[tysen2k]

SUMMARY: This is (yet another) evaluation method to help quantify hand evaluation and reach more accurate contracts. This is a simple and easy-to-use method, yet provides more accuracy than many other “improved” methods.

This is an attempt to convert Evolved Binky Points into some formulas that are easier to remember than huge tables of numbers. This took a lot more work than I had initially thought. Trying to find a method that is more accurate that HCP, but doesn’t use a lot a complicated fractions is just about impossible. In order to keep things simple I’m avoiding fractions wherever I can, but it means there’s a slightly different scale of strength. I initially tried to make this scale to “normal points” so an average 13-point hand would be about 13 points in the new system. I just couldn’t do it without fractions that would be almost impossible to keep straight at the table. This new distribution has ~1.5 times the scale of normal points, so to convert these to “normal” subtract a third. To convert from normal to these values, multiply by 1.5.

A typical opening hand for standard methods should have ~19 points.

You should typically need:
34 for the 3-level
39 for the 4-level
44 for the 5-level
49 for the 6-level
54 for the 7-level

Of course these are just guidelines, like the 26 points typically needed for game. Certain scoring situations (like vul at IMPs) may push you to change these recommendations.

Here is my new method of hand evaluation:

HONOR POINTS (HP):
A = 6
K = 4
Q = 2
J = 1
(This is just HCP + Controls)
* Add 1 point for every suit that has 2+ honors (including the Ten)
* Doubletons: Don’t add the point for 2+ honors and subtract one additional point for QJ. (Don’t subtract one for Qx or Jx as these are already valued low enough)
* Singletons: Honors are valued as the next weakest honor (A=4, K=2, Q=1, J=0)

DISTRIBUTION:
Add points for both shortness and length
* Shortness points: 5/3/1 for void/singleton/doubleton
* Length points: 1 point for each card over 4 in a suit


REVALUATION (AFTER PARTNER BIDS)
This is a complicated matter and I’m working on a lot of different things to try and come up with an easy to use metric. For now use this simplified way:
* Don’t count shortness in partner’s suit (unless you find an 8-card fit elsewhere)
* Give +2 bonus for each trump you have over an 8-card fit

EXAMPLES:

KQJxx
-
QJxxx
AJT
* Initially: 14 HCP + 3 controls + 3 suits with 2+ honors = 20 HP. 5 for the void and 2 for the 5-card suits = 27 total points.
* Partner bids spades: +4 for the 10-card fit = 31 points.
* Partner bids hearts: – 5 for the void in partner’s suit = 22 points.

Kxx
Qx
JTxx
ATxx
* Initially: 10 HCP + 3 controls + 2 suits with 2+ honors = 15 HP. We get 1 distribution point for the doubleton = 16 total.
* Partner bids spades: No change = 16 total.
* Partner bids hearts: -1 for the doubleton = 15 points.


For NT contracts you can still use these HP, but for distribution simply count 1 point if you have a 5+ suit. You should need about:

34 HP for 2NT
37 HP for 3NT
44 HP for 6NT
48 HP for 7NT

Now all I need is a snappy name for my evaluation system. Since my initials are TS, I’ll call them TS points, or TSP. That will do for now I guess.


COMPARING EVALUATORS

I’ll extend my previously posted table of evaluator comparisons to see how much improvement you can expect by using this method:

             ERROR    SCORE
HCP         1.23    -0.49
HCP+321     1.07     0.00
HCP+531     1.05     0.07
Zar         1.05     0.08
BUMRAP+321  1.03     0.14
BUMRAP+531  1.02     0.21
TSP         1.02     0.21
Binky       0.99     0.32

ERROR is the average # of tricks there is in difference between how many tricks we think we can take and how many we actually take.

SCORE is an estimation of the IMPs/board we expect to gain against a team that uses a simple HCP+321 evaluation method. It’s a measure of how much payoff there is for using a better evaluation system.

HCP is A=4, K=3, Q=2, J=1

HCP+321 is HCP + 3 per void + 2 per singleton + 1 per doubleton

HCP+531 is the same with more points assigned to shortness

Zar is HCP + Controls + twice the length of longest suit + once the length of second-longest suit minus length of shortest suit.
http://public.aci.on.ca/~zpetkov/

BUMRAP is a substitute for HCP: A=4.5, K=3, Q=1.5, J=0.75, T=0.25

TSP is the method described in this article. It’s an attempt to find the best evaluator using simple whole numbers.

Binky is Thomas Andrew’s evaluator:
http://thomaso.best....dge/valuations/

[EricK]

When you calculate your figures, do you blindly assume the partnership will bid to the level indicated by the points, or do you assume they will eg check for Aces before bidding a slam etc?

Eric

[Fluffy]

I have heard, that by the time Milton Work (was him right?) invented the A=4,K=3,Q=2,J=1 count, it appeared in Europe another counting method:

A=7,K=4,Q=2,J=1,10=0.5. It was actually much more accurate than the one we are using, but it never became popular due to the need of a calculator to know how many points you have .

[MickyB]

tysen2k, on May 28 2004, 05:01 PM, said:

            ERROR    SCORE
HCP         1.23    -0.49
HCP+321     1.07     0.00
HCP+531     1.05     0.07
Zar         1.05     0.08
BUMRAP+321  1.03     0.14
BUMRAP+531  1.02     0.21
TSP         1.02     0.21
Binky       0.99     0.32

Reposted from RGB:

Cheers Tysen, very interesting.

Two points - firstly, after all that, your comparison shows little
difference between TSP and BUMRAP+531. Any ideas why this is? Could it
be that giving the ten a small value is significant enough to match all
the other changes you suggest?

Secondly, you seem to have gone down the Zar route of suggesting opening
a lot of hands with a couple of top honours and a bit of distribution. I
am convinced this approach is wrong. Say you are dealt Axxxx Axxxx xxx
void. Sure, your hand is great if you find a fit, and there is more
likely to be one than not. But if you pass on the first round, you can
come in on the second round and still reach your major game. If you open
the hand and there is a misfit, you are likely to get far too high and
you won't be able to do anything, as the shape and the top honours are
worth far less in no trump. Even so, Axxxx Axxxx xxx void is close to an
opening bid for me. What I cannot understand is the suggestion that void
xxx Axxxx Axxxx should also be opened. When you open a minor suit, you
are primarily showing strength for playing in 3NT, so why show that when
you don't have it?

[mikestar]

Scoring: IMP

---

This is one of a batch of hands I was using to practice Zar points--I though it would be interesting to compare counting methods.

West counts 14 HCP + 4 controls + 13 Zar DP = 31 Zars
East counts 10 HCP + 3 controls + 2 heart honors + 11 Zar DP = 26 Zars

After East raise hearts, West can add 1 (Zar Ruffing power) or 3 (Zar fit) for the ninth trump--lets compromise and say 2. Total Zar count = 59 which indicates an easy 11 tricks. 4H is laydown barring bizarre splits, but 5H has no chance on a club lead and needs a little luck without one.

Let's see what Tysen's new method says.

West counts 18 for honors, 4 for shape (1 per doubleton, 2 for the six card suit), and 2 for suits with 2+ honors = 24 TSP points.

East counts 13 for honors, 2 for shape (1 doubleton, 1 five card suit), and 2 for suits with 2+ honors = 17 TSP

West can add 2 points after the raise for his ninth trump.

Grand total is 43 TSP, just short of five level values.

[tysen2k]

EricK, on May 29 2004, 02:51 AM, said:

When you calculate your figures, do you blindly assume the partnership will bid to the level indicated by the points, or do you assume they will eg check for Aces before bidding a slam etc?

Unfortunately, yes, but all evalutators are treated the same.

[tysen2k]

MickyB, on May 29 2004, 09:26 AM, said:

Cheers Tysen, very interesting.

Two points - firstly, after all that, your comparison shows little
difference between TSP and BUMRAP+531. Any ideas why this is? Could it
be that giving the ten a small value is significant enough to match all
the other changes you suggest?

They are practically the same evaluator, just with different scales. When I've looked at it, the value of the Ten is pretty insignificant. TSP and BUM+531 are very close in performance, but of course TSP doesn't use fractions, which was the whole point.

Quote

Secondly, you seem to have gone down the Zar route of suggesting opening
a lot of hands with a couple of top honours and a bit of distribution. I
am convinced this approach is wrong. Say you are dealt Axxxx Axxxx xxx
void. Sure, your hand is great if you find a fit, and there is more
likely to be one than not. But if you pass on the first round, you can
come in on the second round and still reach your major game. If you open
the hand and there is a misfit, you are likely to get far too high and
you won't be able to do anything, as the shape and the top honours are
worth far less in no trump. Even so, Axxxx Axxxx xxx void is close to an
opening bid for me. What I cannot understand is the suggestion that void
xxx Axxxx Axxxx should also be opened. When you open a minor suit, you
are primarily showing strength for playing in 3NT, so why show that when
you don't have it?

To each his own. It depends on what you think the point of an opening bid is. If you use sound methods and your goal is to be constructive so you can reach the best games and slams, then that's great. My personal philosophy is to use opening bids to get the first word in before the opponents do. Shapely hands (especially 2-suiters) need the most bids to adequately describe themselves. You've got to start describing as soon as possible while the bidding is low. If your opps find a fit before you do, it may be too late. I'm personally fond of light opening systems on shapely hands and will gladly open 1♥ on xx KTxxx xx AJxx. But partner knows I will have hands like this and won't blindly go to 3NT. But he can preempt much more often in hearts since I'm opening much more often than other pairs...

[tysen2k]

mikestar, on May 31 2004, 09:24 AM, said:

Scoring: IMP

---

Thanks for your support, mikestar. My only hope is more people actually give this a try and see if they like it. Feedback (positive and negative) is always appreciated.

[mikestar]

TSP is clearly a good method, as is Zar. In fact they track quite closely. They produce indentical high card evalutions and after normalizing Zar to 4-3-3-3 = 0 points, they never differ by more than one point in evaluating any distribution among those with at least 1/2 % chance of occurance.

In all but one case where there is a differnce, Zar is one point higher. The anomaly is 6-5-1-1, which Zar evalutes to be 1 point inferior to 6-5-2-0 while TSP
evaluates the two shapes equal. Doing a rough average, Zar is about 2/3 point per hand (4/3 point per partnership) more aggressive than TSP on a 5 points =1 trick scale.

Zar's game target is 52 which reduces to 36 if we remove 8 Zar DP from each hand to normalize to 4-3-3-3 = 0.

TSP's game target is 39 which is 3 points more conservative. Taking both factors into account Zar will average 4 1/3 points more agressive than TSP--a not quite 1 trick average discrpancy in their predictions.

This analysis ignores differences in their respective revaluation methods and focuses on the intial evaluation. TSP's additonal points for each suit with 2+ honors is more aggressive than Zar's valuation for spread honors, so maybe the real gap is about 1/2 trick.

This difference should be wide enough for a reasonably sized sample to show which one corresponds best to reality. I'm slightly inclined to bet on Zar but I'm quite sure it is close--perhaps close enough that it make little difference to your chance of winning. Certainly either method is light years ahead of Goren 3-2-1.

Let the games begin!

[tysen2k]

mikestar, on Jun 1 2004, 04:24 PM, said:

This analysis ignores differences in their respective revaluation methods and focuses on the intial evaluation. TSP's additonal points for each suit with 2+ honors is more aggressive than Zar's valuation for spread honors, so maybe the real gap is about 1/2 trick.

2 average hands together will have 3-4 suits that they get the 2+ honors bonus. Stronger hands will get 4 or more bonus points, so it's not more conservative at all.

[inquiry]

Well of course one hand proves very little. For turnout fair play, here is a hand from Zar's webpage document (page 15)

---

This comes to 52 ZAR points, East 26, and opens 1S, West 26 in support.

How does it do with TSP? East hcp (A=6 scale), 15. +1 for two spade honors, +1 for 5th spade, +1 for doubleton club = 18. According to the scale of opening 19, not even an opening hand. West? 12 hcp, +1 for spade honors, +1 for fifth club, +2 (one for each doubleton). that comes to 16. 18 + 16 = 34, just enough for a three level bid, but 4♠ is cold. If EW were to enter teh bidding somehow, WEST could add two points for the ninth trump, but that brings total to only 36, still well short of that needed for game (39). Zar points out that if EW were to exxhange two small ♦ for AK, that his total would zoom to 62, and slam. Adding AK of ♦, even if to same hand (so you get the +1 bonus) would rasie the 36 to only 47, two short of the needed for slam, and if they ♦ were split, the total would only be 46, 3 short of slam.

I don't mean this as a slight towards TSP. Clearly the example hand was hand picked by Zar to make his point, and fits nicely with his point count system. No doubt on other hands that either method might be better than the other. It is interesting that in both example hands (mikes and this one), Zar was more aggressive than TSP. Aggression fits my style of play, maybe that is why I like ZAR. I wonder if ZAR will always be the more aggressive of the two methods. One thiong for sure, now that tysen isn't using fractional points, his count is easier to apply. Not sure if it is better or even as good as ZAR, time will tell.

Ben

[tysen2k]

inquiry, on Jun 1 2004, 10:04 PM, said:

Well of course one hand proves very little.

Of course...

Quote

It is interesting that in both example hands (mikes and this one), Zar was more aggressive than TSP.

I was hoping to separate accuracy from aggressiveness. I love aggressive methods as well. This post was also for rgb, and when I've posted what I considered to be "borderline" hands there before I distinctly got a lot of pushback that what I considered borderline was "not close to the average poster." So when I published my guildlines for what is opening strength, game strength, etc. it was more in line with what a standard bidder would be used to, not my own personal methods. As I said earlier in this post, my current methods allow me to open 1♥ on xx KTxxx xx AJxx (not close to an opening bid for either method )

[whereagles]

hcp point count is a simple and reliable method when you put together two balanced hands. However, it tends to over-evaluate hands that are highly misfitted and to greatly under-evaluate hands that fit well.

I only use hcp count when investigating a possible 3NT. If I find a heavy misfit, I count cover cards for pard's suit (or vice-versa) and look at my suits' quality. If I have a fit, I use the Losing Trick Count (and Law of Total Tricks for defensive bidding). Both the LTC and the LTT need some adjustments, but I find these methods quite easy to use and satisfying as a whole.

The LTC is the only method I know that doesn't need to add points for extra lenght. I believe that any method that has to add points for extra lenght is going in the wrong direction. Why? Because when you have long suit(s), what matters is how many top tricks you'll lose in the main and side suits, not whether you make the small spots or not.

Most count methods that try and "refine" hcp count take the view of counting winners and add lenght points. That goes against the spirit of the play in trump contracts, where you usually count losers, not winners.

[mikestar]

tysen2k, on Jun 1 2004, 11:50 PM, said:

mikestar, on Jun 1 2004, 04:24 PM, said:

This analysis ignores differences in their respective revaluation methods and focuses on the intial evaluation. TSP's additonal points for each suit with 2+ honors is more aggressive than Zar's valuation for spread honors, so maybe the real gap is about 1/2 trick.

2 average hands together will have 3-4 suits that they get the 2+ honors bonus. Stronger hands will get 4 or more bonus points, so it's not more conservative at all.

Point well taken. And in any event, Tysen could eliminate any residual difference in average agressiveness by using a lighter game target, maybe 1 point lighter would be enough. Then the comparison comes down to hands where TSP and Zar disagree and seeing which is right more often.

Note that the disagreement is always 0 for HCP and never more than 1 point (=0.2 tricks) for DP. Zar is always more agressive on DP when they disagree, except with specifically 6-5-1-1.

TSP will tend to be more agressive with honor location adjustments than Zar.

My suspicion is that with comparably agressive game targets, you will find very little difference in the accuracy of the methods.

In any case, the difference in objective accuracy will be outweighed by user comfort with the method. They seem about equal to me in this regard--if my experience is typical, there is little to choose.

My own preference is for Zar -- doubling the Goren targets is easier than using 34, and changing my entire method of counting is not so hard after all. In particular, Zar is quite correct that the Zar DP values for the more common shapes start sticking in your mind so you don't have to count them. Of course with more practice this might happen with TSP--but I didn't get this effect in such a short time with 1-3-5. Pehaps the adding and subtracting lengths has some mnemonic value, at least for some players.

Nevertheless, I stongly endorse TSP as an excellent count.

[mikestar]

Two hands still prove nothing, but I notice the hand Ben gives from Zar's site is perfect fit -- put the Q of trumps in hearts or even clubs and the chances of game are not as good. But the example I gave (hand dealt) is trump rich--move the Q of trumps to clubs and the chance of 11 tricks gets larger (though so does the chance of only taking 9 tricks).

[inquiry]

mikestar, on Jun 2 2004, 03:01 PM, said:

Two hands still prove nothing, but I notice the hand Ben gives from Zar's site is perfect fit -- put the Q of trumps in hearts or even clubs and the chances of game are not as good. But the example I gave (hand dealt) is trump rich--move the Q of trumps to clubs and the chance of 11 tricks gets larger (though so does the chance of only taking 9 tricks).

I actually noted that this was a perfect hand for ZAR, and no doubt hand picked by him to make his point..... this is why I said.

Quote

Clearly the example hand was hand picked by Zar to make his point, and fits nicely with his point count system. No doubt on other hands that either method might be better than the other.

BTW, I agree on the premise that ZAR points are easier to remember what is needed for game, slam, etc. But that would be quickly replaced by tysens numbers if you worked with them a lot.

Ben

[MickyB]

A few people have been saying they have a slight preference for Zar over TSP, I'd just like to register my strong support for TSP! What I have seen of Tysen's research has impressed me greatly, keep up the good work

[inquiry]

Here is one for the computer geeks to check out.

In my manual checking of ZAR verus the other methods (Bum rap and TSP), what I find is that ZAR points can climb into the stratosphere very quickly with two shapely hands including a fit, and still be off two aces or AK and A. It looks like, to me, these are the hands where ZAR points are off two or three tricks, which gives them the high "error rate" and relatively low "score" compared to Goren.

Now, to beign with, I seriously doubt if the "score" thing that tysen has published here is right. looking at the results of all the hands from ZAR's database (which is publically viewable), and comparing hundreds for real world deals the old fasshion way by hand. But I do believe the ERROR is right, if not actually low, for the way the analysis was done. There is this "flaw" so to speak with ZAR. The points, as noted above, can pile up very quickly, but you may be off one or two quick tricks. So you need to use "blackwood" or a cue-bidding sequence on ZAR hands that are slam going MUCH MUCH more than you ever needed to do with Goren. And since ZAR is more agressive (see above post), it makes sense that lack of a reality check (cue-bidding or ace asking) will result in overbidding on such hands as I described above.

It would be nice, if the test being used (not only for ZAR, but for the others), used an algorithm that tested for off two aces or off two in a suit with neither partner stopping it from running. That is, to assume real bidders were using the systems. Such a built-in check would help all the systems avoid overbidding, but since ZAR is the most aggressive it would help it the most.

Anyway, TSP does seem to do pretty well, but in my examples, TSP

Ben

[Zar]

*** Ben wrote: “Now, to begin with, I seriously doubt if the "score" thing that tysen has published here is right.
<

I am surprised you get these results seriously with no explanation HOW they have been “scored ”, WHAT was programmed and what BOARDS were fed in. No information whatsoever, just plain words and BOOM – we have a GREAT result :-)

Since you already noticed that, I’ll just post the usual results for the Standard GIB boards for Game and Slam, and as usual all records are available for checking.

If you pay attention a bit, you’ll easily see the THREE FUNDAMENDAL mistakes made in this “yet another method” as it was called, which is the natural reason for the catastrophic results on the Stndard GIB boards – both slams and games. If cannot see these three fundamental mistakes, let me know. Here are the results.

For the Games from the Standard GIB boards (63,057 boards total)

================Overall Results ============================

The WTC ( number of tricks > 9) got 19666 contracts
Fit TSN Points ( fit points >38) got 27610 contracts
GOREN 3-2-1 ( HCP+3-2-1> 25 ) got 32688 contracts
GOREN 5-3-1 ( HCP+5-3-1> 25 ) got 41045 contracts
Basic Zar Points (no fit points>51) got 49794 contracts
Fit +3 Zar Points(+3 extra trmp>51) got 55802 contracts


For the Grands from the Standard GIB boards (10,344 boards total)

================Overall Results ============================

GOREN 3-2-1 ( HCP+3-2-1> 36 ) got 1427 contracts
The WTC ( number of tricks > 12) got 1543 contracts
Fit TSN Points ( fit points >53) got 1587 contracts
GOREN 5-3-1 ( HCP+5-3-1> 36 ) got 2913 contracts
Basic Zar Points (no fit points>66) got 3753 contracts
Fit +3 Zar Points(+3 extra trmp>66) got 5729 contracts


And to have no doubt, here is how the calculations were done:

/////////// GRAND slam requirements

if( ZPnow > 66 ) ZP++;

if( ZPnow // HCP + CTRL
+ 3*( max( 0, (L[0][fitCol] + L[1][fitCol] -8) ) ) // FIT points
> 66 ) ZPfit++; // check for Grand

if( Gor > 36 ) Goren++;
if( Wtc > 12) WTC++;
if( p135 > 36) G135++;

if ( TSN // HCP + CTRL
+ 2*( max( 0, (L[0][fitCol] + L[1][fitCol] -8) ) ) // FIT points
+ dN123 + cN123 + dS123 + cS123 // 1-3-5 for N and S
> 53 ) TSNfit++; // check for Grand
/*
34 for the 3-level
39 for the 4-level
44 for the 5-level
49 for the 6-level
54 for the 7-level
*/


Cheers:

ZAR

[MickyB]

I can't see any mistakes in the description of TSP. What do you believe they are?

You have probably answered this elsewhere, but - you have all this data for the proportion of games bid. What about the times that a particular method of evaluation will get you to a game that goes off?

Cheers