[MickyB]

We are very happy with the scans that we use when facing an unbalanced hand. When you have a 17-19 NT opposite a hand with a shortage, you can usually place all honours by

- Finding out about 3-2-1 points
- Scanning the first suit for 0/2 or 1/3 of the A/K/Q
- Scanning the second suit similarly

When opener is weaker, say an 11-13 NT, there are more possibilities, such that it might be best to scan the third suit as well, or carry out some other sort of cross-check. Any thoughts on this? What about when opener has a 14-16 NT?

This method of scanning doesn't really seem to work when responder has no shortage. Any alternative suggestions for scanning 5422s and the like?

[wclass___]

Post your 1NT relay structure.

Why do you prefer 3-2-1 vs 2-1? I don't reckon any 3-2-1 vs 2-1 discussions, but i would like to hear your arguments before i comment on this.

[MickyB]

3-2-1 obviously does a much better job of showing the value of the hand. How are you meant to judge whether to bid slam when you don't know if partner has several queens or no queens, in addition to the aces and kings he is showing?

What are the advantages of 2-1? The main one is, presumably, making it easier to locate the aces and kings later, but we don't have any problems with that in most scenarios, and even in other situations I doubt the gains from this could counteract the disadvantages.

[mgoetze]

wclass___, on 2012-March-28, 10:37, said:

Why do you prefer 3-2-1 vs 2-1? I don't reckon any 3-2-1 vs 2-1 discussions, but i would like to hear your arguments before i comment on this.

Presumably, Queens can also take tricks some of the time.

[wclass___]

MickyB, on 2012-March-28, 11:10, said:

3-2-1 obviously does a much better job of showing the value of the hand. How are you meant to judge whether to bid slam when you don't know if partner has several queens or no queens, in addition to the aces and kings he is showing?

Have I misread OP? I thought responder is relaying and opener who is limited say 17-19 NT describes his shape.

[straube]

mgoetze, on 2012-March-28, 11:14, said:

Presumably, Queens can also take tricks some of the time.

We frequently find that we miss out on a slam because we're unaware of a jack or jack/ten combination. I definitely want to know about queens.

[wclass___]

mgoetze, on 2012-March-28, 11:14, said:

Presumably, Queens can also take tricks some of the time.

Even two of clubs takes a trick once in a while.

[wclass___]

straube, on 2012-March-28, 11:31, said:

I definitely want to know about queens.

You miss the point. I'm not talking of whether you need to know about queens or not, but how to inquire about them in the best way. What are responses to your 3-2-1 ask and how wide it is?

[straube]

wclass___, on 2012-March-28, 11:36, said:

You miss the point. I'm not talking of whether you need to know about queens or not, but how to inquire about them in the best way. What are responses to your 3-2-1 ask and how wide it is?

We use awm's method which I hopefully describe correctly below.

After we show how many queen (relay) points we have, we scan each suit, longest to shortest, tie going to the higher ranking.

We stop with even, skip with odd #. If there's a singleton, we stop with a king or ace and skip without. We then focus on the first odd suit and stop with the A or Q or skip with the K or the AKQ combination. If there are no odd suits, then we skip with the king for the first even suit where we hold two honors. After that, relay captain may specifically ask for jacks if space allows. Usually we don't have room for that.

[wclass___]

straube, on 2012-March-28, 12:07, said:

After we show how many queen (relay) points we have

I was asking for this part. e.g. If i opened 11-15 1♦ and after shape ask for controls responses would 3-4-5-6, how would it look for QP and in other common situations?

[MickyB]

wclass___, on 2012-March-28, 11:30, said:

Have I misread OP? I thought responder is relaying and opener who is limited say 17-19 NT describes his shape.

Sorry, I wasn't clear - I'd just made this post and should have linked to it.

Despite the fact that more is known about opener's hand than responder's, we have responder describe to opener. Relays work best when an unbalanced hand describes to a balanced hand. However, I suppose there might be a case for allowing responder to ask as well, to be used when he has no shortage.

[straube]

I'm sorry, but I'm not sure what you are asking. Are you asking about continuations after 1D-1H, 1N etc after which responder shows his pattern and queen points?

[MickyB]

straube, on 2012-March-28, 12:40, said:

I'm sorry, but I'm not sure what you are asking. Are you asking about continuations after 1D-1H, 1N etc after which responder shows his pattern and queen points?

Yes, that sort of thing. Also 1M:2C relay, 1H:1S Kaplan Inversion leading to relays, some 4SF auctions, etc.

[wclass___]

straube, on 2012-March-28, 12:40, said:

I'm sorry, but I'm not sure what you are asking.

Responses to a bid asking for a QP.
+1=how many QP.. etc.

[straube]

MickyB, on 2012-March-28, 12:48, said:

Yes, that sort of thing. Also 1M:2C relay, 1H:1S Kaplan Inversion leading to relays, some 4SF auctions, etc.

Well, it's easy if responder is captain and opener has a limited hand. We have various relays in these situations and we assume that opener has at least 6 QPs (that's our base anyway). So after we know opener's shape, we ask QPs starting with 6. We zoom at 9 because it's unusual for opener to have 10 and not open 1C.

After 1D-1H, 1N it's impossible for responder to show his exact shape before 3N has been passed. He can show some certainly, but not all. We allow for responder to show his approximate shape, but then we use dual captaincy...meaning that opener registers an opinion for the right strain and how much he likes his hand. I suppose you could continue with relays, however. After approximate shape, opener could ask for queen points. Responder would use S1 as say 10 or fewer and S2 at 11 and so forth. Not sure where the cutoff would be.

[benlessard]

My method for unbalanced hand is

1 -asking/showing shape/relay break
2- setting up trumps & keycarding with one bid
3- scanning
4- rarely cuebidding/asking for controls

I believe there is not a lot of better methods.

However I have no experience with 321 pts however and not enough experience with denials cue/Zooming to have a definite opinion. An all cuebid aproach (Ken R) is also interesting (but since shape isnt knowned its a different style)

The sad thing IMO is that it would take a simple sim of 100 hands with different methods to see what is clearly best and get some answers for recurrant problems. This is for knowned hands shapes all starting at the same level.

Example of a hand where im lacking a specific tool.

opener
x AQxxx AKxx xxx

xxx KJx Qxx AKQx

1H-1S(relay)
2C (D)---2S (art GF)
3D (1543) - 3H (RKC H)
4H (2+Q+K of D no K of clubs 11-14)

here ill be able to scan for Q/Jacks quite easily and ill be able to find some 6C in 4-3 rather than 6H fairly often, I can make a control ask (partner has showned the K of D and a stiff S so my control ask is for clubs) however i have no way to check if partner got the stiff Spade A wich will often lead to a very bad slam (especially when i dont have the HJ). As a general I can ask about secondary honnors quite low and know wich K my partner has and stop at 4/5 but im lacking info about the location of keycards.

In this specific example since i know that partner has 13 out of a maximum of 14 looking for J isnt that useful but on some other hands he could have A+KQ of H as keycards and there would still have place for 2pts outside wich often make the difference between a good 6 and a poor one.

I think we should do a "challenge sim" we each clearly write our methods and bid some hands at the same "shape is knowned" starting point.

[Valardent]

benlessard, on 2012-March-29, 04:53, said:

My method for unbalanced hand is

1 -asking/showing shape/relay break
2- setting up trumps & keycarding with one bid
3- scanning
4- rarely cuebidding/asking for controls

I believe there is not a lot of better methods.

However I have no experience with 321 pts however and not enough experience with denials cue/Zooming to have a definite opinion. An all cuebid aproach (Ken R) is also interesting (but since shape isnt knowned its a different style)

The sad thing IMO is that it would take a simple sim of 100 hands with different methods to see what is clearly best and get some answers for recurrant problems. This is for knowned hands shapes all starting at the same level.

Example of a hand where im lacking a specific tool.

opener
x AQxxx AKxx xxx

xxx KJx Qxx AKQx

1H-1S(relay)
2C (D)---2S (art GF)
3D (1543) - 3H (RKC H)
4H (2+Q+K of D no K of clubs 11-14)

here ill be able to scan for Q/Jacks quite easily and ill be able to find some 6C in 4-3 rather than 6H fairly often, I can make a control ask (partner has showned the K of D and a stiff S so my control ask is for clubs) however i have no way to check if partner got the stiff Spade A wich will often lead to a very bad slam (especially when i dont have the HJ). As a general I can ask about secondary honnors quite low and know wich K my partner has and stop at 4/5 but im lacking info about the location of keycards.

In this specific example since i know that partner has 13 out of a maximum of 14 looking for J isnt that useful but on some other hands he could have A+KQ of H as keycards and there would still have place for 2pts outside wich often make the difference between a good 6 and a poor one.

I think we should do a "challenge sim" we each clearly write our methods and bid some hands at the same "shape is knowned" starting point.

That would be quite interesting.

Best would be to have a series of specific themes on problems encountered by scanning methods.

Members would post "problem hands" which would then be sorted out so to have 5 to 10 hands for each theme.

Each method would then get a rating for each theme i.e. : 5 out of 7 hands resolved gets a 71% rate and so on...

Of course, it would be difficult to assess that a method is overall superior to another when the average of all the ratings are close (bcs some themes-hands will be less frequent than others), but il will certainly give some unbiased information comparing methods.
The aim of all of this is to get a good tool (a preset of hands in fact) to test and rate a scan method and have several other references to compare with.

So why not make a list of the so called "themes" ?

Posted hands should be as "pure" as possible so to not overlap several themes (I don't know if that's possible)

Some themes propositions :

1) The one benlessard just described in his post about hands with one singleton Ace or King. This theme could be divided into 2 parts; one with Ace singleton, the other with Kingleton.

2) Hands were the relayer is compelled to relay with a void

3) One suited-hands where the solidity of the suit makes all the difference (KQ5432 opposed to KQJ109x)

4) Balanced hands (limited to 4333's and 4432's) facing each-other

Any other participating posters would agree about...

[DinDIP]

benlessard, on 2012-March-29, 04:53, said:

The sad thing IMO is that it would take a simple sim of 100 hands with different methods to see what is clearly best and get some answers for recurrant problems. This is for knowned hands shapes all starting at the same level.

I've tried this testing four different methods of DCB but my conclusion was that it's very difficult for a human to do impartially and to get statistically meaningful results.

Impartiality issues:
I started out by bidding each hand four times but soon realised that it was very hard not to be influenced by information obtained from previous auctions using alternative methods. I tried to overcome this first by changing the order of the methods used (so methods 1 then 2 then 3 then 4 on hand 1; method 2 then 3 then 4 then 1 on hand 2 etc). That still seemed flawed to me so I moved to bidding a number of deals (typically 15-25) using method 1, then the same deals using method 2 etc. However, even this way I was still remembering details of the hands.

Sample issues:
In order to generate statistically significant results the sample size needs to be sufficiently large that reach a reasonable confidence level. 100 deals is just not large enough. I was bidding the deals from Kantar's final (??) book on RKC so there were more than 100 -- but this raised the issue that the deals were not random.

My conclusions:
* It is difficult for one person to do such testing impartially so either two (or more) people need to be involved or it needs to be done by a computer program.
* It is important to design the experiment so that it is not flawed. I think this means (1) having a smaple size that is large enough to make the results statistically significant. It is probably also a good idea to (2) look at smaller samples hands of the kind identified as causing problems to see if there are specific problems that one mthod is good/poor at solving. This may make it possible to structure chain breaks, for example, to show known problem hands. (For example, The Way Forward, a British symmetric system from the 90s [well worth reading, available at http://www.clairebri...ay_forward.doc], allowed a strong club hand to make an early step+1 to show a void in the suit responder had shown length in. Relaying continued, up one step, but responder ignored honours in the void suit when showing strength.) Alternatively, system designers may decide it is necessary to do as some systems do (again, TWF is an example) and include a specific-honour asking bid for handling strong freak hands.

David

[DinDIP]

wclass___, on 2012-March-28, 12:22, said:

I was asking for this part. e.g. If i opened 11-15 1♦ and after shape ask for controls responses would 3-4-5-6, how would it look for QP and in other common situations?

A simple rule that is easy on memory and efficient, at least for hands below 15 HCP, is to multiply HCP by 0.6 and round down to get the number of slam/relay/queen points [i.e. 3-2-1 pointa]. (0.3 is the number to use for controls [2-1 points].) For stronger hands the rule is still helpful but the percentage of hands with the minimum number falls away. What this means is that the SP range for 11-15 hands is 6, 7, 8, 9, 10. Each partnership needs to make a decision whether hands with 11SP (so AAAK) must open 1C or can make a limited opening. (My partnership's rule is that all 11SP hands open 1C.)

One important addition to this is that testing (and ATT experience) shows it is better to treat kingletons (singleton kings) as 1SP and stiff Qs as 0SP.

David

[benlessard]

Quote

I've tried this testing four different methods of DCB but my conclusion was that it's very difficult for a human to do impartially and to get statistically meaningful results.

Judgement in scanning methods is mostly at the end of the sequence before that its mostly mechanics, so even if its not totally impartial its going to have a lot more values than opinions.