[kenrexford]

I don't have a simulator myself, but I'm curious if anyone has even ran this question through.

Assume that Opener starts with a 2NT opening (I assume 20-21, but you could probably analyze this for different ranges) and Responder has a hand with at least five of at least one major, and makes a transfer (presumably).

A proposal was that a Jacoby Transfer be forcing (to make slam sequences easier, etc.). Is this a good idea?

I am curious how often game would not be bid by Responder if given the choice (Responder has a hand where he would pass). I am curious how often this occurs when pener does not have a classic "super acceptance" hand anyway. I am curious how often in this scenario game nonetheless makes, and how often it makes if Opener has 3-card or longer support. (Maybe JT should be forcing IF Opener has 3+ support?)

I am also curious what % of the time when Opener has 3+ support game would not be bid and would not make. (If 0.00004%, I could live with that occasional bad result.) It might also be relevant what % of the time the 3-level contract goes set also.

Any takers?

[Phil]

I agree, I think this would be a worthwhile exercise. There are some partnerships that have this agreement and I've wondered about its effectiveness.

This sim also takes out some of the variability of the effect of lower level auctions, so the results should be rather 'pure', although some might disagree on what a proper superaccept entails.

[dellache]

Difficult to program the superaccept, so I'll just give you the results depending on the number of trumps of the 2NT opener.

DEFINITIONS :
2NT = 20-21.6 using usual evaluation tools (+0.8 for 5xxx, +0.2 for Ace or ten, etc.)
+Includes 5m422 with guarded doubletons
+Includes 5M332
Responder = 0-3 HCP, 5Mxxx, not 5-5 (the "not 5-5" doesn't matter anyway I think)

RESULTS:

________Nogame    4M  3NT   Both-Games Total
Trumps
2       704       30  62    15          811
3       495      145  32    59          731
4       155       65  19    31          270
5        96       64   2    26          188
Total  1450      304 113   131         2000

Number of generated hands : 23451888 (23 millions)

The situation "2NT+responder weak-5M" occurs once every 10000 hand (you need to deal 23 millions hands to generate 2000 cases)

Then I think the above results for THOSE 2000 cases are clear. You can conclude what you like from this DD analysis. Maybe the superaccept represent half of the 4 trumps line.

Regards,

[Jboling]

I did a quick one (fortunately slightly different than dellache), I generated 1000 deals with the following constraints:
north: 20-21 balanced, and 3 card spades (5 card H possible)
south: 5+spades

And the DD-results were:

tricks north south
6    0   1
7   16 17
8   48 52
9  129 140
10 210 209
11 283 279
12 229 221
13  85 81

So if you always superaccept with 3 cards you will end up in a failing 4M contract about 20% of the times. One could compare that with 4+card superaccepts:

tricks north south
6    0   0
7   4    4
8   22 31
9  114 117
10 222 222
11 317 316
12 225 217
13  96 94

That is a failure rate of about 15%. Not that big difference. With 2 card support you have at most nine tricks 33% of the time with spades as trumps. 3NT is actually a slightly better contract, it fails only 32% of the times.

[nigel_k]

You'd need to agree on a maximum hand that would transfer and pass? Or maybe a few different types. Also, are we saying that a really bad hand eg xx xxxxx xxx xxx will pass 2NT or still transfer. If that's a pass then you also need to decide when responder will be 'too good' to pass with a five card major.

Doing this would allow me to build a simulation to calculate the cost of this approach. But calculating the benefit would be harder I think.

[kenrexford]

Yeah -- that's a good problem. Some 5332 hands with utter garbage might be better handled with a pass, perhaps, which could affect the results.

Any thoughts on the right question, then?

I mean, the idea is to have one of three schemes:

a. traditional (or aggressive traditional)
b. always super-accept with 3+ support (with a fit)
c. transfer is forcing

[Phil]

kenrexford, on Nov 11 2009, 03:25 PM, said:

Yeah -- that's a good problem. Some 5332 hands with utter garbage might be better handled with a pass, perhaps, which could affect the results.

Any thoughts on the right question, then?

I mean, the idea is to have one of three schemes:

a. traditional (or aggressive traditional)
b. always super-accept with 3+ support (with a fit)
c. transfer is forcing

I'm more interested in #3, but the sims don't appear to address this.

You need to compare IMPs / MPs gained from better slam bidding versus IMPs lost due to having to play 2N v 3M with drek.

The 2nd part shouldn't be that difficult. The 1st part is pretty subjective.

[dellache]

Honestly, I think this is not worth any complications here. On the boards of the sims, even if you *always* trf, and if you always bid game in any strain (what a silly strategy !), you will be wrong 75% of the time (and I don't know the precise cost of it : I don't know if you would make 2NT or score 3M). Stupid strategy but let's see.

Let's simplify matters by saying it's only a matter of 4M versus "3M always making" or 3NT vs "2NT always making". Let's say we get never doubled. Let's say we are white. Then the **maximal** cost of forcing to game, even with 2-carder is 2.5 imps in average on the 2000 boards of the sim. It's a total of 5000 imps lost every 23 million boards. So this is a maximal average lost of 0.0002 imp per deal played.

In reality, of course, there are garbage hands that you would pass, and so on, and the loss on the garbage hand is probably less than 0.0001 imp per deal played. This is meaningless.

I remember I once spent a lot of time designing a scheme like this on the 2NT using sims :
2NT 3M-1, Then :
3M = bad hand 2 cards, NF
Other = superaccepting 3+ cards.

And finally, the gain was so negligible in theory that we rejected this idea. In fact we do prefer that :
2NT 3M-1, then :
3M = 3+ cards always
Other = Misfit.

The 2NT 3M-1 3M(fit) is actually better in practice, imo. If you only want to be in game (often the case), there is no need to tell anything to the defense (like bidding a Control), and no need to complicate matters by retransfering or whatever.

My 2 pence.

[benlessard]

My interpretation of the result is that super-accepting seems like a dubious propositon and that playing GF transfers is awful.

The fact that 4M score way better than 3Nt clearly demonstrate that a weak hand with a long suit should play in the suit and not in Nt. So playing 3M instead of 2NT is clearly a good strategy even with a dull 5332.

Also facing a strong 2NT i think that 0-3 pts are probably as frequent as slammish hands. So the gain you have by having more slammish sequence is completly annihilated by loss of not being able to play 3M.

[hatchett]

Quote

that playing GF transfers is awful.

Quote

Also facing a strong 2NT i think that 0-3 pts are probably as frequent as slammish hands. So the gain you have by having more slammish sequence is completly annihilated by loss of not being able to play 3M.

Pretty sure this is incorrect. The weaker hands would be a lot rarer than the slam hands. If you play GF transfer you have a big advantage in the slam zone plus the fact your gains are double you losses say 11 IMPs for finding a thin slam versus lose 5-6 for playing the wrong partscore.
Of course a fair amount of the time you make 3M you will also make 2NT.

[Jboling]

I did a new dd-analysis, with in all cases north has 20-21 balanced including 5M332, and the constraints for south are given in the table below.

 
south    : 2NT m : 3S m : 4S m : 3S m, 2NT f: 3S f, 2NT m : 3S f, 2NT f : 2NT m, 3S m
-------------------------------------------------------------------------------------
5S, 0-1p :  137  : 401  : 141  :     303    :     39      :    560      :   98
6S, 0-1p :  182  : 759  : 375  :     587    :     10      :    231      :   172
5S, 2-3p :  449  : 683  : 319  :     327    :     93      :    224      :   356
6S, 2-3p :  457  : 913  : 634  :     474    :     18      :    69       :   439

Here m=makes and f=fails. Not that surprisingly 3S seems on average to be the safest spot in all cases, but more interesting 4S is not that bad contract at least from a imp-point of view, especially with 6 spades. With 6 card spades it seems to be better even from a match-point view to play 4S than 2NT, if 3S is impossible that is. But 3S can be possible if you can stop in 3S if opener has 2 spades. The most problematic case is the one with 5S and 2-3 points, but 2NT is not that bad in that case.

[mikegill]

I once did this simulation because I played that opener always superaccepts with 3+ before. I looked at about 200 hands to see about whether this was a win or not. I found that the (usually very bad) case of opener having exactly 3-card support opposite a responder with 0-2 happens ~2% of the time. Obviously I only saw 200 hands but a few percent is probably about right. Here are my thoughts on this subject:

1) The inability to get out in 3M on these hands is small loss, but definitely nonzero. Obviously if you make your transfers forcing then you can NEVER get out in 3M, which is also certainly a (more nonzero) loss.

2) If you are using the 3-card superaccept to set the suit then
you are losing opener's ability to make judgments about 3NT when he has 3-card support. Does he have to superaccept with xxx or with 4333? If the answer is yes, then you'll be playing the wrong game sometimes (this will always happen over spades - over hearts maybe you can still get to 3N if you use 3♠ as the 3-card superaccept). If no, then can you get back to 4M if opener doesn't superaccept? If yes, then you're taking up a bid to allow opener to get back and you aren't gaining that much from this treatment. If no, then you'll play the wrong game sometimes.

3) This is taking up a bid that could have been used for a more descriptive 4-card supercaccept, and preventing you from finding a 4-4 or maybe 5-4 minor suit fit if you happen to have one too. If you aren't setting the suit with this then, well, you aren't really gaining anything for your slam bidding.

4) Overall, I concluded this was not worth the trouble, nor the increased variance in score. The hands where it gained were very rare indeed, at least with the followup methods I was employing.

[kenrexford]

The theory problem seems perhaps more critical than the safety problem, I agree.

If a transfer is forcing IF Opener has 3-card support, the "middle" approach, then Opener could super-accept by bidding 3NT with the hand that would pass an offer to play 3NT. However, that might then reduce available options to better describe the true super-accepts.

So, maybe the answer is as suggested -- who cares? Establishing a force or a conditional force may not actually gain anything useful.

Of course, the next logical question is whether a conditional force or an unconditional force makes sense if 2♣...2NT (22-23) is the range, or in a Kokish sequence, etc. The theoretical might be affected somewhat here, especially if Responder has already made a positive.