smerriman

Yes, double dummy play, which is obviously a bit flawed in real life, but does match the exact statement of "can set the opener", vs "will" :)

Assuming: - RHO opens 1NT with any balanced 15-17 HCP (ie they don't upgrade 5 card suits) - partner never runs with a weak hand I get the following: If you have a balanced 15 points, 50.2% If you have a balanced 16 points, 53.0% If you have a balanced 17 points, 60.4%. Someone else can check my numbers but that sounds about right.

I said that basic GIB doesn't run simulations during bidding. The entire cardplay algorithm is based on double dummy analysis of simulated opposition hands (as is the case for all bridge robots).

People have definitely experimented with learning algorithms. See eg https://arxiv.org/pdf/1607.03290.pdf . The catch is that the best bidding system for a robot is completely incomprehensible for a human, and as stated near the end of that PDF, trying to make it human-understandable usually gives a much worse result. (GIB itself was designed to play Moscito Byte, a much simpler system for robots than 2/1, but much more complicated for humans).

Your 4♠ bid is a cue bid in spades, promising the ace or a void (GIB requires showing first-round controls before second-round controls). Since it has the ace, you must have a void, despite the 2NT opening bid (I'm not sure how it decides which lie to believe when you have contradicted yourself). If you have 20-21 points and a void in spades, 7♦ looks a good bid.

2♦ shows 4 or less clubs, not 4 clubs. When virgosrock posts money bridge hands, something seems to go wrong with the formatting. Unfortunately GIB has a tendency for bidding irrelevant 4 card suits, even at high levels.

Basic GIB passes both, so I guess you just got an unlucky set of simulations on each.

That's the point - the book bid is wrong and should be fixed, but advanced GIB, if simulations had been turned on for the double (perhaps they aren't?) should have been able to figure that out and correct its mistake.

Argh, I'm an idiot. I was surprised originally too but thought I had double checked things. Forgot to put a lower limit on West's point range :( making +1: 68 making =: 186 down 1: 168 down 2: 65 down 3: 11 down 4: 2 Not doubling = 16 cents to the defence; doubling = $4.27 to the declarer. So, I'm completely wrong. Thanks for correcting me. If GIB were able to simulate in this situation it probably wouldn't have doubled too.

It follows a gigantic set of rules. You can see what it used to look like many many years ago here: http://orig.gibware.com/bidding/ It's not a good idea playing with GIB if you don't know its system! How it plays Blackwood is clearly outlined here: https://www.bridgebase.com/doc/gib_system_notes.php

( Ignore everything I wrote :( )

The first board has nothing to do with simulations; all non-human bids are book bids. 4♦shows 5+ ♦, 3- ♥, 3- ♠, 13+ total points. Looks like a perfectly good description to me? It's not like you want to be playing in 4♦ knowing you both have minimums. 3♠ down 1 looks the best spot (and that's where GIB would have reached). Only yourself to blame here. In the second board, GIB without simulations would pass your double and not bid 4♥. 4♥ looks like a superior bid to me. I ran a rough simulation of 500 hands myself - giving east a weak 2 in diamonds, west any 4 diamonds, and south any normal takeout double (excluding off shape very strong hands) - and 4 hearts makes 75% of the time. Absolutely clearcut win for GIB's simulations. 5♣ definitely looks pushy, but don't forget you've lied about your strength by bidding 4♠ with that hand. Suppose you really did have 19+ total points. Another simulation of 500 hands showed that there was a slam making in either clubs or spades over 50% of the time - and that's not even taking into account the fact that you could sign off at the 5 level. I'm siding with GIB here.

I appreciate you getting involved with this forum, but please get to learn a bit more about how GIB works before making these sorts of assertions. This has absolutely nothing at all to do with simulations - this is basic GIB who runs no simulations during the bidding. And 6 clubs is ludicrous over 4NT, because it shows an odd number of keycards and a void in either diamonds or hearts.

The key with daylong tournaments is that they're best hand, so here you know nobody has more than 11 HCP. That weighs decisions far more towards always opening (and often passing "unlimited" responses).

[hv=http://www.bridgebase.com/tools/handviewer.html?sn=smerriman&s=SAKQ82HK7DKTCAKQ8&wn=Robot&w=SJT94H96D872CT765&nn=Robot&n=S3HAQT82DQ54CJ432&en=Robot&e=S765HJ543DAJ963C9&d=e&v=e&b=6&a=P2C(Strong%20two%20club%20--%2019+%20HCP%3B%2023+%20total%20points%3B%20forcing%20to%202N)P2H(Positive%20--%205+%20%21H%3B%20%21HKQ%3B%208+%20total%20points%3B%20forcing%20to%202N)P2S(5+%20%21S%3B%2019+%20HCP%3B%2023+%20total%20points%3B%20forcing)P2N(5-6%20%21H%3B%205-%20%21S%3B%20%21HKQ%3B%208+%20total%20points%3B%20partial%20stop%20in%20%21C%3B%20partial%20stop%20in%20%21D)P3C(4+%20%21C%3B%205+%20%21S%3B%2019+%20HCP%3B%2023+%20total%20points)P4N(Blackwood%20%5BC%5D%20--%204+%20%21C%3B%205-6%20%21H%3B%205-%20%21S%3B%20%21HKQ%3B%209+%20total%20points%3B%20partial%20stop%20in%20%21D)P5C(Zero%20or%20three%20key%20cards%20--%204+%20%21C%3B%205+%20%21S%3B%2019+%20HCP%3B%2023+%20total%20points)PPP&p=H9HTHJHKCAC6C2C9CKC7C3D3C8CTCJH3C4H5CQC5SAS4S3S5SKSJD4S7SQSTD5S6H7H6HAH4HQD9S2D2H8D6S8D8H2DJDTD7DQDADKS9]400|300[/hv] Cost me 8 IMPs which would have placed me 3rd in the daylong instead of 11th :( (Well of course, probably less since the same thing happened at a number of other tables.)

Rather than looking at the Results page, if you log onto BBO and go to My BBO - Hands and results - Recent Tournaments - My Hands - Other Tables, you can see a full list of everyone who played the hand, and how. PS - be warned - you may be disappointed to find that it's something as trivial as playing a different card from equals which results in a different switch from GIB.

The process manudude03 described here works perfectly fine. I'm not sure what you're doing differently in to order to get weird plus symbols etc - if you're following that process exactly, where are you copying and pasting the link from?

Fair enough. I suspect it would still be considerably more complex than you expect, especially when you consider opposition interference, but getting rid of all conventions altogether would definitely make for a simpler robot, albeit not one that I'd ever be able to play with. Simulations are only used by advanced robots, only for a handful of the bids, and even then, only rarely change the resulting bid (it still uses rules to come up with its first choice of bid). It's what causes the biggest improvement over the basic bots, so not something you'd want to get rid of.

This isn't true at all. The logic behind the play and defense algorithm is actually reasonably straightforward; coming up with bidding rules to cover every possible situation is incredibly complicated.

GIB is definitely buggy with respect to a game try in hearts. See http://www.bridgebase.com/forums/topic/77124-gib-passes-game-try/ for another example (can still replicate that bug).

As in, the extremal case - it won't even bid 1♦ when holding a 9 card suit to the AJ and 5432 in hearts (as opposed to, say, it only doing it when it had heart honors, or a 6 card diamond suit and not 7, etc).

I can't really see much wrong, other than 7♦ which seems risky to rely on the queen. GIB has exactly what it promised.

Taking this to the extreme, if you give GIB ♥5432 ♦AJT987654, it will bid the same way - 1♥ then pass. This doesn't seem right.

For what it's worth, the option isn't called 'blacklist', it's called 'ignore'; and when read like that, it's more obvious why chat is 'ignored'. But it probably makes sense that chat shouldn't be blocked in *tournaments* - as the TD can always ban them if they are being abusive. Not being paired against someone you've ignored doesn't make sense; people could then game the system by ignoring good players. Incidentally, there's a little known feature where if you right click someone's username, you can set a category and have them show up in a different color without ignoring them. Useful when you want to mark players as those you don't want to play with, without ignoring them entirely.

That's a formatting issue with the way the handviewer link has been embedded. Most of the + symbols represent spaces - the actual description is 4+ ♣, 4+ ♦, 5- ♥, 5- ♠, 11- HCP, 12 total points.