[dake50]

I'm trying to chart 531 count as is done for HCP for 8-20 points.
Need help running sims or calculations. eg. take 12-14 HCP -> 531.
The basics of 531.
**A=5**, except A-single is 4.
**K=3**, except K-single is 1.5.
**Q=1**, except Q-single,Qx is 0.
J=0
xxx,xxxx is -1.
Q=2 if with A or K 3+length.
J=1 if with Q or K or A 3+length.
I have confidence 531 is better (at least for 1st evaluation) but
I am trying to get a handle on the impact of this new evaluation.
Is it so small, that my preference is just that, or big enough to
publish/persuade experts?

[Free]

I don't want to give you the impression that this point count is awful, because it looks quite logical so it may actually be worth something. It upgrades various honor combinations while it downgrades some other holdings. However I don't think you'll be able to persuade many experts because:
- an expert takes the suggested 'corrections' into consideration without really paying attention to it (for example they all know that Jxx is worth less than 1HCP while QJx is probably worth a trick).
- it's much more complicated than HCP count, which would result in more effort to evaluate a hand. This extra effort required during the bidding may even result in worse performance during play and defense.
- if 531 count can't be mapped to HCP in an elegant way, experts would have to modify their systems based on a point count which hasn't proven anything (yet?).
- the overall added value for an expert is probably very limited (or even non existent).

Beginners on the other hand may benefit from this, but their focus should be elsewhere imo. Perhaps something nice for intermediates, dunno.

[Stephen Tu]

Take a look at
http://bridge.thomas...com/valuations/

People who have spent a ton of time looking at computer sims and various point counts basically concluded that for NT, 4/3/2/1/0.5 (for A through ten) is quite good already, with 4/2.8/1.8/1/.4 only a little better. For suit contracts, a lot have settled on 6/4/2/1 for the relative value of honors, or normalized to 4.5/3/1.5/.75/.25 to get something on the same scale as standard point count.

So 5/3/1 in your scheme seems likely to overvalue aces, and undervalue queens and jacks, on an average basis.

Also agree with Free's point that the value of tweaking these things is ultimately rather limited. Probably of more use when designing bidding software.

[jogs]

dake50, on 2013-March-29, 19:37, said:

I'm trying to chart 531 count as is done for HCP for 8-20 points.
Need help running sims or calculations. eg. take 12-14 HCP -> 531.
The basics of 531.
**A=5**, except A-single is 4.
**K=3**, except K-single is 1.5.
**Q=1**, except Q-single,Qx is 0.
J=0
xxx,xxxx is -1.
Q=2 if with A or K 3+length.
J=1 if with Q or K or A 3+length.
I have confidence 531 is better (at least for 1st evaluation) but
I am trying to get a handle on the impact of this new evaluation.
Is it so small, that my preference is just that, or big enough to
publish/persuade experts?

Axxxx, Axxx, Axx, A

Every one of those aces are worth a different value.
It's mostly not worth the effort to use another
point count. The variance of any and all methods
of evaluation is extremely high.
The secret of good bidding is making the proper
adjustment during the subsequent rebids after learning
more about the cards around the table.

[dake50]

Take a look at
http://bridge.thomas...com/valuations/

*** That's actually a source along with Aces Scientific (don't count stray Jacks).
Instead of comparing evaluators why not assemble a best evaluator?
If Andrews evaluator-tester has enough hands to make his correlation test reliable:
tricks won to evaluated, doesn't it have inherent a 'best' evaluator? 531 is that.

[dake50]

'jogs'
Axxxx, Axxx, Axx, A

Every one of those aces are worth a different value.
It's mostly not worth the effort to use another
point count. The variance of any and all methods
of evaluation is extremely high.
*** The variance is "extremely high" at an observed value of A=5 +-0.22 in Kxx=3-points??
That's 1/5 of a point! The "extremely high" variance occurs with quacks.
The very reason to start valuing them lower.

The secret of good bidding is making the proper
adjustment during the subsequent rebids after learning
more about the cards around the table.

*** Absolutly! But adjust down to less than a minimum opening can't be shown by any rebid, can it?
Responder bids to help opener promote working quacks in responder's suit(s).
That uptick in evaluation CAN be shown in a rebid.
Basically I'm saying adjustment from a poor start can't match adjustment from a good 1st evaluation.

[PhilKing]

dake50, on 2013-March-30, 18:39, said:

Take a look at
http://bridge.thomas...com/valuations/

*** That's actually a source along with Aces Scientific (don't count stray Jacks).
Instead of comparing evaluators why not assemble a best evaluator?
If Andrews evaluator-tester has enough hands to make his correlation test reliable:
tricks won to evaluated, doesn't it have inherent a 'best' evaluator? 531 is that.

I like his "Interactive evaluator." I tried this hand on it: ♠AKQJ ♥AKQ ♦AKQ ♣AKQ. "Binky HCP NT" count rated this hand at 34.8 but "Binky HCP suit" count rated it 37.4 - I fear some head-shaking if spades don't break.

More impressively, Binky trick count confidently predicted 15.06 tricks in no trumps - it knows a good hand when it sees it.

[rhm]

PhilKing, on 2013-March-30, 19:32, said:

I like his "Interactive evaluator." I tried this hand on it: ♠AKQJ ♥AKQ ♦AKQ ♣AKQ. "Binky HCP NT" count rated this hand at 34.8 but "Binky HCP suit" count rated it 37.4 - I fear some head-shaking if spades don't break.

More impressively, Binky trick count confidently predicted 15.06 tricks in no trumps - it knows a good hand when it sees it.

you might be interested in this response:

http://bridge.thomas...rstandings.html

Rainer Herrmann

[jogs]

'jogs'
Axxxx, Axxx, Axx, A

'dake50'
*** The variance is "extremely high" at an observed value of A=5 +-0.22 in Kxx=3-points??
That's 1/5 of a point!
....................

Think you misunderstood what I meant. That ace stiff is worth only
one trick. The ace from Axxxx potentially makes the fourth
and fifth 'x's' worth a trick. Therefore the ace from Axxxx
is worth much more than the ace from ace stiff.

[dake50]

Axxxx, Axxx, Axx, A

Think you misunderstood what I meant. That ace stiff is worth only
one trick. The ace from Axxxx potentially makes the fourth
and fifth 'x's' worth a trick. Therefore the ace from Axxxx
is worth much more than the ace from ace stiff. -- 'jogs'
*** Did you miss A-single is 4 not 5? So already noted as you suggest.

[jogs]

dake50, on 2013-April-14, 08:52, said:

*** Did you miss A-single is 4 not 5? So already noted as you suggest.

The singleton ace is taking only one trick.
The consensus expert opinion on hand evaluation
seem to think there is only one variable for
estimating tricks. I think they are all wrong.

There are two independent random variables.
Power for high card points and its location
within the partnership hands. Pattern or joint
pattern pair of the partnerships is the second
random variable. Shortness(singletons and voids)
is a component of pattern, not power. It isn't
worth any number of points. It may be worth extra
partial tricks.

Our tricks.

E(tricks) = trumps + (HCP-20)/3 + e

'e' is the error of the estimates from normal(μ,σ²).

The next term of this equation will adjust for
shortness.

E(tricks) = trumps + (HCP-20)/3 + f/s + e

The flatness/skewness adjustment.

[mikeh]

jogs, on 2013-April-14, 09:52, said:

The consensus expert opinion on hand evaluation
seem to think there is only one variable for
estimating tricks. I think they are all wrong.

lol

Sorry to lol, but this comment reveals a profound misunderstanding of expert hand valuation. I have never partaken in a serious bridge discussion with any expert, and I know and have even on occasion played with some truly wc players, in which the expert discussed any hand in terms of any metrics beyond saying something like: 'I held a good 17 count' (usually followed by giving the hand), or 'I had 3 working cards so I....'.

The giving of the point count is usually a shortcut start to the description of the hand, not the end of it, and nobody, in these discussions, goes on to say: it was really worth about 18.5 because of the interior sequence, or I discounted it to 16 because the A was stiff, etc.

I know of no expert who assesses the hand in terms of point count as a guide to the final contract, other than in the rare purely quantitative hands such as 1N 4N.

Experts tend to count tricks, losers and sometimes even ideas of how the play may unfold....not adding or subtracting points to come up with a numerical strength of the hand (again, with the exception of the purely quantitative).


In response to the OP, using a 531 with the suggested tweaks might (or might not) be useful in designing a simulation programme, at least until programming/hardware allows for the development of more real-life expert reasoning to be available in such programmes, but I can't se it having any relevance to humans. It significantly complicates matters for players who haven't yet acquired the ability to do proper valuation, and will likely do more harm than good, while better players simply don't need it.

[jogs]

mikeh, on 2013-April-14, 11:05, said:

The giving of the point count is usually a shortcut start to the description of the hand, not the end of it, and nobody, in these discussions, goes on to say: it was really worth about 18.5 because of the interior sequence, or I discounted it to 16 because the A was stiff, etc.

I'm not the one advocating a more precise point count system. On the opening bid only interested in knowing if the hand falls within the range of the chosen bid.
What I am suggesting is during the auction when a suit fit has been found, players should convert to counting tricks, not points. Use joint pattern of partnership to estimate tricks not to fine tune points. Combined trumps and shortness are components of joint pattern. Have not seen the experts suggest this approach to evaluating the tricks of a partnership on a specific board.