[Cascade]

A simulation suggested you were a little too conservative.

        <25     25     26     27     28     29     30   >30     Sum
  <9    311     20      5      1      0      0      0      0    337
   9   1107    153     38      3      1      0      0      0   1302
  10   1814    557    245     66     15      1      0      0   2698
  11   1100    964    659    279    103     25      3      1   3134
  12    272    329    491    410    279    129     32     10   1952
  13     22     33     79    122    104    115     64     38    577
Sum    4626   2056   1517    881    502    270     99     49  10000
        6.3%  17.6%  37.6%  60.4%  76.3%  90.4%  97.0%  98.0%  25.3%

I generated 10000 hands with a combined total of 24 or more hcp and a guaranteed 8 card or longer fit somewhere.

I then simply found the longest fit (combined length between the two hands) and calculated the double dummy tricks.

The results are tabulated against QPs + combined trumps.

[Cascade]

1000 hand simulations similar to the above with additional criteria for short suits

QPs + combined trump length with no shortage

        Low     24     25     26     27     28     29     30   High    Sum
Low      52     14      3      0      1      0      0      0      0     70
   9    109     68     32     12      1      1      0      0      0    223
  10    110    120     82     38      8      5      0      0      0    363
  11     20     38     64     51     34     12      1      2      0    222
  12      1      3      5     26     28     16      7      1      1     88
  13      0      0      1      2      7      8      6      5      5     34
Sum     292    243    187    129     79     42     14      8      6   1000
        0.3%   1.2%   3.2%  21.7%  44.3%  57.1%  92.9%  75.0% 100.0%  12.2%

QPs + combined trump length with a singleton in the shorter trump hand (or either hand if equal length)

        Low     24     25     26     27     28     29     30   High    Sum
Low      18      7      0      1      0      0      0      0      0     26
   9     75     25     10      3      0      0      0      0      0    113
  10    105     78     56     10     10      0      0      0      0    259
  11     51     91     83     55     34      7      3      0      0    324
  12     12     19     51     55     48     29     13      2      1    230
  13      0      0      4     10      7     12      8      5      2     48
Sum     261    220    204    134     99     48     24      7      3   1000
        4.6%   8.6%  27.0%  48.5%  55.6%  85.4%  87.5% 100.0% 100.0%

QPs + combined trump length with a singleton in the longer trump hand

        Low     24     25     26     27     28     29     30   High    Sum
Low      19      8      1      1      0      0      0      0      0     29
   9     60     42     19      3      0      0      0      0      0    124
  10     92     92     68     19      8      1      0      0      0    280
  11     45     70     90     70     35     15      6      0      0    331
  12      6     17     30     52     32     22     16      1      0    176
  13      1      0      6      5      8     17      7     13      3     60
Sum     223    229    214    150     83     55     29     14      3   1000
        3.1%   7.4%  16.8%  38.0%  48.2%  70.9%  79.3% 100.0% 100.0%

QPs + combined trump length with a void in the shorter trump hand (or either hand if equal length)

        Low     24     25     26     27     28     29     30   High    Sum
Low       5      0      1      0      0      0      0      0      0      6
   9     40     11      4      1      2      0      0      0      0     58
  10     61     49     25      2      1      0      0      0      0    138
  11     91     86     81     31     15      2      1      0      0    307
  12     52     56     62     59     45     14      8      1      1    298
  13      8     20     30     51     32     26     15      8      3    193
Sum     257    222    203    144     95     42     24      9      4   1000
       23.3%  34.2%  45.3%  76.4%  81.1%  95.2%  95.8% 100.0% 100.0%

QPs + combined trump length with a void in the longer trump hand

        Low     24     25     26     27     28     29     30   High    Sum
Low      16      3      1      1      0      0      0      0      0     21
   9     66     34     14      8      2      0      0      0      0    124
  10     92    104     52     29      8      1      0      0      0    286
  11     37     84     88     61     30     17      5      0      0    322
  12      5     16     39     42     39     23     12      2      2    180
  13      0      0      2      9     17     10     15     11      3     67
Sum     216    241    196    150     96     51     32     13      5   1000
        2.3%   6.6%  20.9%  34.0%  58.3%  64.7%  84.4% 100.0% 100.0%  

QPs + combined trump length with a singleton or void (non-matching) in both hands

        Low     24     25     26     27     28     29     30   High    Sum
Low       4      1      0      0      0      0      0      0      0      5
   9     42     12      4      1      0      0      0      0      0     59
  10     66     63     20     11      3      2      0      0      0    165
  11     77    104     91     41     23      3      1      0      0    340
  12     20     39     70     83     49     29      8      5      1    304
  13      4      7     10     24     32     17     16     13      4    127
Sum     213    226    195    160    107     51     25     18      5   1000
       11.3%  20.4%  41.0%  66.9%  75.7%  90.2%  96.0% 100.0% 100.0%      

[straube]

Free, on Apr 28 2010, 03:02 AM, said:

straube, on Apr 28 2010, 07:47 AM, said:

Ok, I looked up what Mike Lawrence had to say in "I fought the Law".  Assuming a trump fit, he looked at the partnership's two shortest suits.

Let's say 5431 is opposite 4144.  We have a trump fit (spades) and shortness in clubs and hearts.

Our Short Suit Total is 1+1 = 2.

Looking at that 5332 opposite 5332 our short suit total is 3 + 2 = 5 because you can't be looking at the same suit (clubs) twice.

I gathered that the SST difference equates to roughly a trick and a trick is roughly a king or 2 QPs.  So perhaps if I subtract twice the SST from the QP total I can get a number and then I can use this number as a check before deciding whether to dcb and venture into the 5-level.

You were suggesting that 5332 opposite 5332 needed 21 QPs.  So if we subtract 2 * 5 from 21 we'd get 11.

With your 5143 opposite 5413 you suggested we might need 15. If we subtract a SST of 2*2 from 15 we get 11.

Am I on the right track?  Suggestions?

Problem is that you can have enough with 15 if you have the 5431s, but it's not always the case. It depends on what partner holds in our short suit, and what we hold in partner's short suit.

I'm not sure where you're going, but am I correct to assume you consider 11 some kind of constant?

I totally get visualization. The nice thing about visualization is that you can rule out slam for certain hands. You can give partner the perfecta and if it's still not enough, you don't explore. Of course, sometimes partner can have several combinations of cards that are sufficient for slam. But the fewer QPs one's side holds, the more perfectly aligned they have to be to produce slam and then one has a probability decision...what is the likelihood that partner's QPs are useful vs what is the risk of them not being useful and getting too high trying to find out? That's what this is about for me.

I don't think 11 is a constant exactly. I was trying to come up with an equation that results in a number that would signal the likelihood of a slam being present. With this particular rule, 11 seems to be the point at which slam has a likelihood. Of course, that's just looking at 2 hands.

Here's 2 more from last night

AKx KQ KQTxxx Axx vs Tx Axxx Axxx QJx
SST=4 so 21- (2*4)=13

AQx AKQxx Qxx Qxx vs xxxxxx x AJx Axx
This hand feels funny because the shortness of one hand is opposite length of the other. anyhow...
SST=4 so 18- (2*4)=10

I still think that length in trump matters. Obviously it doesn't matter if the hands are mirrored or there is no shortness, but hands with shortness and only an 8 cd fit can have handling problems that 9 cd fits don't have. Perhaps SSTs indirectly take trump length into account.

[straube]

Cascade, thanks a lot for your work. Seems like a degree of shortness is about a trick.

If you're interested in seeing how the SSTs bear on the solution, maybe you can run QPs (not QPs + trump length) for each scenario. It can confirm or disprove the SST rule.


4333 vs 3334 (specific) for 6 SSTs

5332 vs 3442 (specific) for 5 SSTs

4432 vs 2443 (specific) for 4 SSTs

5431 vs 2443 (specific) for 3 SSTs

5431 vs 1543 (specific) for 2 SSTs

5440 vs 1543 (specific) for 1 SST

[straube]

I've looked into Lawrence's method even more. For trick estimation he says that 13-SST=tricks expected if we have 19-21 hcp. For each 3pt range (22-24, 25-27) he adds 1 trick.

hcp.....tricks..QPs
19-21.....7.....12
22-24.....8.....14
25-27.....9.....16
28-30.....10...18
31-33.....11...20
34-36.....12...22
37-39.....13...24

Is that right? I'm not sure.

So converting hcps to QPs...

(QP total)/2 + 7 - SST = trick expectation.

let trick expectation = 12 (solving for small slam)

(QP total)/2 + 7 - SST = 12

(QP total) + 14 - 2*SST=24

QP total - 2*SST = 10

So if the total is ten or higher then slam should be on. All of this assumes that the QPs are working and that may not be knowable until after dcb. So perhaps the more we rely on distribution and perfectas the more reluctant we should be to use this rule.

[Free]

Would have to analyze hand by hand, but your calculations seem right. Your rule obviously is at most as accurate as Lawrence's method.

[Cascade]

I think this is what you asked for.

I changed the orientation of the table - I hope that is not too confusing.

The distributions shown are precise 5431 means five spades four hearts three diamodns and one club etc.

Frequency 4333 4333:
        Low      9     10     11     12     13       Sum
  12      0      0      0      0      0      0         0
  13      8      3      2      0      0      0        13
  14     67     41      9      0      0      0       117
  15    108     90     19      2      0      0       219
  16     75    121     47      2      0      0       245
  17     24     71     72     10      3      0       180 2%
  18      4     28     56     23      2      0       113 2%
  19      1      4     18     31      4      0        58 7%
  20      0      0      4     17      4      0        25 16%
  21      0      0      2      3      9      1        15 67%
  22      0      0      0      3      7      1        11 73%
  23      0      0      0      0      2      1         3 100%
  24      0      0      0      0      0      1         1 100%
Sum     287    358    229     91     31      4      1000

Frequency 5332 3442:
        Low      9     10     11     12     13       Sum
  12      0      0      0      0      0      0         0
  13      7      6      0      0      0      0        13
  14     24     36     36      4      0      0       100
  15     43     87     82     26      0      0       238
  16     18     66    127     44      3      0       258 1%
  17      4     20     69     61     10      0       164 6%
  18      1      5     28     52     31      1       118 27%
  19      0      0     11     26     22      3        62 40%
  20      0      0      1     12     16      6        35 63%
  21      0      0      0      0      6      1         7 100%
  22      0      0      0      1      1      1         3 67%
  23      0      0      0      0      0      2         2 100%
  24      0      0      0      0      0      0         0
Sum      97    220    354    226     89     14      1000

Frequency 4432 4243:
        Low      9     10     11     12     13       Sum
  12      0      0      0      0      0      0         0
  13      5      4      1      0      0      0        10
  14     20     51     35      3      0      0       109
  15     27     84    112     14      0      0       237
  16      4     68    130     37      0      0       239
  17      3     20     78     74     13      0       188 7%
  18      0      2     31     46     18      0        97 19%
  19      0      0     11     20     30      4        65 52%
  20      0      0      1     13     25      2        41 65%
  21      0      0      0      1      5      4        10 90%
  22      0      0      0      1      2      1         4 75%
  23      0      0      0      0      0      0         0
  24      0      0      0      0      0      0         0
Sum      59    229    399    209     93     11      1000

Frequency 4531 4243:
        Low      9     10     11     12     13       Sum
  12      0      0      0      0      0      0         0
  13      2      6      2      1      0      0        11
  14     19     23     36     16      0      0        94
  15     16     46     99     81      8      0       250 3%
  16      1     33     86    106     26      0       252 10%
  17      0      4     42     75     42      3       166 27%
  18      0      3     15     41     45      2       106 44%
  19      0      0      1     12     42     11        66 80%
  20      0      0      2      3     25      8        38 87%
  21      0      0      0      0      6      6        12 100%
  22      0      0      0      0      1      3         4 100%
  23      0      0      0      0      0      1         1 100%
  24      0      0      0      0      0      0         0
Sum      38    115    283    335    195     34      1000

Frequency 4531 5143:
        Low      9     10     11     12     13       Sum
  12      0      0      0      0      0      0         0
  13      0      1      7      2      0      0        10
  14      2     10     33     45      9      0        99 9%
  15      1      9     56    124     47      2       239 21%
  16      0      4     27    102    114      6       253 47%
  17      0      0      5     59     98     11       173 63%
  18      0      0      1     21     75     17       114 81%
  19      0      0      1      6     34     15        56 88%
  20      0      0      0      0     14     21        35 100%
  21      0      0      0      0      3     11        14 100%
  22      0      0      0      0      0      7         7 100%
  23      0      0      0      0      0      0         0
  24      0      0      0      0      0      0         0
Sum       3     24    130    359    394     90      1000

Frequency 4540 5143:
        Low      9     10     11     12     13       Sum
  12      0      0      0      0      0      0         0
  13      0      1      1      0      1      0         3 33%
  14      0      3     14     27     35      8        87 49%
  15      0      4     23     81    126     34       268 60%
  16      0      1     12     71    136     60       280 70%
  17      0      0      6     29     79     57       171 80%
  18      0      0      0      6     44     42        92 93%
  19      0      0      0      1     11     41        53 98%
  20      0      0      0      1      6     24        31 97%
  21      0      0      0      0      2      9        11 100%
  22      0      0      0      0      0      2         2 100%
  23      0      0      0      0      1      0         1 100%
  24      0      0      0      0      0      1         1 100%
Sum       0      9     56    216    441    278      1000

[straube]

Thanks for running those hands. They seem to bear out that that SST rule is not useful.

I've come to doubt that tracking SSTs and not trump length is faulty because you need both trump and shortness to take extra tricks.

Say I have 6143 opposite 1534. I have here a SST of 2 but not even an 8-cd trump fit.

So if there's a rule of thumb after QPs have been found to help decide whether to dcb (go to the 5 level), it might look something like...

QPs + (value for shortness times value for extra trump length)= a certain threshold

[DinDIP]

SST could be helpful but you need to invert it. Simpler is to use something like this:
Investigate slam (i.e. move into DCB) if QP + shortness is >=21

For shortness count
if an 8-card fit: each void = 2 and each singleton = 1
if a 9-card fit: first void = 3, first singleton = 2, first doubleton = 1, second void = 2, second singleton = 1
if a 10+card fit: first and second void = 3, first and singleton = 2, first and second doubleton = 1

Note: shortages in both hands are counted but not when they are opposite one another (i.e. if you have a singleton opposite a doubleton and a 9+card fit you only count for the singleton).

Obviously this is only a guide: if you have an 8-card fit but a solid side suit then you should count void = 3 and singleton = 2. And if you have hands that fit well -- partner's singleton is opposite your three small -- then count more (3 in that case as you know the partnership has no wasted K or Q opposite partner's singleton).

Most importantly, this should only be used when you are starting to relay. With experience you'll find that rules/guides like this are of little (if any) value: I've been relaying for nearly 30 years -- on and off -- and never use this or any other guide. Instead, I just use a simple test:

If I locate partner's QP in the most favourable places is slam solid? If it is, can I find out that partner has that hand by the five level or, more accurately, can I stop at a safe level if partner doesn't have that hand?

This is where it gets a little tricky: if there is only one honour permutation that makes slam worth bidding, how likely is that permutation? That's difficult to judge -- it helps to look at the relative frequency of the different honour permutations when the hand opposite has 15/16+. (I.e. if partner has 6QP how often is that AA, KKK, AKQ, KKQQ, AQQQ and KQQQQ.) But that is onlya general guide: the exact frequencies are affected by the specific honours the relayer holds. In general, however, honour combinations with lots of one honour are less frequent than ones with some of all honours. So, if partner shows 9QP and slam is only good if he has three aces then be wary of DCBing unless it is clear that you can stop safely. (Note that in most DCB methods partner will bypass more steps with more honours so you need to be very wary if you're looking him to hold AAA rather than AKKQQ.)

Try this in practice: using the earlier example of 8 QPs and 3-4-4-2 opposite AQxxx x KQx AQJx. (Much better to have described the UNBAL hand but . . . ) Here Kxx Qxxx Axxx Kx is a great slam even with a wasted HQ so you should be thinking of investigating. But whenever partner has the HK you know slam is poor (at best a finesse and a 3-2 break, unless partner turns up with the SJ, when it's still at best a finesse). And it's also the case that partner is more likely to have the HK than the SK or CK. So, unless your methods allow you to distinguish between partner's different honour holdings in H then I'd be conservative and sign off in game. Which is exactly why this hand shouldn't have relayed when it discovered partner was BAL; rather, it should have shown its shape or, at least, its shortness and S length.

The best thing to do is practise a lot: you don't even need partner. Just get old hand records or bulletins from major tournaments and bid all the hands. (I try to bid at least 500 -- of all types, not just slam hands -- before playing in any serious tournament.)

David

[straube]

Thanks for a very useful post. Your rule of thumb seems right in that it accounts for both shortness and trump length. It also seems pretty easy to calculate at the table. Did you make this rule of yourself or did you find it somewhere? What do other folks think? Are the point awards about right? Seems about right to me.

[DinDIP]

straube, on May 1 2010, 12:01 PM, said:

Thanks for a very useful post. Your rule of thumb seems right in that it accounts for both shortness and trump length. It also seems pretty easy to calculate at the table. Did you make this rule of yourself or did you find it somewhere? What do other folks think? Are the point awards about right? Seems about right to me.

My once or twice a year partner or teammate (shevek) suggested many years ago, when others were learning our symmetric system, that the rule should be QP + shortages (in both hands, counting 3-2-1). I thought that was a bit much when we only had an eight-card fit so did a little bit of testing -- not much I hasten to add (because I don't use the guideline) -- and the results seemed to work.

But I can't emphasise strongly enough that more practice is the best recommendation.

David

[straube]

these are pulled from Cascade's simulations....


...............................fit...............shortness.............QPs.................slam success......qps under rule
4333 vs 4333............7...................none.................21...................63%..................21
4432 vs 4243............8...................db, db...............20...................65%..................21
4531 vs 4243............8...................sg, db...............19...................80%..................20
4531 vs 5143............9...................sg, sg...............17...................63%..................18
4540 vs 5143............9...................vd, sg...............16...................70%..................16

Your rule works pretty close. It seems like you need to give a bump for the 8 cd fit if there's a ruffing value. Notice that there's only 15% difference between hand sets 2 and 3. I imagine this is because only 1 ruff is usually to be had with an 8 cd fit.

I'm also a little surprised that 4333 vs 4333 yields a 63% chance of slam with only 21 QPs. Seems like most people's experience would suggest 22 are needed for balanced mirrored distribution.

[cherdanno]

awm, on Apr 12 2010, 02:56 PM, said:

I'll just mention that I've played a QP-based relay system for quite a while now and I really don't think of things in this way. My tendency is just to visualize possible hands for partner once the QP total is known and try to figure out what we can make opposite various holdings. It really depends a lot on the shapes and how the hands fit.

I still think this is only good advice in this thread. All of the other rules will make you a worse bridge player in the long run (compared to trying to visualize partner's hand in all such auctions).

straube, on Apr 12 2010, 03:37 PM, said:

So say partner has shown 8 QPs and 3-4-4-2 and I have AQxxx x KQx AQJx.  That's 19 if I did the math right.  Do I ask?  Still looking for rules of thumb here.

Well if you give partner perfectly fitting cards (♠K ♥A ♦A) then grand could be on a finesse. If you give him terribly misfitting cards (bad trumps, ♥KQ ♦A ♣K) then you could go down in 5, but opposite most badly fitting hands (♠K and heart wastage, or the previous hand with better trumps) small slam will be on a finesse. So it must be right to investigate slam.

However, you may also notice that it's better to be relayed than to relay when you are unbalanced. Also, the time to wonder whether you should investigate for slam is before you have told opponents his complete shape.

[straube]

The good advice I've received on this thread has always been accompanied by the admonition that visualization is the way to go. I've agreed, but I thought it might be useful to see if some guidelines could help.

We haven't discussed much such things as bad splits. That in combination with partner not having the right cards can also make the 5-level unsafe.

We also haven't talked about the probability of partner having the right cards based on his known shape. For instance, if partner has 3-4-4-2 and 8 QPs and I have AQxxx x KQx AQJx odds ought to favor partner having red honors over black honors. It's pretty important on this hand that partner have the SK.

So when I visualize, I have to factor in bad splits and the likelihood of partner having useful cards.

Visualization and experience are great, but I think doing simulations (like Cascade has done) is also useful. So is having a general idea of how many QPs on average are needed for various hand patterns.

My system does a pretty good job of having the balanced hand do the relaying and we all know that this is important. Even so, sometimes the unbalanced hand winds up captain and then getting a feel for whether the 5-level should be ventured is useful.

It's hard to determine whether to explore for slam until I know partner's exact shape. Obviously, distribution counts a lot toward deciding whether a slam can be made. My partner and I reverse relay when opener has a limited hand and sometimes we can end auctions early. I think more important is to organize our system so that the captain hand declares.

So yeah, I get it. I'm not going to get a rule of thumb that has much utility, but I think I've learned a bit by trying. Appreciate everyone's help.

[DinDIP]

straube, on May 2 2010, 11:23 PM, said:

The good advice I've received on this thread has always been accompanied by the admonition that visualization is the way to go.  I've agreed, but I thought it might be useful to see if some guidelines could help.

<snip>

So yeah, I get it.  I'm not going to get a rule of thumb that has much utility, but I think I've learned a bit by trying.  Appreciate everyone's help.

In my experience, it's hard to find rules that are useful, easy to use and sufficiently accurate. But the process of working out why they don't work is very valuable -- with the proviso that the process involves bidding lots of hands to work out why. (Simulations can help but are less valuable than bidding actual deals.) I found when I did this years ago -- trying to work out a better method of hand evaluation -- that I never did discover a new metric that met the necessary criteria (useful, easy to use and sufficiently accurate) but my hand evaluatiion skills improved significantly because of all the deals I examined, trying to ascertain why the new value did (or did not) work on a particular deal. So, keep on bidding those hands!

David