Point counting method

[Zelandakh]

Any chance of running the same analysis but separating out a queen or lower with another honour versus without and also singleton versus doubleton versus 3 or more? I think you should also separate out a singleton ace or king versus 2 or more. Ideally we would analyse all honour combinations rather than individual honour values here I think. Indeed I thought this had been done by someone a while back but if so I cannot remember any significant results from it (other than the normal stuff anyway).

[gwnn]

http://www.bridgebase.com/forums/topic/32125-hand-evaluation-for-suit-contracts-investigated/ has some fun stuff. Oops, suit contracts.

[Zelandakh]

http://www.bridgebase.com/forums/topic/32125-hand-evaluation-for-suit-contracts-investigated/ has some fun stuff. Oops, suit contracts.

I have seen a lot more analysis on suit contracts, mostly because of the shake-up that Zar points caused. I have not seen so much detail devoted to NT contracts, probably because Milton is good enough here and there is less scope for improvement. Nonetheless, I think such an analysis would be interesting and could (theoretically at least) produce a set of "standard" upgrades and downgrades which would be great for both B/I players and perhaps also computers. Of course, on a personal level it would also be of great interest to see how my own list of amendments fares against reality.

[bluecalm]

For better analysis, we need a single dummy solver, but I'm not aware of one.

 

Double dummy solver with single dummy leads would be a big step in this direction. You would define what information the opening leader has from the bidding and amount of tricks would be determined after the program chooses 1st lead on every hand. Even if it requires simulating 100 additional hands just for lead decision it's entirely viable on modern hardware (it seems available simulators these days don't even do simple parallelism which would instantly speed them up by a factor of 4/6/8 and hands for the lead simulation could be dealt in such a way to make solving faster (spots closer together)). Besides that determining decent 1st lead could probably be implemented decently without any simulations.

 

I still hope someone writes one before I do :)

[cherdano]

If somebody suggested that 5-4-3-2-1 is a better scheme than 4-3-2-1-0 or 4-3-2-1-1/2, and didn't tell you that those numbers are based on an analysis that doesn't take into account whole hands, would you all still be so negative about it? For that matter, did the analysis on which Work is based (was there such an analysis?) take into account whole hands?

 

Imagine playing a game of money bridge with the following set up: In odd boards, I will be dealt a 4333 with 4 aces. In even boards, you will be dealt 4333 with QJ QJ QJ QJ. All other hands are random. Only NT contracts are allowed.

Since this is replacing an Ace by QJ in the same suit, rather than by Qxx Jxx, this is actually a very favorable setup. In addition, my aces aren't supporting any lower honors. So, if you think Banzai count is reasonable, QJxx QJx QJx QJx will win a lot of money against Axxx Axx Axx Axx. Fancy a game?

 

I think what you are missing that the judgment of good bridge players is much much finer than just using point count. They don't just use point count, they also have an intuition which hands are slightly undervalued or overvalued by point count. If Banzai points were right, than someone would have found out you can win by upgrading hands with many queens and jacks.

[aguahombre]

...

[blackshoe]

If Banzai points were right, than someone would have found out you can win by upgrading hands with many queens and jacks.

Hm. Okay, that seems to make sense.

 

Interesting. I took Statto's 'pure value' numbers and normalized them to 4 and 5 for the ace. Got:

 

A-4.000, K-2.817, Q-1.686, J-0.889, 10-0.361 = not much different from the pure numbers, unsurprisingly.

A-5.000, K-3.522, Q-2.107, J-1.111, 10-0.451 = this looks to me like a 5-4-2-1-1/2 count, and a 50 point deck. But I probably have no clue what I'm talking about. :blink:

[Cthulhu D]

That would make sense and supports the view that work tends to slightly overrate Qs +Js. To take the Axx Axx Axx Axxx equivalence above, now the equal hand is KQJT KQT QJT QJT and that's possibly equivlent.

 

3NT is cold opposite xxxx Axx Axx xxx or xxxx Jxx Kxx Kxx which is about the least partner could have to drive to game

[bluecalm]

To take the Axx Axx Axx Axxx equivalence above, now the equal hand is KQJT KQT QJT QJT and that's possibly equivlent.

 

Intuitively the 2nd one is much stronger. I mean, it would make 3nt opposite 8hcp balanced more often (and by a lot).

If someone feels I am incorrect here I can run a simul.

[Cthulhu D]

To take the Axx Axx Axx Axxx equivalence above, now the equal hand is KQJT KQT QJT QJT and that's possibly equivlent.

 

Intuitively the 2nd one is much stronger. I mean, it would make 3nt opposite 8hcp balanced more often (and by a lot).

If someone feels I am incorrect here I can run a simul.

 

Probably need to add in a bonus for honour sequences.

[0Filou]

I think that you will find the answers to your questions in the book : Optimal Hand Evaluation (recently published by Master Point Press and available for ordering through Baron Barclay or the Bridge World on-line bookstore). In that book, the author clearly identifies all the flaws of existing point count methods and demonstrates the vast superiority of the Optimal point count. It should become "The Standard" in Hand Evaluation point count.

Oh, by the way, it also shows that the Banzai point count has no statistical validity whatsoever and is actually quite absurd. Good reading !

[Zelandakh]

Hm. Okay, that seems to make sense.

 

Interesting. I took Statto's 'pure value' numbers and normalized them to 4 and 5 for the ace. Got:

 

A-4.000, K-2.817, Q-1.686, J-0.889, 10-0.361 = not much different from the pure numbers, unsurprisingly.

A-5.000, K-3.522, Q-2.107, J-1.111, 10-0.451 = this looks to me like a 5-4-2-1-1/2 count, and a 50 point deck. But I probably have no clue what I'm talking about. :blink:

I realise this was posted in 2012 but, as it was resurrected, here is the full set of normalised values for K=3, A=4 and A=5:-

 

Card Pure. All.. Pure K3..... All K3...... Pure A4..... All A4...... Pure A5..... All A5

A... 4.005 4.276 4.2591279688 4.6461427019 4.0000000000 4.0000000000 5.0000000000 5.0000000000

K... 2.821 2.761 3.0000000000 3.0000000000 2.8174781523 2.5827876520 3.5218476904 3.2284845650

Q... 1.688 1.574 1.7951081177 1.7102499095 1.6858926342 1.4724041160 2.1073657928 1.8405051450

J... 0.890 0.837 0.9464728820 0.9094530967 0.8888888889 0.7829747428 1.1111111111 0.9787184284

T... 0.361 0.365 0.3839064162 0.3965954364 0.3605493134 0.3414405987 0.4506866417 0.4268007484

9... 0.173 0.158 0.1839773130 0.1716769286 0.1727840200 0.1478016838 0.2159800250 0.1847521048

8... 0.061 0.030 0.0648706133 0.0325968852 0.0609238452 0.0280636109 0.0761548065 0.0350795136

 

Edit, and extended also for A=6 and K=4:-

 

Card Pure A6..... All A6...... Pure K4..... All K4......

A... 6.0000000000 6.0000000000 5.6788372917 6.1948569359

K... 4.2262172285 3.8741814780 4.0000000000 4.0000000000

Q... 2.5288389513 2.2086061740 2.3934774903 2.2803332126

J... 1.3333333333 1.1744621141 1.2619638426 1.2126041289

T... 0.5408239700 0.5121608980 0.5118752216 0.5287939152

9... 0.2591760300 0.2217025257 0.2453030840 0.2289025715

8... 0.0913857678 0.0420954163 0.0864941510 0.0434625136

[MaxHayden]

I realise this was posted in 2012 but, as it was resurrected, here is the full set of normalised values for K=3, A=4 and A=5:-

 

That was helpful. Do you have a stats package with a double-dummy solver set up that you could run a full test on the reliability of a few different count methods? The Darricades book recently referenced contains a ton of information and the ultimate list of adjustments is fairly complex. So I wonder how much of an improvement it provides over TSP or BUMRAP, and specifically, which elements provide the most bang-for-the-buck. I don't think people will be using it at the table any time soon, but it's good from a bidding system design standpoint to know what hands are or aren't statistically similar.

 

Everything passes the "gut-check", but I haven't had a chance to do something rigorous. My understanding is that the author does not have a formal statistical background and got his results by manually checking something like 8000 hands. So I'd be interested in seeing a computerized validation of his results. For better or for worse, the current situation has left me swamped with work. So I keep putting off testing it.

 

I discussed this back last year when I first got a copy of the books in question. The person going around bumping threads to mention it should have just posted there. They are giving a promising book a bad impression by doing a bunch of thread resurrections.