Talk:Hand evaluation

From Wikipedia, the free encyclopedia

 This article is part of WikiProject Contract bridge, an attempt to build a comprehensive and detailed guide to bridge on Wikipedia. If you would like to participate, you can edit the article attached to this page, or visit the project page, where you can join the project and see a list of open tasks.

I am aware that there are some articles covering parts (Point count and High-card point) of this article but I believe there is a need for a single article covering this subject in total. I hope others watching bridge articles will agree but if you don't please let me know ... I am new at this but very keen - be gentle!Abtract 22:05, 8 May 2006 (UTC)

I'm grinnin' its a pretty good presentation. Terryeo 08:01, 20 May 2006 (UTC)

Contents

[edit] HCP - one fifths

This is growing into a nice overview article. What do you folks think: should we add hand evaluation rules (alternative to HCP) derived from double dummy analysis on large sets of hands? Or is this too theoretical? (I'm thinking about Thomas Andrews' work.) JocK 12 June 2006.

To be honest, I don't know anything about it, so I'd be interested. However, if the section becom too long, you could just give an overview here and link to a new article. Duja 06:51, 13 June 2006 (UTC)
I have never heard of it so it may well be a useful addition (for me anyway!) but if it is as theoretical as you suggest, then I agree that a separate article linked here might be best - like Zar Points.Abtract 08:37, 13 June 2006 (UTC). I just found this http://bridge.thomasoandrews.com/ which may be useful - no time to read it yet.Abtract. I see this is a private site so I think we ought to know JocK if you are Thomas Andrews or closely associated with him? Abtract 09:04, 13 June 2006 (UTC).
Me, Thomas Andrew? That would be sneaky! :) No, I don't know the guy. From a theoretical point of view on hand-evaluation, his research seems very useful though. For instance, it leads to hard data on what is the best point evaluator for no-trump contracts (http://bridge.thomasoandrews.com/valuations/cardvaluesfor3nt.html). The optimal high-card point evaluation rule appears to be: A = 4.0, K = 2.8, Q = 1.8, J = 1.0, T = 0.4. (Quite close to standard HCP!) Obviously, the result has limited practical significance as with 40 points per board you can not bid your hand accurately within 0.2 point. JocK 13 June 2006.
OK as you will see I have added his site in as further reading. If you want to do more go ahead but bear in mind it is hardly mainstream. Abtract 18:08, 13 June 2006 (UTC).
thomaso I am Thomas Andrews, and a quick note. The 4.0/2.8/1.8/1.0/0.4 evaluator is optimized for 3NT contracts, not notrump contracts in general.
The problems with fractional valuations is that, while they can be 'optimized' we still have to base bidding on a 'yes/no' response. For example, after a 1NT opening, when do you leap to game? There's no realistic way to always bid game if you have 24.2 or more points, combined, and not bid game with fewer combined, because you really have three paths: Pass, Invite, Jump. If the 1NT opening is 14.6-17.4, then you obviously want to jump to game with 9.6 points. You want to pass with 6.8 points. But do you really want to invite with 7 points? With 9.4 points? No. So you are going to miss some hands where you and partner have at least 24.2 points, and bid some games where you and partner have less than 24.2.
So, while this evaluator is interesting, I'd hesitate to ascribe any practical value to it.
Well you heard it here on this page first - maybe I was a little quick in adding it to further reading, but honesty surely deserves its own reward?Abtract 20:12, 13 June 2006 (UTC).
Hi Thomas, nice to meet you. You are too modest: if your 'fifths evaluator' is optimised for the decision whether or not to bid 3NT, it *is* optimised for notrump contracts in general (i.e. for deciding whether to stay in partscore or head for game). If you mean to say that an avaluator optimised for the decision 'game-or-not' is not necesarily suited for the decision 'slam-or-not', you are obviously right. But then: which evaluator is?
Clearly, no-one is proposing to use such non-integer evaluators in practice. The practical value of having obtained such an optimal 'vector-evaluator' lays in the fact that it shows the Work-count to be close to the optimal evaluator for deciding whether to or not to stay in part-score on notrump hands.
As an aside: ever contemplated evaluating the newer methods described here? (Zar, NLTC, ...) JocK 13 June 2006.

[edit] LTC - one thirds

Here's a simple modification that should make the LTC somewhat more accurate.

After the initial calculation, subtract one "tick" for each A, add one "tick" for each Q and add 2 ticks for each J. Three ticks = one LTC. Then AKx, AKx, AKx, AKxx = 4 LTC- 4 ticks = 3 LTC - 1 tick But KQx, KQx, KQx, KQxx = 4 LTC + 4 ticks = 5 LTC + 1 tick And KQx, KQx, KQx, KJxx = 4 LTC + 5 ticks = 5 LTC + 2 ticks

Another suggestion is to round up with 2 surplus ticks and round down with one.

NOTE: In the revised version, Jacks are counted in the initial calculation.

Jillbones 02:17, 16 November 2006 (UTC)

I expect you looked at 5.3 and 5.4 in the article where you will have seen very similar improvements to the LTC method and citations as to the source. If you know publications that show this 'one third point' improvement, then you should include it in the article at the appropriate place. If, on the other hand, this improvement is your own idea or that of a friend then this would count as original research and be against policy so shouldn't be included.Abtract 10:15, 16 November 2006 (UTC)

[edit] Law of Total Tricks

I'd like to direct your attention to the article on DONT, which references the Law of Total Tricks. Clicking on that link brings the page here, which is both unsatisfying and misleading. First of all, the Law of Total Tricks is not a hand evaluation method - simply counting the combined trumps cannot tell you if you are likely to make 3 tricks or 13. Secondly, at the very least, clicking on the link should put you at the Law of Total Tricks section of the article. If anyone could help resolve this issue, I should greatly appreciate it.Eljamin 18:05, 20 November 2006 (UTC)

Excellent idea I have improved the link as you suggested.Abtract 23:13, 20 November 2006 (UTC)
I've split the article from this one and I'll fix the link in DONT; with due respect, I think that Abtract was a bit overzealous in merging the topics into this article; by itself, the LoTT is a tactics rather than a method of hand evaluation: it doesn't say how worth is your hand but which tactics to employ in competitive auction. Plus, it's now long enough that it definitely deserves a separate page. Duja► 12:49, 21 November 2006 (UTC)
Perhaps you might like to discuss the merits of splitting it off before you do so and refain from discussing how over zealous I may or may not have been. Abtract 13:03, 21 November 2006 (UTC)
OK, here I am. Actually, I only recently noticed myself (before Eljamin's suggestion) that LoTT was merged here (and I don't recall you "discussed the merits of merging before doing so", as the LoTT article had a pre-history before it ended up here, but let's leave that behind us). But I can't find a good reason why it should remain merged, as LoTT has well-defined scope outside of mere hand evaluation, as I said above. Actually, the contexts of wikilinking ("mine" was from Balancing (bridge), Eljamin's from DONT) had nothing to do with hand evaluation. Yet the reader ends up here. Besides, I expanded the LoTT article (and it has potential for further expansion), so that its backmerge here would really result in a huge article. In addition, LoTT section was in the middle of this article, right between LTC and playing tricks, which are, I think you agree, on a fairly similar line of reasoning while LoTT appear to came straight from Mars in the context. I don't mind a brief coverage about LoTT here, but not more than a sentence or two. We also have Zar Points in a separate article. (which, I still maintain, are more related with the subject of Hand evalauation than LoTT) Duja► 11:48, 22 November 2006 (UTC)
As I said I am not against a separate article ... leave the old stuff here and I will cut it down and refer to the main article.Abtract 13:17, 22 November 2006 (UTC)
OK I have made it just a brief summary and linked to the main article.Abtract 13:36, 22 November 2006 (UTC)

[edit] LTC baseline

This last edit http://en.wikipedia.org/w/index.php?title=Hand_evaluation&diff=174582171&oldid=171495298 suggests that there is a clear rationale between the basenumber 24. That is a misconception. True, a 4-3-3-3 Yarborough has 12 losers. And such a Yarborough opposite another 4-3-3-3 Yarborough leads to a total losers count of 24. So far so good. But, for LTC to be applicable these two 4-3-3-3 hand need to have their 4-cards in the same suit. And when these hands get played in the 4-4 trump fit, one may expect to take one trick. so, if there would be a systematics to it, the LTC baseline number should be 25 (like in NLTC). The baseline, however, is selected as 24 simply because that fits best with emperical data. I suggest we revert this last change. JocK (talk) 16:56, 29 November 2007 (UTC)

I don't think that is correct. Since there are by definition never more than 3 'losers' in any one suit, there can only be 12 'losers' in each hand ie 24 in total - this is where the number comes from. Of course you are right that if there are matching 4-card suits a trick or two may well be made but this is not how the method works - it is entirely dependent on a semi-artificial count of missing aces kings and queens. I like the edit and think it should stay. Abtract (talk) 17:03, 29 November 2007 (UTC)
Ok, you don't understand. No problem, I'll try it in a different way: using the baseline number of 24 LTC predicts that ♠xxxx ♥xxx ♦ xxx ♣xxx opposite ♠xxxx ♥xxx ♦ xxx ♣xxx when played in spades takes zero tricks (24 - 12 losers - 12 losers = 0 tricks). Do you believe this is a realistic estimate?
In other words: the example does not prove that the baseline is 24, if it proves anything, it would be a baseline of 25 (25 - 12 losers - 12 losers = 1 trick). Whether you like it or not, the baseline number of 24 in LTC is just a fitting parameter. JocK (talk) 23:01, 29 November 2007 (UTC)
I'm sorry it's you that doesn't understand ... as is clearly stated in all the literature no suit can have more than 3 losers, it is the absence of A K and Q that constitute a 'loser' (not to be mistaken for a real loser). Thus Axx, Kxx and Qxx each have 2 'losers' even though their trick taking potential is clearly quite different. Suits with more than 3 cards can only have a maximum of 3 'losers' since they can only be missing the A, K and Q ... so Qxxx has only 2 'losers' and so on (this is clear in the article). In the example you quote each hand does indeed have 12 losers, making 24 in total although naturally I agree with you that in combination and chosing spades as trumps there is a pretty good chance of making one or two tricks (as I said originally). The method recognises this total of 24 and uses it in the subsequent calculations as a tool only there is no suggestion that this is in anyway "real". Its unimportant that you think the baseline number should be 25 because that isn't what the method as publish actually uses ... it uses 24 because no suit can have more than 3 'losers'. I hope this helps. Abtract (talk) 23:46, 29 November 2007 (UTC)
the first time I encountered it I heard it explained as Abtract does - the maximum number of losers a hand can have is 12 as the 4th card is a 'winner' as a length trick. Therefore the maximum number of losers is 24, so that is what is is subtracted from. However, just because this is used as a justification does not make it the true reason... Anyway, losing trick count is rubbish. I prefer winning to losing... :-) Cambion 17:18, 30 November 2007 (UTC)
Indeed, fully agree, Cambion. Abtract, please read the recent literature. Ensure you get a copy of TBW. Ah, well why do I bother? Cheers, JocK 20:16, 30 November 2007 (UTC)
So we have agreement as Cambion says "the maximum number of losers a hand can have is 12 as the 4th card is a 'winner' as a length trick. Therefore the maximum number of losers is 24, so that is what is is subtracted from" and JocK says "Indeed, fully agree" - excellent. However you both seem to be implying that despite this being the standard explanation of the 24, we should reflect some other view because you both think "Anyway, losing trick count is rubbish. I prefer winning to losing... ". Give me a clue what TBW means and I will read it or better still tell me what it says, if relevant - always eager to learn. Abtract 21:45, 30 November 2007 (UTC)

"We have an agreement that the maximum number of losers in a hand is 12"? C'mon! No-one ever disputed that. This: "Therefore the maximum number of losers is 24, so that is what is is subtracted from", however, is complete nonsense. And that is my point. If you still don't understand, then let's stop the discussion. JocK 23:49, 30 November 2007 (UTC) (you haven't tried a wikisearch on 'TBW'..?)

The recent edits by Abtract made things worse. Have removed all the recent edits and changed into one simple (and above all less-suggestive) statement. JocK 00:13, 1 December 2007 (UTC)

since it is a subscription magazine that i dont subscribe to i cannot. Why don't you explain what your point is clearly on here ... you agree that 12 is the maximum number of losers in one hand but you disagree that the number 24 is simply the sum of two hands at 24 each - is that what you are saying? Abtract 00:49, 1 December 2007 (UTC)
Haven't you read my remarks al all? I started of by stating "a 4-3-3-3 Yarborough has 12 losers". You then started arguing repeatedly that I was wrong because a single suit can only have three losers so that a hand could have no more than 12 losers. This is embarrasing.
I also cleary stated: "a [4-3-3-3] Yarborough opposite another 4-3-3-3 Yarborough leads to a total losers count of 24". And now you are asking me whether I agree that a hand of 12 losers opposite a hand of losers adds up to a combined number of losers equal to 24? This is unbelievable.
I am always willing to explain, but sometimes it's better to give up. Let me give it one last try, I then leave it to others to make attempts to demonstrate why removing the word coincidentally from the text is wrong. (In case you didn't notice: that is what it all boils down to.)
Suppose the article said:
The "losing tricks" in a hand are added to the losing tricks in partners hand and the resultant number is deducted from 24 (a number that equals the maximum number of hours a day one can play bridge); the net figure is the number of tricks a partnership may expect to take when playing in the established fit.
Wouldn't you agree that one needs to remove the text between brackets or at least put in the word coincidentally:
The "losing tricks" in a hand are added to the losing tricks in partners hand and the resultant number is deducted from 24 (a number that coincidentally equals the maximum number of hours a day one can play bridge); the net figure is the number of tricks a partnership may expect to take when playing in the established fit.
Note: in this analogy no-one is arguing against the number of hours in a day being 24. Neither is anyone arguing against the fact that according to LTC one needs to subtract the combined number of losers from the baseline number 24 so as to arrive at the tricks one may expect to take. What I am arguing against is that the number of hours in a day is in any way related to the baseline number. JocK 07:25, 1 December 2007 (UTC)
The word "coincidentally" is a personal opinion (yours) and should not be included. This method is all about losing tricks not hours in the day so while your analogy was vaguely amusing it was hardly relevant. However in the spirit of cooperation I will agree to your other suggestion that we remove the phrase altogether since it adds very little if anything to the article. Abtract 10:29, 1 December 2007 (UTC)
Pfffff... finally! Your remark suggests you still don't get it, but doesn't matter: the erroneous suggestion is now removed from the article. JocK 11:09, 1 December 2007 (UTC)
Is that what passes for a gracious "thank you" in your part of the world? Abtract 11:53, 1 December 2007 (UTC)
In your part of the world one expects a "thank you" without a favour being done? Curious. BTW: No thanks, it was realy fun to go through the hassle of explaining the obvious (and you kept me off the street as well- ahhh that must have been the "favour" - thankzzz!!) JocK 15:14, 1 December 2007 (UTC)

I must say I never expected the particular addition to cause so much controversy. Realize that in Losing Trick Count, it doesn't say how many losers you have, only it defines a heuristic which calls certain cards "losers." I think it's important to realize that distinction. Therefore, by the definition of the LTC, there are by definition 12 losers per hand. By definition you have 24 losers in both hands. The number of tricks you can take is by definition 24 minus the number of losers (via LTC) you have. More often than not this has relevance in the real world in terms of LTC losers being the same as non-LTC losers. It doesn't always work. Buoren (talk) 17:48, 12 December 2007 (UTC)

Depends what you mean by "by definition". Whatever definitions/rationale you write down for LTC, these should mutatis mutandis also hold for NLTC. So:
In LTC, there are by definition no more than 12 losers per hand. So you can not have more than 24 losers in both hands. The number of tricks you can take is set to 24 minus the number of losers
In NLTC, there are by definition no more than 12 losers per hand. So you can not have more than 24 losers in both hands. The number of tricks you can take is set to 24+1 minus the number of losers
JocK (talk) 18:22, 12 December 2007 (UTC)

[edit] Math

The calculation of 68% uses the table on page 287 of the Official Enc of Bridge, as follows: for the trump suit declarer and dummy each hold 4 cards, leaving 5 with defenders. These will split 3.2 67.83%, 4.1 28.26% and 5.0 3.91. The contract will only succede with a 3.2 split because by definition x cards will never win a trick if there is an opposing card out. Therefore the odds of Axxx and KQxx making 4 tricks is only 67.83% ... no ruffing is possible because of the 4333 ditribution. Abtract 13:50, 1 December 2007 (UTC)

"no ruffing is possible because of the 4333 ditribution"..??? Áre the opponents not allowed to ruff because you have a 4-3-3-3 distribution? JocK 15:01, 1 December 2007 (UTC)
You have a "got out of bed the wrong side" attitude today jock. I meant that declarer and dummy cannot ruff because they have a matching 4333 shape - of course defenders could ruff on first round with 7.0 Abtract 16:27, 1 December 2007 (UTC)
And I meant that your result "the odds of Axxx and KQxx making 4 tricks is only 67.83%" is wrong. It ignores the fact that under certain circumstances the opps can make an addition trick by ruffing. Go back to bed, you need some sleep... JocK 16:59, 1 December 2007 (UTC)
Quite right I forgot that ... so point it out, but why the attitude? Abtract 17:20, 1 December 2007 (UTC)

[edit] main article status

The LTC section has reached sufficient size to merit its own article so I have (re)created it, and will greatly shorten the entry here. Abtract 18:46, 1 December 2007 (UTC)

[edit] Control count

This sounds interesting but I have never heard of it as such ... and I cannot find it in the Off Encl of Bridge. Anyone know how it works in practice? and has anyone a citations for it? Abtract 23:01, 1 December 2007 (UTC)

Thanks next edit helps but it still doesn't explain how to discover the number of controls in a hand, how to actually use the theory. According to the OEoB, controls are assumed as direct result of HCP but presumably there is more to it than that? Abtract 23:46, 1 December 2007 (UTC)
To discover the number of controls in a hand is trivial: Controls = 2*NumberofAces + NumberofKings. According to Rosenkranz' table HCP and controls roughly relate as HCP = 2.5(Controls + 1). JocK 00:09, 2 December 2007 (UTC)
That's what I thought (it says it in the OEoB) but in that case how does the example in the article have any validity? Both west hand have the same number of HCP. Abtract 00:14, 2 December 2007 (UTC)
A few more edits but it still isn't clear how a player is to discover whether his partners hand is control rich or not ... other than using a form of blackwood for example. If controls are generally dependent on hcp, there must be a method of discovery hidden somewhere. Abtract 01:35, 2 December 2007 (UTC)
We could relate this to the Norman four notrump convention, but in most cases you don't need to know partner's precise control count. You and your partner both evaluate your own hand and convey a HCP count (where appropriate adjusted for excess or lack of controls) to your partner. What is perhaps not sufficiently clear from the current write-up is that the control count is no more than an additional metric used to adjust HCP count? JocK 02:04, 2 December 2007 (UTC)
Yes the section certainly seems to have grown out of proportion to its current day importance ... I wonder if it would be better in the Slam-seeking conventions article? Abtract 10:59, 2 December 2007 (UTC)
I agree, relative to the section on HCP itself, it has grown too big. However, I would propose that we beef up the section on HCP first (history, context, examples, etc.), and then decide what to do. By the way: utilising control count as a metric complementary to HCP is certainly mainstream amongst expert players (not only for slam hands, but also for game-invitation hands with a fit). JocK 11:18, 2 December 2007 (UTC)

[edit] Proposed new order

I am going to re-order the sections, hopefully other editors could give me say an hour to do this - then we can discuss it, change it or revert it if that's the thought. Abtract 16:25, 2 December 2007 (UTC)

OK thanks, I have finished for now. Feel free to amend or discuss, it probably isnt perfect but I think it helps read what is becoming quite a long article. Abtract 17:14, 2 December 2007 (UTC)
It's now follows a more logical structure; certainly an improvement. You dropped the introduction to the visualisation section. (Because of its style?) The section needs some intro though. Have added one - more encyclopedic (?) - sentence. JocK 18:21, 2 December 2007 (UTC)
Good. Yes it was 'style' but I am afraid your replacement is not any better. It is very much a pov - it may be correct but we cannot put our own opinions in, just well cited opinions of (in this instance) bridge experts. I will leave you to remove it or provide a citation. Abtract 18:40, 2 December 2007 (UTC)
OK, give me a bit of time to find a proper citation. Should not be too dificult. I don't think we can do without such an introductory sentence. (If you can render the sentence, non-POV, be my guest - but don't just take out.) thanks, JocK 19:33, 2 December 2007 (UTC)