smerriman

There are lots of places where doubles mess up auctions unfortunately. It's nothing to do with "forgetting", but simply that the database defines continuations to 3♣ P rather than 3♣ (P|D) like it should. Chalk this one up as another easy to fix bug, but bugs aren't being fixed so you have to live with it :( The latter is a non-issue IMO. Law 41d was not written with graphical interfaces in mind where screen space is limited (esp on mobile, which uses an identical GUI - the old standalone desktop version used to lay out cards in separate columns) and where it's possible to provide different views to different players. It was written to prevent inconsistency from one hand to the next. The fact that card symbols are rotated so that you can read them is also no doubt illegal, but sensible. The current layout - left to right, top to bottom - makes more sense to me online than what the law would say, and much of the law still needs fixing for online play.

If your partner held the ace of spades instead of a void, they would respond 5♠ and you'd be in exactly the same awful contract. Sorry, but 110% of blame to South here. If you truly believe North is at fault, post this as an ATB on BridgeWinners and if you don't get a unanimous response, I will be stunned.

I think it's unfortunate they're both called lebensohl as I wouldn't even consider them related conventions. In one partner is locked to a very narrow 15-17 range, so you know when to game force and sign off. In the other, partner could be anything from a 10 count to unlimited. If 3 level bids are game forcing when partner could have the minimum, it seems unplayable.

That's a page about responding to a 1NT opener. There a 3 level suit response is clearly game forcing. But unrelated to responding to a takeout double.. Edit - found Larry talking about lebensohl over weak 2s here, where he has the standard non-GF definition of a 3 level response.

But most continuations have no equivalent. For example, after an initial 1 suit response with a game forcing hand: - raising opener's second suit normally shows 4 card support. That's impossible here - does 3♠ show 3 cards, or something else? - you can also rebid your own suit or bid the fourth suit as an artificial force. Neither exist here, so what are the precise definitions of 3♣ and 3♦? Do you also give up the natural 2NT invite - almost always unnecessary after a 1 suit response, but almost always what you'll actually have after a 1NT start?

Actually, the bigger problem is that the thread there, while excellent in itself, skips over reverses after a 1NT response entirely, and they're actually quite different from reverses after 1 of a suit. I haven't seen any definitive coverage of the 1NT case anywhere. The answer to this specific case will also depend on what 1NT means (forcing or weak). With a known lack of a spade fit, and a mainly wasted queen of clubs, I don't think I would reverse on this hand. But it's still worth defining what responder's continuations mean.

I don't even see how the first condition is met. The only way declarer can go wrong is by believing South has J, 9, 9, as he did. With that hand, South has the most trivial decision of any of the situations discussed. If South is the type of player who does not need any time at all to remember all played cards, then South will absolutely have been aware that a change in tempo would *not* work to their benefit. Even though it did*. But I don't think East is ever going to able to make a case that South knows East well enough to conclude he would make a completely illogical conclusion. *In fact, it's not clear that the hesitation did work to their benefit, since East's reasoning that South would make a poor play of throwing a low spade sounds like he would have made the wrong decision regardless.

Argine's 4♥ feels like a homage to GIB's biddable suits :) [hv=https://www.bridgebase.com/tools/handviewer.html?lin=st||pn|smerriman,Argine,Argine,Argine|md|4S93HADK8CAQJT9872,SKTHKQ8DQ9754CK54,SJ852HT76532DJTC6,SAQ764HJ94DA632C3|sv|e|rh||ah|Board%206|mb|1S|an|4+!S;%20HCP%2011-20;%20natural|mb|3S|an|HCP%2010-17;%20Gambling,%20solid%207-card%20minor%20and%20asks%20for%20stopper%20in%20!S%20-%20Forcing|mb|D|an|HCP%206+;%20penalty|mb|4H|an|6+!H;%20HCP%2014-;%20natural|mb|P|mb|P|mb|D|an|HCP%206+;%20penalty|mb|P|mb|P|mb|5C|an|6+!C;%20HCP%2010-17;%20natural|mb|D|an|HCP%206+;%20penalty|mb|P|mb|P|mb|P|pc|SK|pc|S2|pc|S4|pc|S3|pc|ST|pc|SJ|pc|SQ|pc|S9|pc|SA|pc|C7|pc|CK|pc|S5|pc|C4|pc|C6|pc|C3|pc|C2|pc|DJ|pc|D3|pc|D8|pc|DQ|pc|HK|pc|H2|pc|H4|pc|HA|pc|CA|pc|C5|pc|S8|pc|S6|pc|CQ|pc|D7|pc|DT|pc|S7|pc|CJ|pc|H8|pc|H3|pc|D2|pc|CT|pc|D4|pc|H5|pc|D6|pc|C9|pc|D5|pc|H6|pc|H9|pc|C8|pc|D9|pc|H7|pc|HJ|pc|DK|pc|HQ|pc|HT|pc|DA|]400|300[/hv] Also had the opponents open a weak 2♦, only to find a 3♦ overcall was natural. I've forgotten how unusable non-RKC Blackwood is too. Whose idea was to teach that to beginners? Still, it's fun to play with a new system. And how nice it is to see some actual signalling?

Tricky. Usually the answer at MPs is to not make safety plays, but if spades are 5-3, then the diamond is more likely than not to be offside. So I think I'm going to take the IMP line anyway and duck twice, then play ace and 9 towards the jack.

I guess the real question is, where did you get the lin file from? It appears BBO's handviewer supports suits and honor cards being in lowercase, but BridgeSolver does not. And your lin file has them in lowercase. Everywhere that I can see BBO providing lin files, they're in uppercase. So either: - you used a third party tool to extract a lin somehow, and that tool is at fault - some area of BBO provides them in lowercase, and that area of BBO may be at fault - lin is supposed to support lowercase, and BridgeSolver is at fault Edit - ah, yes, it must be a glitch specific to challenge-a-robot, as the hand records with GIB at all 4 tables do indeed result in all of these being lowercase, while all other records throughout BBO, including the human side of the same challenge, do not.

I know IMPs is "all about making the contract" but I really can't see this providing enough information to make up for the guaranteed IMP you'll be throwing away when the club queen does drop.

It's a complete guess, unfortunately. While there is some truth to the idea that GIB doesn't bare a king, it's only because it assumes that when it does, you'll know to drop it - while other lines often are theoretically better double dummy. You would think therefore, that GIB must have kept its spade king protected, because you might hold Axx in spades. But no, to GIB if you held the Ace of spades, you would be too strong to bid 4♥. So it actually places that card in its partner's hand, making its choice of discard irrelevant, and thus made at random.

Whenever there's a question like this, Richard Pavlicek usually has an answer somewhere on his site :) http://www.rpbridge.net/8j25.htm 33 and 37 probably came about as a simple guideline that you're not off two and one aces respectively (with AK of the same suit much less likely).

Oh my.. it's not GIB.. it's Argine..

Um, wow? This appears to be gigantic news? https://doc.bridgebase.com/Help/Acol-robot-system-notes.html But why this of all things? The date doesn't appear to be April 1..

Yes, of course - but that's what I was referring to earlier about it being too late by then. With lots of distribution, there is almost always an overlap (or underlap would be a better description) in ranges - too weak in high card values to force to game, but too strong in distributional values for all of the possible follow-up bids. There isn't a possible way to show extra values the second time, because the assumption is with more values it would have game forced the first time. It thus only has weak descriptions available. Simulations should fill these types of holes.

Don't know what the exact values used here are, but if, for example, you defined that you should normally have 9 HCP to game force, then a robot without simulations has to show it as weaker. The equivalent of "obviously this is stronger than that" or "does not fit the hand" to a human is simulation for a robot, so with that off its rules are often generally quite useless when it comes to these types of hands.

I did warn you back in August via email not to hope that, because it wasn't going to happen :)

Bidding table uses free robots only. Teaching table uses the ones you rented. Not surprised at all that basic robot would get this wrong, since they're all about points and never can bid properly with lots of distribution; "seeing" a game is good relies on simulations. But more surprised that advanced robots actually do the same here. Old version of GIB used to simulate its first bid and could comfortably see it's better to game force. (If you forced it to start with 2NT, it would bid the same 3♦ the next time, because it's forced to do so - doesn't have other options built in at that point). But current version of GIB appears to have simulations now turned off for the first bid (but the second bid was still forced, so it was too late by then.) No idea why BBO have weakened the robot since the old version. Maybe too much server load.

It's not actually anything to do with an abused bar bid convention. The book bid here is actually pass over 1N, which is fine since of course the robot will always figure out it's better to bid diamonds instead. So it's not bidding 3♦ because it thinks it actually has a bar bid, but because it thinks that faking a bar bid will lead to a better result than bidding 2♦. When I run this with the older version of GIB, it bids 2♦ most of the time, precisely due to the fact that this will occasionally get you playing a better 2♠. But it does bid 3♦ occasionally, because in those simulations, a few of the hands when it went low, the opponent robots actually got into the auction by competing in clubs and you ended up worse off. With the opponents cold for game in your deal, maybe GIB's bid was genius :) But probably just an edge case in the long run.

I know I wrote this after you posted this thread, but I'm quoting it just because this is a really good example: GIB has been programmed that fourth suit forcing shows 13+ total points, and it only has 12. It also has a well known poor priority of rules for giving preference to diamonds over clubs, and of course doesn't match rules for other bids like bidding no trump. This hand happens to be one that falls through the cracks where nothing matches. But that's not really a big deal at all, since after the database tells it to pass, the paid robot runs various hands for South and gets to pick from all of the above options anyway and see what works best. Running this through the older version of GIB 40 times, it chooses 3♥ 35 times, an invitational 3♦ 4 times, and pass once. Harder to test the frequencies with the up-to-date robot because teaching tables lock in a fixed seed - there I actually do get a pass both vulnerable, but 2♥ at others - it may be that the seed hit the 1 in 40 shot, or something else is going on. You could reduce the lower limit on 4SF but 1 point and this hand would no longer be a problem, but then you'd easily find a slightly weaker hand (with poorer placed honors) and could post that as a bug that it was too weak to bid 4SF. And then you could reshuffle the priority for preferences, which should probably be done anyway, but all in all the robot should virtually always get this one right, so all those changes aren't likely to achieve much.

Sure, something is probably off by a point somewhere. There are lots of places in the database where it may come up with the wrong bid (a more common case is where there are two possible options - like pass or give preference to an earlier suit, and the wrong one has slightly higher priority.) But GIB was never designed to work with half its algorithm disabled. That half is deliberately intended to fix small issues on the fly. In the latter case I mentioned above, I expect it's even intentional that no effort was put into prioritizing those two continuations, as a lot of decisions can't be based on simple facts like points alone and you know that GIB will get to choose between both options on a case by case basis. Database issues like this only need fixing if there is an *earlier* bid in the sequence where simulations have been turned on - because it uses the database for the followup. For example, if GIB in South's position is told to simulate over the 2NT bid to decide whether 4NT is a good choice, then it would be important to ensure the database accepted correctly. But that's not a simulation spot, which is why there is zero reason to change the database here. Other than to placate people using a free service, but I think if BBO ever got back to improving GIB, paid customers would be the only ones worth considering. I've been assuming most of your forum posts have been with the proper robot, but if not, that does nullify quite a lot of them. (Obviously many are still insane as the above situation doesn't apply.)

This only ever happens in the anonymous games, never playing when logged in. I think the anonymous games just use a lesser server, and sometimes some data gets lost along the way.

OK - you should really specify if you're using a free bot. Like discussed in similar threads in the past, there is 0 upside to changing the book bid here, and the advanced robot gets it right, so there's nothing to fix.

Once again I can't replicate this with advanced bots.. passes at all four vulnerabilities. Are you sure you're not playing with the free bots?