bluecalm

Isn't it completely obvious to play E for Q♦? I mean he has more diamonds on average and that should be the end of analysis. Seems so basic that I am afraid I am missing something :) He may still be 3-2-5-3 I think. Or maybe he realized that if partner had J♦ it doesn't matter what he pitch and if he doesn't then small diamond may well confuse declarer. Wouldn't he play an 8 from AT8 ?

I always thought those are unplayable at matchpoints because you end up in 4-3 minor fit instead 1NT on many sequences. Am I missing something here ? Also does it mean you automatically bid 2S with 3-5-(2-3) ?

It's forcing. 15pc + 8pc = game for them. Their system is such that you have two ranges for gf hands: (15)16-18 and 19-21 and can relay for shape/range. If you have 14, you bid 2H if you have 15 you are in GF auction opposite 8+. If you have 7 hearts and invite (or some 6-4 with KQJTxx or w/e) you bid 3H after 1S. I guess it may not be very good at matchpoints to automatically be in game with 15-8 and 6hearts but you can tighten 2D relay up if you are worried about it (and precision have the same problem anyway). If it goes: 1H - 1S 2C - 2H You can still decide if you have an invite opposite 5-7, most of the time not really and you play level lower than you otherwise would. Also, to make the system complete: 1H - 1S 2C - 2D 2H - 3H = invite, so: 1H - 1S 2C - 3H = slam try with support (game hand bid 2D and then 4H after weak response)

I have found AMBRA write-ups available having issues. Top Italian pairs certainly don't play the AMBRA way and I think for good reason. If it starts: 1H - 1S 2C - ? and you have a weak hand it seems natural to play 2N as minors (4-1-4-4/4-0-4-5/4-1-(5-3) and 3m as 6carder analogous to 1H - 1N and 1S - 1N. Pass. One needs to do the math but my feeling is responder has 6 spades most of the time as you only bid 2S having 5 spades with 5-1 majors. Even if it's not most of the time it's still tragic to bid if responder has 6S and finding a better contract is not guaranteed opposite 5S. No, of course not. 2D is 8+ GF opposite strong hand so any 15+ bids somethign different here.I mean 2H is 11-14 5H-4+C and other bids are strong. 15 is somewhat borderline I guess you can bid 2H with 15 but no way you do it with 16 let alone 17. 3H promises 7 hearts. At least Italian pairs play it this way. The whole point of Gazilli is to avoid jumping to 3 level with 6 carders and invitational hands. As to the structure: 1H - 1S 2C - 2D I can give you what Lauria - Versace and Sementa - Duboin play but those are relays for exact shape/strength. I think simple structure would do: 2S - 15+pc, 3S 2N - balanced without 3S (could be 2-5-(4-3) if localization is right) 3C - 5H-4C 3D - 5H-4D 3H - 6H without 3S or 4m 3S - 4S-5H 3N - 18-19 2-5-3-3 It seems like there is too much space as you still have some after 2N.

But those numbers have nothing to do at all with how much the system is worth. Look at this situation: Pair A never bids 50% slams and always bids 50% games. Their cardplay is perfect. Pair B always bids 50% slams and never bids 50% games. Their cardplay is such that they lose makeable contract one time in 10 and it costs 10 imps every time. Now if we play a lot of hands with 50% games and slams vast majority of swings will come from bidding not cardplay. In fact cardplay is as good 9 times out of 10. If you apply author's argument here you will end up with some non-sense like bidding contributing significant % to total pair A's edge. Still, cardplay is 100% of the edge of pair A. The author just measures the wrong thing. When you try to approximate something at least give argument which makes some sense and works on simple example data.

I think that the key to good bidding program is fast function to estimate how often the contract make given two hands. It doesn't need to be very accurate, just accurate in average case. People can do that without long analysis usually so I think could a computer. I have some ideas for this but maybe this work is already done/started ? Is anybody aware of any efforts in this direction ?

Imo the author's argument is weak total junk. I don't want to quote the whole chapter here but his reasoning isn't even a serious try to quantify anything. He makes an argument that on hands with the same contract "cardplay swing" is 1.76 on average while average swing on any hand is 4.1 so from that follows that cardplay accounts for less than 50% of swings. What is worse he then goes from that to conclusion about how much bidding contribute to total edge of given pair. I mean, can I have some of what the author is smoking ? :) One fact which almost all authors ignore is this: Just because something contribute to most of the swings doesn't mean it's significant part of pair strength. It might be just variance. Of course the biggest swings come from bidding but the difference in EV is often very low. If you bid 52% slam or not or if you bid thin game which happen to make opposite actual partner's hand (but wouldn't opposite many other possible hands) doesn't matter much. Reasoning as: "average cardplay swing is 1.76 and average swing is 4.1 so bidding contribute to more than 50% of pair strength" isn't "food for thought" it's just nonsense. Misleading nonsense to be sure.

What I tried to do is to take all hands played by given pair and calculate double dummy result after the 1st lead and then compare it to minimax on given hand. This way you account for revealing bidding and it *SHOULD* more or less have similar EV as real level of bidding. Variance is obviously huge though and not all things are factored (ie first lead is, subsequent defense isn't). I realize this method is far from perfect but at least it's a try as opposed to: Which I agree is just nonsense. Also Fantunes played much more hands than 2700 on vugraph. I have 7119 hands right now on my computer and I guess that's not all of them as I haven't downloaded 2012 (and end of 2011) hands at all. Laughable sample size and probably bad methodology. Every argument like that I saw for bidding judgement being the most important thing at high level suffered from the author not understanding probability and being very result oriented ("uh-oh they bid 3NT here which is 50% contract but 4S is only 15% contract and 3NT maked so +12 for judgement!")

S would certainly bid 5♠ though :) As to the auction. E has no-brainer double. W has some decision to take and I think bidding on is not obvious although it seems to be better action to em. I have nothing to support this claim, just intuition.

Maybe you don't need full single dummy solver to estimate the best play. I mean even humans can do that on some hands on a good day :)

How good is GIB at constructive bidding ? I mean, if you play with it is it clear it's worse than your average expert ?

So I am thinking something along the lines: Keep all possible combinations of hands of opponents grouped by categories (say shapes). There will be like only 240 possible shapes and they will be pruned quickly as informations from bidding/play are included. Now for every possible shape there are possible combos of suits so in every shape thing you keep possible combos of spades/hearts/diamonds/clubs. It's still small so far. Now you add a number which represents number of all combos for given shape (you multiple number of combos for every suit for each shape) and sort the whole thing from the biggest amount of combos to the smallest one. Now you add some weights. Say if opponent leads 3/5 and you see 2s you assign very small weight to every combination of spades with 2s where it's not systemic lead and keep weight unchanged for the ones where it's systemic lead. You do the same for shapes with stiffs if the stiff is not lead etc. etc. After applying those rules you sort from the most likely shape to the lest likely one and start running simuls for the most likely hands opponent might have. After every trick you repeat the process again applying the rules and again sorting to find the most likely hands to run simuls on. I have no clue how gib works but my experience from writing some poker simulations and solving double dummy problems on huge databses tells me it *MIGHT* be computationally feasible to construct system like that. Also my intuition tells me it can potentially be very strong. I am quite hyped up about it right now :)

Btw, can current state of the art bridge programs solve all bm2000 problems assuming we tell them what hands are possible from the bidding ?

My understanding is that bidedit.exe needed to open those is automatically installed with old BBO client which you may find here: http://www.bridgebase.com/intro/v2problems.php You should be able to just double click .bss files to open once you install it.

My understanding from barmar's posts is that GIB doesn't even try this one ply search (due to computational power limitations of decade old computers). It also doesn't understanding basic signalling. My intuition is that it could be fixed in simulation based framework. Solving it is long way away and maybe not possible ever. Playing better than top humans is much easier though. Humans suck. Even at top level they make numerous blunders in card play which would never be made by fresh/rested/not pressured advanced player. Computer player doesn't need to be even close to solving anything to beat that imo. No idea about Jack but if Ginsberg stopped GIB development 10 years ago the chances are he missed many solutions/improvement due to his intuition of what is computationally feasible being developed at the time where hardware was 1000x slower than today. I appreciate all the knowledgeable people posting in this thread. I accept that my intuition on this one might be completely off so I will shut up and maybe code something.

Wow my intuition on this one is completely different. I am shocked nobody wrote a card playing program better than top humans by now and I think it's only for the lack of incentives (not that current programmers are bad, just that lack of incentives to write good bridge playing programs makes the field stagnant compared to say chess programming). Computers are fast these days they can simulate a lot. I think 1 or 2 levels of "assuming he plays double dummy and I do that the result is" are feasible and that could potentially be very strong. Anyway, I might be completely off on this one. Can't wait to finally have time to try some ideas and see how they fare.

I will check the book out :) As my contribution: all constructive FN biddings from vugraph in tree format (.bss) with alerts included which you can browse using BBO .bss viewer: https://dl.dropbox.com/u/86311885/fantunes.bss :)

I personally find rule based approach not too interesting. For me the point of writing bridge playing (or analyzing) program is to discover something new or something human players get wrong. If it's based on "expert system" and "huge database of knowledge" then it's limited to what opinions humans already hold about the game. So, for example, if there is a rule in constructive bidding: "with 12-14hcp if it goes 1D - 1S we raise to 2S with 4 spades" then it's instantly not interesting because there is nothing new we can learn from that. On the other hand if simulations or similar process shows the bid being effective then that's something - even if only confirmation in this somewhat obvious case. I want my program to answer questions like : "is it better to bash 3NT here or to go via stayman and try to discover 4-4 major fit" or: "if I bid 3NT here, how often I will make in real play", or "what is the difference in EV between preempting to 3S and 4S here". Once it's rule based it will just tell me what common wisdom here is which - especially that you are proposing that many people contribute to the knowledge base - about useless in my mind.

I am still to prove it to myself :) One thing which is bothering me when it comes to dmpro (don't know about other software) is that it doesn't run in parallel on multiple cores. As the problem of calculating double dummy results for many hands should scale linearly that would be huge boost on modern processors (like 8x faster on i7). It would require just few line of code more (assuming the software is written in C or C++) but without the source I can't do anything about it. Same goes for dealing software, although I admit I have no idea if for example deal runs on multiple cores. I am involved in writing simulation software for other card game now but once I am done with it I think I am going to spend some time on my own simulator just to have it run on multi cores (and deal cards faster I think dealing programs are really bad when it comes to rare hands). Yeah, bridge playing program would be cool. It's high time to knock GIB out of BBO :) What I would be very interested in for starters is a simulator which can solve defensive problems after several tricks are already played. It would require a lot of human input but could give valuable output about what play is the best at given stage. As far as I am aware there is no such software available but the concept is not that complicated.

Here I thought we should invite but I think I am starting to see the light. Full hand: http://www.kongres-slawa.pl/2012/k2040.html

We were discussing it with my friend and he thought passing is insane, I thought it's rather obvious. I guess he was temporarily unable to thing rationality (they ended up in 3N on this hand). Full hand: http://www.kongres-slawa.pl/2012/k2041.html

It seems that the last one was too easy, so let's try next one: matchpoints: [hv=pc=n&s=sqj862hak6dj98cak&d=s&v=0&b=11&a=1sp1np2cp2sp]133|200[/hv] 2♣ - gazilli (natural or any 16+) 2♠ - weak, up to 7hcp 2N/pass ?

[hv=pc=n&w=sjhj2daqjt76ck832&d=n&v=0&b=1&a=1d1sp1np2sp]133|200[/hv] What are our options ? What do we do ?

Imo double is t/o and I think it's way better to play it as t/o. Partner couldn't bid anything below GF strength but odds of him having a 5-6carder somewhere opposite our say: xx AKxx AQx AKxx are great. I mean, they have heard about 20-22 balanced part and still choose to overcall, it's not like we are collecting 500+ more often than once a century. In Vanilla 2/1 they are, in some decent system they aren't. Just open 1C and bid some multimeaning 2D reverse next round, wtp ? Even playing vanilla you could open 1D and reverse in hearts. It sucks a bit to have 4-4 instead of 5-4 but vanilla standard always suck with strong hand anyway, this lie is not even that big. If they bid spades you will have easy double all the way describing your hand perfectly.

I agree, can you think of better one ?