smerriman

Your aim isn't to make the contract - but to do better than everyone else who is in exactly the same position as you. So if you have two losers in a just declare daylong and the opponents cash them straight away, you're not going to get a bad score - you'll get the same average 50% that everyone else will. It's true that if you want to 'win' a daylong, you won't be able to do so if you get dealt one of those flat 50% hands, because someone else will likely have fluked hands where they can get 80-90% on lots of them. But that's more luck of the draw, rather than anything to do with skill / making an 'easier' daylong. What you should do is just focus on your score for each hand individually, and not worry too much about where that places you on the combined leaderboard.

In an earlier post I debugged my way through a hand where robots made a 0% play on defense. The cause was the surprising failure to calculate double dummy scores accurately. Today I remembered I had seen one other 0% play reported on BridgeWinners - this time by an advanced robot declarer. Here was the original hand as posted: [hv=https://www.bridgebase.com/tools/handviewer.html?d=w&v=b&a=1S2H4SXPPP&sn=Human&nn=Human&en=Human&wn=Robot&n=SA42HAKJT4DJ87C93&e=SJ96HQ9DA965432C5&s=S75H6532DKTCKJT86&w=SKQT83H87DQCAQ742&p=HAH9H5H7HKHQH2H8C3C5CTCQCAC9D2C6DQD7DADTD6DKS8D8C7DJS6C8D9S7STSAHJD3H3S3C4H4SJCJS9S5SKS2C2S4D4CKHTD5H6SQ]400|300[/hv] After 10 tricks, you have a trivial cross-ruff to take the remaining 3 with high trumps. Instead, GIB plays a trump, which guarantees the contract going down on any lie of the cards. Not as much debugging output is available for declarer play in the 10-year-old command line version of GIB, so I didn't expect to make much headway with this one. You don't get to see any of the hands that are being simulated, and GIBson's single dummy algorithm is more complicated to understand. However, you do get some insights, such as what opponent's holdings GIB is "playing for" and how many tricks it expects to take for that holding. After 7 tricks, GIB correctly calculates that the contract has no hope of making with accurate defense, with 9 tricks the limit. But when human North errs by ruffing with the Ace, GIB.. *still* thinks the limit is 9 tricks, and continues down the path it calculated to its desired goal of 9 tricks. Now, you're probably thinking - this doesn't sound remarkable. Just some weird glitch where it can't see a winning line has arisen. There's more. While debugging this in the command line GIB, I was messing around with going back and forward a trick or two and trying different options to see if it would ever come up with the winning line. So I'd undo a few moves, try something else and so on. And I found something strange - after GIB concedes the contract with the trump lead at trick 11, I undid a trick and repeated it - and it discovered the diamond play for 10 tricks. If I undo back to the point before North erred, and then go back forward.. it can only find 9 tricks. And most importantly - this is replicable today in a teaching table on BBO. Predeal the above hand, put an advanced robot in West, and play through the hand matching the cards above from the other three seats. It will throw the contract at trick 11. Hit undo once.. and it'll suddenly be able to make the contract. Hit undo again once.. and it'll still play the diamond. No matter how many times you try this (clicking undo once only each time), it'll find the winning line every single time, so it's not randomly switching between them. But hit undo several times in quick succession to get back to before North plays the Ace, then play forward again - and it'll be back to the guaranteed failing line. GIB is clearly using cached information that it shouldn't be. If a simple undo is enough to clear the cache properly, then this is trivial to fix.

But the key is finding grand when you have the right two aces, so those numbers probably aren't that meaningful.

No, the odds of seeing the same hand are very low.

I don't see where we denied anything in diamonds?

Ah, OK, I wasn't sure on what 4NT would show. In that, assume partner bids 4NT.

That's exactly how I'd start too. Partner continues with 4♦, and you presumably bid 4♥. Partner doesn't have the ♠A so bids 5♣ (turns out they have the ♠K, but knowing you're short I expect it's right to skip a second round control). 5♠ seems the only appropriate continuation, but I always get a bit lost at this point when it comes down to a queen or jack here or there..

[hv=https://www.bridgebase.com/tools/handviewer.html?d=w&s=HKJ9DQ92CAKQJ862&v=b&a=p1NP]200|300[/hv] You probably weren't expecting your partner to open 1NT (15-17) when you picked up this hand. How do you find out everything you need to know?

Firstly, basic robots never simulate during the bidding, so are guaranteed to bid the same. Only advanced robots simulate during the bidding. Secondly, the seed is only equalised during robot tournaments, not for the main bridge club. So two options: a) (Most likely) One table was using a basic robot, one had paid for an advanced robot b) Both tables were using advanced robots, and simulations differed due to the seeds not being equalised.

If the auction is anything like the one you described, then I suspect this was your mistake, not the robot's. Your bid of 2♦ shows a weak hand (~6-9 points) and doesn't promise diamond support.. anything after that is likely to be very weird. On top of that, if GIB had 7 diamonds and a singleton spade, and 4 clubs for the 2♣ bid.. then they're never bidding 2NT; they're certainly going to rebid diamonds, and you still wanted to play in spades. So you probably miscounted the cards as well.

https://webutil.bridgebase.com/v2/tview.php?t=ARDCHALLENGE:43bea1a4.121e.11ed.897d.0cc47a39aeb4-1659415671 https://webutil.bridgebase.com/v2/tview.php?t=ARDCHALLENGE:793afde5.10dd.11ed.897d.0cc47a39aeb4-1659277892&u=smerriman&v3b=web&v3v=6.3.8 smerriman 7.5 - 8.5 gib smerriman 7.5 - 8.5 garant_7

The diagram shows you in South. Are you GIB :o OK, I assume you were disconnected, and a basic GIB took over. GIB doesn't make assumptions about what a play does or doesn't show; finessing is the best at trick one, but on the next trick it considers East ducking from Kx just as likely as anything else. Matt Ginsberg's original paper specified that it should calculate conditional probabilities of whether opponents have certain cards based on whether a certain play earlier on would have been an error, to handle situations like this. But apparently it made the program way too slow, and it was taken out (on reflection, not too surprisingly; if GIB often struggles to deal any hands that match a bidding sequence, and believes you way too much when you do something a tiny bit unusual, imagine if you restrict it further by removing a whole heap every time someone plays a non-optimal card..)

Descriptions are entirely meaningless at that level. Double says "double dummy simulations say we're going to beat this enough times". But East is incapable of interpreting this.

https://doc.bridgebase.com/BiddersHands2022-9.pdf Found by taking the URL from last time and changing 8 to 9 :)

Why would anyone defend it? Of course it's silly for 5♣ to show 29-30 points, and it's a bug that has been brought up in the past, but you can't change how GIB works. All you can do is check the description before you bid it, see it shows a silly hand, and give up on the idea of sacrificing. (Or in this case, if you've decided you wanted to sacrifice at the 5 level, you can bid 4NT instead of doubling diamonds.)

Final: bgm vs kgr The final consists of 4 x 16 board challenges. The deadline for round 2 is Tuesday, 6 September, 11:59pm EST. I've updated the spreadsheet for round 2. Have asked Povratnik the status of his robot challenge - will put crazy4hoop through if I don't hear back within the next day. https://www.dropbox.com/s/9h1jov60prn2j4a/Event%2024.xlsx?dl=0 Complete one 16 board, non-best-hand, MP challenge against everyone in round 2 who you haven't yet played. Do not check the Advanced Robots checkbox - but all friend challenges will automatically be upgraded to advanced robots. Use "challenge a robot" to complete the challenge against GIB. This will not be automatically updated to advanced robots, but you can choose whether or not to do so as you wish. If you challenge GIB multiple times, only the first one will count. The top 9 in each pool will carry their scores forward to round 2 where those 18 will finish off a round robin against the other 9, before the top 4 will play in a knockout format. The deadline for Round 1 is August 19, 11:59pm EST. As always: - If a challenge hasn't been accepted in a day or two, cancel and reissue it to make sure you don't time out after the other player accepts. - Please only accept a challenge if you are positive you will be able to complete it in time. Partially completed challenges are annoying for everyone. - (At least) one player should report the scores of each challenge here by pasting a results URL. You can get the URL via History the tab, clicking Results, then the icon at the top right to open the results in a new window.

Hiding because this is a little offtopic to what the real question was:

Studies about double dummy have shown that on average, humans do slightly better than double dummy in all cases *except* for slams, when they do worse. This is for fairly obvious reasons - the biggest gains come from suboptimal leads, which more rarely cost at higher levels. And the idea that declarer guesses balance out defender guesses doesn't apply when defenders aren't going to have the lead. I forced East to have a balanced 20-21 with AKxx of trumps (surely the best case) and it was still only making slam ~40% of the time double dummy. I then looked at the first 10 of these hands manually. All 6 of the cases where it was down double dummy would be down in real life - but of the other 4, half would have gone down with normal play, which put it closer to 20%. For this specific dummy, I would expect larger datasets to show similar results, well below the levels required to bid it.

Right - your post sounded like the fact you can make game/slam was relevant to whether you should open, where it's not really the important aspect at all. If you provided just one hand and asked whether you should open, that's a better question. (And I would, but not in 4th seat.)

It is meaningless to provide *both* hands when asking questions like this. If I were in fourth seat with the second hand, then there is a chance that partner has the absolute perfect cards and we're cold for game.. and there is also a chance that the final contract will end in the opponents making some number of spades. The question is not whether the first chance exists.. but which is more likely? Of course, some people will open every 11 point hand, and that's fine too.

While a nice idea in theory, cheating would be so trivial I would find it hard to believe any pairs near the top of the leaderboard weren't doing so. If you rent robots and play in the main bridge club with your partner, you are compared against other tables renting robots too (mainly just a human in South, but others may have a human in North too). So that's the easiest way to play with a partner against robots.

The hand in question shows how inappropriate double dummy is in this scenario.

Bumping with 1 day left just in case anybody missed this and still wants to play.

Mike covered all the reasons things could go wrong if partner doesn't superaccept.. but here even when partner superaccepts, slam is pretty hopeless, even though it makes double dummy. So there seems to be no need to even give partner a chance to do so.

I'm in the 2NT forcing camp, a la the regularly referenced thread on BridgeWinners, so that you have an extra step to describe your strength. If I don't want to be in game opposite a maximum weak no trump, I wouldn't show an invitational hand on my first bid (just like I wouldn't in other auctions).