tnevolin

Hi, kwiktrix. Sorry for late reply. I kind of abandoned this thread for a while. 1) Combined, of course! It is a partnership who takes trick in bridge, not the individual player. 2) Yes, they are. 3) Absolutely. That is the core of my analysis and calculation. In fact, it is not even a comparison. I directly adjust coefficients to match the reality better. I don't use DD or SD emulation, though. The only thing I do is comparing my results to actual tricks taken in the game (average across the table, that's it). I am not sure how to describe degree of fit because the approach is complex by task nature and my program is not 100% academic. I do some tests those seem to be enough to convince me in good match but nothing fancy like Xi-square, etc. Correlation does not go through (0,0). 4) Sure. See below. Here is the folder with all related documentation https://drive.google.com/drive/u/0/folders/0BxM2JfK2YtucYmhjdThEOHhXTU0 Specifically, Calculation rules https://drive.google.com/open?id=0BxM2JfK2YtucRzVOTjZJcXlfa0k Method history and description (plus some relevant information) https://drive.google.com/open?id=0BxM2JfK2YtucWW10bndERmd2NzQ

Hi Everybody. Sorry that I disappeared for a while. I was regrouping my calculation to optimize score improvement as well as trick prediction. Essensially I am trying to maximize the potential score (MPP and IMP) that my evaluation system would give to players. It is not finalized yet and I stumbled accross one formalization problem and I hope you can help me. I am formalizing hand features for NT contract. Most of them are simple like high cards, long suit, high cards in long suit (the source of extra tricks). However, there is another important features affecting NT games outcome: stoppers. Usually it is the weakest and shortest suit that is the most dangerous for NT contract. Unfortunately, there are many factors and it is not easy to mathematically determine which suit is most beneficial for defenders. For example, you have short suit with very strong cards (AK) or long suit with very weak cards (T9876) - which one is to select as worst stopper? Another problem that stopper is a combined partnership effort. Partners surely can exchange each one best stoppers but it still unclear how to predict which suit will be defenders attack target because mere flag "stopper or no stopper" doesn't disclose information about partner's suit strength. What if partner showed stopper and you have another 3 low cards in it - does it make it significatly stronger? With partner's Kxx it doesn't, but with partner's Jxxx it gives only 6 cards to defenders lowering the risk of attack. And many more. I would appreciate any suggestion on the topic. Try to think in formal terms (card face values, suit length, plus vague information about partner's declared stoppers) so that others can apply your rule without ambiguity. Thank you.

Good suggestions. I accualy tried something similar just more generic. Very roughly estimated number of tricks: for NT = HCP/3 + 1, for suit = HCP/3 + combined trumps/2 + 1. Then if result differs from estimated by 6 tricks I throw whole section away assuming that the same person entering the data could make other mistakes. You can look at examples here: http://www.bridgebase.com/vugraph_archives/vugraph_archives.php. Most common errors are typos, missing results, missing hands, impossible hands (duplicate cards, missing cards). The only one non typo error I found is swapped N-S, E-W hands. So the contract says something like 4HE but it was S who played it. Grrrr. I don't remember the exact vugraph but it is one on their page. You can even see the movie for it with this error right in plain sight.

One good news and one bad news. I found a large game archive on BBO vugraph page. It is quite extensive: 400+k boards. Unfortunately, they all are manually entered and there are a lot of typos and mistakes. Some of them are easy detectable like impossible deal. Some of the are very hard to detect like one vugraph I found had N-S and E-W hands swapped. A very nasty error. When such error enter into analysis they pollute results. I am trying to devise empiric rules to filter such mistakes out. Anybody can propose such rule? I would prefer to filter out extra to make sure that rest of the observations are clean rather than leave incorrect entries. Thank you.

Parsed both pages for 2k records. Not too much. Some text on the second page says there are a lot of free PBN data in Gib or some other sites but I never found them in open access. Anybody uses Gib? Is there a free PBN database there?

I scanned all the events, rounds, tables and Boards there are. Apparently they just started recording boards since 11. That's all they have.

I finally downloaded them! Thanks for interesting source. The sad thing is that there is only 10k boards there. A tiny amount. :( Anybody knows some similar sources?

I like the link and the study. I even send the man a message but he didn't reply yet. Do you know him personally?

I know this research. This is specially true for NT contracts. It could be explained by lack of entries to weak hand and, therefore, lack of maneuver. I tried to factor it into the calculation. Didn't work well for two reasons. First, the effect is very subtle and there are not much hands with large skew like 24-0. Probably need larger dataset to catch it. Second, and more important one, is that this is second iteration factor. I.e. you need to evaluate your hands first and then do a second iteration for this adjustment. I decided not to include them for now as it complicates evaluation rules tremendously.

Yep, the idea exactly. Help people to make most difficult and important decision there in the game.

You results and conclusion correspond to my findings exactly. My 3NT contract requirements are 25 for IMP and 26 for MP. I just often replace 25/26 with 25 for the sake or simplicity. Then if we shift it 2 points down, as I explained in my initial post, we would get the results for Evolin points without constant which almost exactly corresponds to HCP. And here you get your 23(IMP)/24(MP) borderlines! If you use pure score and calculate what probability you need to win 3NT to still profit on average, you would get 40% (nv), 33% (v). Jamming v and nv together it'll be somewhere 37%. Of course, game points is not equivalent to IMP but they are good approximation.

You lost me. Isn't it what we started with?

I know some. The problem here is that there is a correlation between number of tricks and contract level. So if someone bid 3NT that means they have around 9 tricks. Maybe some more or less but not the whole lot. So if I restrict whole set to only those bidding 3NT I will get a conditional result that in simple human words would read: "If we bid 3NT and have so much points - how many tricks we have?". When you rephase it like this it is obvious that you don't even need statistical analysis to know that the answer will be somewhere in 8.5 - 9.5. I actually did what you proposed. Tried to narrow down to the interval of interest. Like if I would want to predict 3NT games better and don't care about 1NT, I would just cut out interval where people bid (or get) 8-10 tricks. Sounds promising but it didn't work. The coefficient values becomes ridiculous. Like I would get one huge constant of 9 tricks plus some very small coefficients for other features to account for slight variation around 9 tricks. And the overall prediction around 3NT got worse comparing with the case where I considered whole range of games.

You are right, jogs. This is a inherent skew that all analysts are pointing out. Number of tricks depends on contract because both declarer and defenders tacktics is driven by it. Unfotunately, I don't know a cure. I also didn't see anybody posting a good idea how to filter this skeweness out in analysis.

Nice link. Let me study it. Maybe I indeed have some statistical error inside.

OK. Let me try to find another game source and see if it holds. From the other side, 2 points is a huge difference that would be very difficult to explain by certain constant dumbness of players in thousands of games. By the way, can you point me to where it was "most thoroughly examined in past hand-eval studies..."? I never found one after extensive search for few years. That might resolve many of questions right away.

To all forum buddies. I need your expert opinion. m1cha proposed to shift scale for trump model down so critical contract requirements will match popular values. Upon some thinking I agreed with him that even though such shift distorts the exact trick estimate bidders are much more often interested in critical contract condition rather than generic trick estimate. So the convenience is huge enough to outweigh the incorrectness. So I did and updated evaluation document accordingly. Again thanks to m1cha for pointing this out. With this is done I continued thinking about further improvement in this direction to make point to contract requirement transition is even more convenient. Here is the background info for you. I have analyzed standard HCP evaluation model and found out that 3NT contract requirement is actually 23 points, not 25. That number marks the breaking point where declaring the contract become profitable on average, not that contract will be made 100%. This is an interesting finding and I believe it's correct because my recent NT model calculations suggest to add 2 points as a constant value to combined strength for better trick count estimate in NT contracts. My critical contract strength table lists 25 points for 3NT (IMP). With 2 points constant value in mind this is equivalent to 23 points 3NT requirement in HCP model. So far the math matches. Now let's go back to the convenience. This 2 points constant value is practically the only difference between Evolin NT model and HCP. The long and strong suit rule happens quite rare. So if we would remove the constant value these two models would match exactly 99% of the time and in the rest of the case they would differ by 1 point max. Which would be a huge in game convenience. The only consequence of this change would be shifting NT critical contract requirements down 2 points. I understand that this would drift away from popular values. However, let me reiterate again that HCP 3NT 25/26 points requirement is incorrect one. The correct one is 23/24 points anyway. Please let me know which approach would be more convenient in your opinion. If we decide to shift NT model scale the same should be done with Trump model to keep critical contract requirements in sync. Trump model shift doesn't present a challenge, though. I would just add/subtract the value for number of trumps and this would do it. Trump model is already quite complicated so such shift doesn't make it more or less complicated anyway.

There are two graphs on pages 9 and 10 depicting experimental average depending on theoretical prediction for SAYC and Evolin points, correspondingly. In simpler words, I estimated number of tricks for each hand in observation - that would be the "estimated tricks" (horizontal) scale value. Then for each observation I got real tricks. That is the "actual tricks" (vertical) scale value. Then I averaged results by whole tricks. So the graph point corresponding to 10 estimated tricks represents all hands that were estimated in range 9.5 - 10.5 tricks. And vertical value of this point is the real trick count average for all of these hands. Now back to residual quadratic mean. Residual is the difference between experimental and theoretical values. The standard measure of match quality is the sum of residual squares. The lesser the value the better the match. This number is good when you run the optimization but is it not good for understanding how much experimental values deviate from theoretical ones on average. For that the <residual quadratic mean> = SQRT(<sum of residual squares> / N) is used. What I meant in previous post is that I calculated residual quadratic mean across all observations as a single number instead of splitting it to buckets as I did on prediction accuracy graphs. Graphs are for visual perception only. It is nice to see how two lines goes close to each other. Whereas for residual quadratic mean you need just one number which show you an average error. So taking that residual quadratic mean is 0.8, one can draw two supporting lines on the prediction quality graph. One is 0.8 tricks above and one is 0.8 tricks below. Then you can say that 70% of experimental dots would fall into this band. I just didn't want to actually draw them on the graph to not overload it with heavy math.

Oh, I like disagreements and opinions. That's why I opened this discussion!!! What I meant is quite often this discussion steers toward "this new system is no better than existing ones". I didn't plan it to be better. I planned it be special. It is based on principles I never saw in others. This is sort of experiment to see if these principles are viable or just bullshit. The most important part of it is a Guidelines with principles description not the bidding chart.

That's correct. I'm not sure I am in line with the mainstream bidding theory here. Most of them try to be very exact on distribution description simplifying opening rules. Unfortunately, this never works well. There are always "awkward" distributions those do not fit in a simplifying rules but still need to be shovel somewhere. That produces monsters like 1 minor openings in SAYC or 1D in precision or 2D (multi) in Polish system, etc. Those openings are pretty hard to follow, they require a specialized bidding sequence to eventually uncover their complex meaning. In my system I favor integral hand evaluation to specific distribution description. Therefore, it is allowed to make your own judgement on suit bid selection. Your goal is to find better fit as soon as possible rather than pass an exact hand shape to partner.

Gentlemen, I am sad that you keep beating me on my head for the "attempt to create another system". :( Let me reiterate that this is merely and attempt to illustrate how bidding system can use the full strength of said valuation method. This being said, besides of just wrapping valuation method in the bidding system, I tried to make such system as workable and elegant as possible to make it potentially usable too. However, the small details in the bidding system are irrelevant. They could be changed any time.

Thanks for posting these important questions here. Not in attempt to discard your points but to bring our discussion up to speed, I'd recommend you read my article and corresponding valuation model topic on this forum. Also, this is a perfect question for valuation model topic as well. In short, these are EXACTLY questions I had in mind while I was working on my method. I did not play to solve them all in entirety. However, I can proudly say that I solved them much better than other contemporary methods do. If you look at hand evaluation document, you'll see that many factors you mentioned above are accounted there already. Like different price for high cards in trump and side suits, different price for high cards for NT vs. trump contract, even such factors as interference between yours and your partner hands are there! :) In addition, I analyzed the famous "Which parameters have the greatest effects on tricks?" question during few years trying tons of various factor combinations. Some of them are much more complex than those you touched in your post. I can surely say that the resulting valuation method I present is the answer to your questions.

There are few concepts described in "Guidelines". This is essentially part of system description. I recommend to start reading it first. Unfortunately, there are quite a few concepts there. I couldn't possibly include them all in written form into the bidding chart. Sorry about that. In web publication both these part would probably go together: guidelines first, then chart. I've separated them in word documents for ease of navigating and updating. In short, one of the major concepts is that the suit length is not a strict requirement. It is specified for matching bid selection guidance only. Bidder is allowed to loosen this restriction by 1 card in exchange for other compensating features (distribution, suit strength). Yes, your partner will be uncertain about exact number of cards in your trump suit and you may end up with 7 card fit only, but the point is - it is irrelevant as long as they know your hand strength which incorporates your trump length already. With slight variation in suit length it is up to bidders to prefer one trump suit over another. Your example can be bid in two ways. 1345 - Generally open clubs even if you have only five. The singleton makes your hand sharper and, therefore, more desirable to play clubs versus NT. 4315 - Either clubs or spades depending which one is stronger. If you have pretty strong hand and expect to bid twice, the clubs - spades sequence is better because your second bid of spades shows exact length.

Good point, Stefan. Indeed it would. And this is exactly what I pointed out in my article and in words anywhere else this question pops up. See comments regarding bidding system at the end of the article. Here are methods of applying evaluation model to bidding system in order of ease: Use them in existing system without modifying system at all. Use them in existing system slightly modifying it to avoid very distributional descriptive bids. Create new system that may resemble existing one but redone completely to utilize full benefits of evaluation model. I don't need a topic for #1 and #2, obviously. #3 deserves a topic to gather people opinion. That doesn't mean it is the best way to go. :) (edited) By the way, I would greatly appreciate anyone's help in trying #1 and/or #2 out! I believe it would bring more practical results than just this theoretical discussion. Anyone is up to it?

That's correct. I used these terms since all bridge dictionaries explain them the same. So I assumed they are standard. Regarding whether to separate defensive bidding from non-defensive one. This is a huge topic by itself. In short, I tried to standardize non-competitive and competitive sequences as much as possible. IMHO that is a big oversight all bidding system designer allow. Obviously, designing competitive part of the system is much harder as it allows for more variations and it has to account to opponents' holding as well. Naturally, this part requires more description and more analysis and more guidance from system designers. Yet they spend 99% of efforts to perfect the easiest part of it leaving the rest undeveloped. I believe this is irresponsible and selfish attitude. Competitive bidding is much harder to navigate and it occurs in 2/3 of the auctions. This is where players need most help from system designer but they are abandoned and on their own. :( I think the right approach to bidding system design is to establish a general principles those would guide player through auction regardless of whether opponents dare to say anything. Only after that is settled, designer can spend more time on perfecting specific auction branches like one side bidding or slam bidding, etc.