HCP and probably should be keeping the auction alive. Needs more explicit rules that bidding one more in a strain previously passed out in a competitive auction is a signoff, shouldn't show much.

It's not a matter of "contradicting common sense rules". Remember for a computer there is basically no such thing as "common sense". Rules are either explicitly defined or not. If they are missing, or poorly defined, the computer has to grope based on default rules that are triggered, there is no "common sense" it can rely on, and it often does poorly in such cases.

[iandayre]

I don't wish to make a personal attack, but this appears to another one of your typical "ignore my bad bidding, focus on GIB" hands. Maybe if you tried bidding sensibly you wouldn't create so many problems for the robots.

[Stephen Tu]

iandayre, on 2016-September-13, 12:07, said:

"ignore my bad bidding, focus on GIB" hands.

We SHOULD be doing this. Ignore what human does, even if it is questionable or downright awful. Imagine you only see GIB's hand and the auction.. Bidding 2c on 3433 11 count instead of doubling 1nt is pretty bad IMO. Trying 3s after partner competes with 3c after passing out 2c is also really bad, 3c should be competitive and partner should not be bidding again uncontested. These are still bugs regardless of whether the 3c bid itself was reasonable or not. Bids should be analyzed isolated from the content of the other hands, whether their bids are normal or not.

[iandayre]

Stephen Tu, on 2016-September-13, 12:28, said:

We SHOULD be doing this. Ignore what human does, even if it is questionable or downright awful. Imagine you only see GIB's hand and the auction.. Bidding 2c on 3433 11 count instead of doubling 1nt is pretty bad IMO. Trying 3s after partner competes with 3c after passing out 2c is also really bad, 3c should be competitive and partner should not be bidding again uncontested. These are still bugs regardless of whether the 3c bid itself was reasonable or not. Bids should be analyzed isolated from the content of the other hands, whether their bids are normal or not.

Well Stephen, if the only purpose of this forum is to communicate with BBO about GIB's imperfections, I would agree with you. But since this is the only forum exclusively for GIB users, I don't think it is inappropriate to suggest actions that maximize GIB's effectiveness and to discourage those that don't. The point about GIB's problems is still there to be seen.

[Stephen Tu]

iandayre, on 2016-September-13, 13:54, said:

Well Stephen, if the only purpose of this forum is to communicate with BBO about GIB's imperfections, I would agree with you. But since this is the only forum exclusively for GIB users, I don't think it is inappropriate to suggest actions that maximize GIB's effectiveness and to discourage those that don't. The point about GIB's problems is still there to be seen.

To me this seems to be mainly a forum for GIB bug reports and discussing how it could be improved since there doesn't seem to be any other venue other than the send robot report button in BBO. It's not a "help humans make better bidding decisions" forum, that's what the other sections of the forum are for. If someone makes a GIB bug report, and it's pretty clear GIB should have done better and it is a bug, IMO we should ignore any human errors. Unless human poster is specifically asking if their own bids were OK or if they are asking in general how to score better in bot tourneys or something of that nature.

I'd only bring up the human bidding if the human perpetrated some auction that really shouldn't exist and it's not reasonable for GIB to have to know how to continue. Like if we were discussing GIB maybe going haywire after say 1nt-p-3nt-p-4♦-p-?, then I'd think you have a point. But on this auction, GIB made two egregious mistakes, the human overbidding is irrelevant to the fact that GIB should be improved here. If humans refrain from "creating problems for the robots", then we avoid finding bugs. If South had passed instead of opening, the hand gets passed out, we don't find out that GIB makes bad decision to bid 2♣ instead of doubling. To me it's good to create problems for the bots, that's how bugs get exposed and hopefully eventually get stamped out. Avoiding GIB's weak auctions doesn't make GIB better. Getting GIB into its problematic auctions and reporting the bugs can.

[lycier]

Honesty is the best policy. The best beautiful lies wouldn't keep long time.

[iandayre]

Stephen Tu, on 2016-September-13, 16:07, said:

To me this seems to be mainly a forum for GIB bug reports and discussing how it could be improved since there doesn't seem to be any other venue other than the send robot report button in BBO. It's not a "help humans make better bidding decisions" forum, that's what the other sections of the forum are for. If someone makes a GIB bug report, and it's pretty clear GIB should have done better and it is a bug, IMO we should ignore any human errors. Unless human poster is specifically asking if their own bids were OK or if they are asking in general how to score better in bot tourneys or something of that nature.

I'd only bring up the human bidding if the human perpetrated some auction that really shouldn't exist and it's not reasonable for GIB to have to know how to continue. Like if we were discussing GIB maybe going haywire after say 1nt-p-3nt-p-4♦-p-?, then I'd think you have a point. But on this auction, GIB made two egregious mistakes, the human overbidding is irrelevant to the fact that GIB should be improved here. If humans refrain from "creating problems for the robots", then we avoid finding bugs. If South had passed instead of opening, the hand gets passed out, we don't find out that GIB makes bad decision to bid 2♣ instead of doubling. To me it's good to create problems for the bots, that's how bugs get exposed and hopefully eventually get stamped out. Avoiding GIB's weak auctions doesn't make GIB better. Getting GIB into its problematic auctions and reporting the bugs can.

True, it's great for the GIB community as a whole to learn how the robots handle a wide variety of auctions. But I am assuming that when most of us play in GIB tournaments, we are trying for a good score, not to perform experiments. I certainly don't comment every time I see human bidding I don't like, but when I see a pattern I don't think a comment is inappropriate. To each his own though.

[jdonn]

It's a bit of a problem that passes in competitive auctions are often not defined. So like 1C P 1H 2D, P 2S P P, 3C might have the same definition as 1C P 1H 2D, 3C (I'm just making this auction up / guessing, this exact auction might not be a problem but this is the type of auction that has the problem) because the pass over 2D doesn't particularly have a definition. It would be a massive undertaking to define all kinds of passes like this, but it's on our radar that the problem exists.

Personally I have no problem with people reporting bugs on hands where the human bid or played 'inferior', as long as their actions were still plausible. For example, I have had a report where someone opened 1NT with like a 9 card suit, then bid the suit 4 times and GIB kept overruling them to its own 6 card suit. Sorry, no luck there, you can't say you have one hand then try to undo it and expect GIB to work it out. But like on this hand, even if 3C wasn't the best bid, I am still glad to get the report if GIB did something wrong since 3C is still a plausible bid.