smerriman

This one of GIB's most well known play issues, and comes up regularly in the forum. Your 2♣ bid promised 4 clubs, West believes you, and doesn't include this hand in its simulations. It concludes it makes no difference what card it throws. Actually, when I ran GIB and it ran 290 simulations (way more than the BBO robot does), 289/290 times it didn't matter what it played, and on one singular deal it did, thus it played a club. But on BBO, it's highly likely that hand will never come up. Regularly suggested that in end positions GIB could easily deal all possible arrangements and gives ones it thinks are impossible a tiny weighting, just to break ties. But not implemented.

By finding and reading every bit of information about GIB that has ever existed online :) You can see the bidding database as it existed in 1999 well before BBO took over GIB and starting improving it here: https://web.archive.org/web/20200130215544/http://orig.gibware.com/bidding/ The old Windows downloadable version of BridgeBase included an mb.txt file which is what the robots used around 10 years ago (but it's the one with all the readable macros replaced out, which makes it a lot harder to read. Also looks like BBO has added a lot of extra features that weren't documented back then, so lots of experimentation and guesswork to figure out what all the extra bits may do.)

I would have agreed.. and as always, any comments I'm making recently are applying to an old version of GIB, so may have changed.. but in this case, not only does my version of GIB bid 4♠ here too, but it's told it's not *allowed* to simulate to choose a different bid. If I turn off that no-simulation flag, it simulates and passes. I think the flag is due to the values it says 3NT promises and the fact it still thinks both players are wide ranging. 1M - 3N is different, where GIB passes only if it has exactly 5332 and 12-14 points.

Joshua Donn, who was the GIB developer who would interact with the forums many years ago, said in 2018: Then he left and hasn't been back since :) But in this auction, 4NT isn't Blackwood anyway. I think the problem here is that the bidding database is very poor when it comes to slam auctions. Which isn't a great surprise as most grands need to be bid based on counting tricks rather than points / rules - which its what its simulations are for. But when it simulates, it plays out the rest of the auction using database rules only, without knowing that it'll be able to simulate to get to a better result later (recursive simulations are too time consuming). If it starts with 4♦ the database will probably tell it the auction will end up at 6♦ even when 7♦ is stone cold. So when 7♦ is making a lot of the time, it decides blasting to grand will get a better average than the database, with all of the making grands outweighing the ones where it goes down. But that's just a guess; the old version of GIB is far weaker and North bids an immediate 6N over West's double of 2♣ :) Even if I hack the database a bit to get it to match yours, it still picks 4♦, so there have probably been a lot of changes here.

Yes, it distinguishes equals in situations like the opening lead based on its lead conventions: My earlier post on analysing why it randomly breaks these conventions may be explained by this 'weighting for signalling correctly' parameter that I wasn't aware of back then. It definitely doesn't play attitude discards though, so you probably just notice it when it happens to be true.

If you look at the 'GIB Release Notes' thread at the top of this forum, you'll see GIB underwent regular updates from 2011 to the start of 2019 - on average they were released about twice a year (more regular early on). Then all development seemingly stopped and nothing has been officially heard since. This was primarily bug fixes, but also new conventions, etc. The only known change to the play logic over that time period was in 2014 where GIB was forced to cash an ace against 7NT. I believe it's written in C, though the bidding database itself is a custom format, very incomprehensible without a lot of study. Based on some very old documentation, there's one slightly more readable version of the bidding database that is then processed by some search + replaces to give a more raw version that GIB uses. In the raw version, the rule that told a 2012 version of GIB it was OK to bid 3♣ in this recent thread was: 0d e[12]g.:-:a[+e]b[^N]~b<=g&(a)&&(#0<0||(#0<H&&b>D))&(3,13)<25&&(#10(12,b)<4&(3,b)<4)&&(#0>=0||(((m=#3(N)>=0)*#0(2,m)+(m>=0)*(n=#3(m)>=0)*#0(2,n))<9&(#5(#0(2,C),#0(2,D),#0(2,H)),#0(2,S))<6))~$$ 27 @114#b~(#1<0||(#1<H&&b>D))&(a,b,2)~ ~#0(a,b,2)~ #532 The bidding database uses fixed rules, so a basic robot will bid exactly the same way with the same hand (can vary based on vulnerability when it comes to preempts / two suited overcalls), though the "rules" are far more complex than the ones you see when you click on the descriptions of bids. Advanced robots are allowed to run simulations at certain points in the auction (though there are some bids marked as 'not allowed to simulate'). It selects some alternatives from the bidding database that a robot might have bid with a different hand in its position. Then it builds a set of deals, and for each alternative, finishes off the auction under the assumption all bids from there strictly use the bidding database. If a bid works out better than the book bid, it'll make it. That leads to some 'random variation' in bidding. Play is fully based on simulations so there will always be random variation in that sense. In the very original version of GIB that Ginsberg wrote, there was a parameter for 'weighting given for signalling correctly', so it is likely there is some random variation in the 'unimportant' cards it selects to prevent it being too predictable.

It depends what you mean by fail. If succeeding is never having silly definitions for bids, certainly there is no hope.. But if a particular bug occurs time and again when playing with the robots, and gets reported numerous times on the forum, and then suddenly it's fixed.. that increases the enjoyment of playing with the bots immensely. Failing is not even getting a response 6 years after reporting a problem.

East's bid is based on a simulation too, so the human construct of "downgrading" is included in that. From East's perspective, simulating hands where West has 12-14 points and at least a queen in diamonds, and South has those clubs, still shows game making almost every time, so it's not even close to a reject. It's true that GIB can be a bit aggressive, since East's simulation doesn't take into account that West may have simulated too.. and conversely, West's simulation assumes East will only accept when it has a book-accept.. but there's really nothing you can do about that.

Not sure there's much to this one. 2♦ is an advanced robot bid; if its simulation of hands where you have rebiddable clubs finds that a game is worth a shot if partner has an accept of a diamond game try (it surely increases the odds, knowing your club honors are less valuable to you), then I can't argue with it. With better breaks it might have made..

GIB never redoubles, the basic bot will bid a non-forcing 3♣ even with an ace extra (while the advanced robot would just bash out 4♥ and hope for the best). GIB has a set of rules telling it how all of the continuations are defined after partner accepts the transfer with 2♥ (and separate rules after a superaccept). This section doesn't include any rules for what to do if partner *passes* the transfer (despite the fact North has its own rule saying pass with 2 hearts, accept with 3 hearts). So as with most competitive auctions, it falls back to a set of default rules that have nothing to do with transfers. There is a rule which says in general auctions like this, a new suit at the 3 level shows game forcing values. However, it has another rule that says if you're forced to bid and have no fit, a cheap new suit doesn't show anything. It combines the two, making the club bid but unable to promise any points. Often fixing robot bugs is difficult because a slight change can break a bunch of other sequences. This one would be pretty easy to fix by adding a few new rules for continuations after this double that promise strength, after which default rules for later bids will work fine.

It looks like the old 1-character bug has been fixed - yay :) Unfortunately, an extra line of code was added at the same time which is causing more regular issues. Here's a silly hand diagram demonstrating the problem clearly, though it actually pops up on nearly every hand. [hv=https://www.bridgebase.com/tools/handviewer.html?a=1NPPP&d=e&s=SA32HAKQDAKQcAKQJ&w=SKQT9754H432D32C2&n=S86HJT98DJT9CT987&p=SAS4S6SJS3]400|300[/hv] Click through to after the lead to trick 2 and click GIB to see missing results for the T, 9, and 5. Specifically, any time two cards that weren't originally equals now are due to the intermediate cards having been played on earlier tricks, the lower cards never show a double dummy score. This is due to a new line of code which been added (wasn't there last time I checked) - the long line with the comment here: The server request only returns scores for cards that really aren't equals (SK and S7), with this loop in the handviewer code meant to apply the same score to the remaining equals. But the commented line specifically tests if the next card was played by another hand on an earlier trick, and stops adding scores. It's only meant to stop if the card hasn't been played, or was played by another hand to the *same* trick, which is what the later inThisTrick part does. So that line of code shouldn't be there.

Yes, advanced robots only both pass and 4N when simulating and choose between those two options. But over 4N and without the double, West does bid clubs, so that part did get fixed. Just not the equivalent with a double.

https://webutil.bridgebase.com/v2/tview.php?t=ARDCHALLENGE:94c0c1ca.1a25.11ed.897d.0cc47a39aeb4-1660298423&u=smerriman&v3b=web&v3v=6.3.8 smerriman 11.5 - 4.5 shuba

https://webutil.bridgebase.com/v2/tview.php?t=ARDCHALLENGE:1f23cca3.17bc.11ed.897d.0cc47a39aeb4-1660033226&u=smerriman&v3b=web&v3v=6.3.8 smerriman 10 - 6 Povratnik

The challenge times out 3 days (not a week) after the challenge is first issued, whether you have started or not. Note this means that if you issue a challenge, and it takes the other player 2 days to accept, it'll disappear after one more day. (For robot challenges I believe it's one day rather than 3.) Is it possible therefore that this is your issue after all?

Amusingly, 10 year old GIB would make exactly the same 4NT bid.. but its partner would *always* pass, even without the double. It had no followups programmed at all, and thought that 5♣ or 5♦ (by West, or even by East if doubled and passed back) showed enough combined strength for game. In one of the last GIB updates in 2018 they fixed some other sequences involving 4NT bids (like unusual 2NT followed by 4NT) which GIB would also always pass. My guess is they added in a rule for 4N P ?, and forgot about 4N X ?.

Which part in particular were you wondering about?

Very odd - GIB definitely used to have a simple rule that told it to bid 1♥ here. Who knows what changed in the last 10 years.

This is a well known GIB flaw. GIB can handle offshape doubles if the opponents stay silent, and partner makes a simple suit bid. But if partner shows strength like here, or the opponents compete, none of the followup are programmed correctly, each of them locking in the classic takeout double definition. So you basically have to avoid using them as much as possible.

I agree - they aren't opposing statements :) Most + some = all.

https://webutil.bridgebase.com/v2/tview.php?t=ARDCHALLENGE:2427277e.15be.11ed.897d.0cc47a39aeb4-1659814191&u=smerriman&v3b=web&v3v=6.3.8 smerriman 10.5 - tim_ucin 5.5 https://webutil.bridgebase.com/v2/tview.php?t=ARDCHALLENGE:14dd0c47.16e3.11ed.897d.0cc47a39aeb4-1659940008&u=smerriman&v3b=web&v3v=6.3.8 smerriman 8.5 - 7.5 muddylane

I'm going to answer various questions. Ignore ones you don't want to know the answer to. Should GIB beat the contract regardless? Yes, of course, West should overtake the heart Ace and play another one, but with double dummy analysis it never knows the difference. Why does East continue with a heart here, and not if South responds 1♠? With robot logic, East reads nothing into the fact the heart king won the trick; anyone could have the heart ace. In the cases where it matters what East does, they have to switch to a club if West has the club ace, and a heart if West has the heart ace. (There's no chance of a heart ruff like you suggested, since that would give North a 1♥ opener instead) In your auction, North has shown honors in clubs. This makes it very likely North has the club ace, so a heart is a clear favorite for the robot. If South starts with 1♠, North will rebid 2♦. I'm not sure how the auction got to slam after that, but suppose South jumped to Blackwood and then went to slam. Now North hasn't said anything about clubs at all. There are more clubs left in the deck, so West is more likely to have the club ace than the heart ace. Thus a simulation will tell the robot the club is the best choice. Could I have predicted any of this when deciding what to bid? No. Should I bid Soloway with this hand? Without shortness, it's really too weak. GIB is extremely unlikely to be raising your spades (some people even play that you only raise with an honor), so by jumping, you're giving up on the chance to play a spade contract (diamonds are fixed as trumps unless partner raises immediately). Your spades are so nice here - and diamonds weak, along with the two doubletons - that I'd want to play in a spade contract most of the time, rather than diamonds. 1♠ will give you a chance to do so.

The bots can make some very bad bids, and there are lots of examples in this forum. Humans playing with bots can also make some very bad bids, and think the bot has messed up when it hasn't, and there are lots of examples of that too in this forum. Without posting specific hands, there's no way of knowing which applies to your situation, but note that even when playing with humans, you can't "sacrifice" over the opponents' preempt - overcalling shows a good hand.

I'm with AL78. If the diamonds are 3-7, the chance of picking up the heart suit for 3 tricks is a tick under 50%, and that's not considering ruffs. Maybe clubs 3-3 gives us a tiny extra chance, but looking at the two hands, I'm happy with 4♠. Though of course, that's only looking at the two hands. Actually bidding it out without knowing each other's hand could easily go either way.