[smerriman]

johnu, on 2024-November-15, 22:31, said:

Matchpoints or IMPs? In my simulations, NT was clearly better at matchpoints, but worse at IMPs.

mycroft's statement was about MPs, so I was using that. But agree re IMPs.

Cyberyeti, on 2024-November-15, 13:23, said:

I would invite slam by whatever method I had, likely 4♣ which has the advantage of denying a spade control which means partner doesn't get in any way excited with Qx, KQxxx, KQx, Jxx but nor do we play in 3N-1

jillybean, on 2024-November-15, 20:37, said:

Puppet provides responder with a lot of information about openers shape and may be the start of a slam probe. Perhaps only hands with some slam aspirations should use puppet here?

Suppose after hearing 3♥ you bypass 3NT, and plan to bid slam if partner has 12-13, a spade control, appropriate keycards (2, or 1+Q), and can somehow do that without ever ending in 5♥. A simulation has you still only scoring 38% vs signing off in 3NT, simply because of all of the times you're in 4♥ losing to 3NT.

If you bypass 3NT, then peek at partner's cards and only play slam precisely when it's makeable, that maximum you can achieve is 58%. I don't think any system will come close to that.

If I have a stronger hand I suspect playing 4NT or 6NT will turn out better than Puppet for similar reasons.

[DavidKok]

johnu, on 2024-November-15, 22:23, said:

If not double dummy odds, then what???

In real life, defenders don't make perfect leads, they sometimes make bad/unfortunate opening leads. They may open up a frozen suit, expose partner's queen in a 2 way guess position, fail to lead a suit setting up a trick which allows declarer time to set up 12 tricks, etc. They may get pseudo squeezed, or discard poorly and expose partner to a squeeze.

Are double dummy odds accurate? As a gross evaluation, I think close enough. If the double dummy odds are 40%, and you need 50% to be a reasonable contract, then a reasonable interpretation is that the contract is not favorable. If the double dummy odds are 60%, then a reasonable interpretation is that the contract is favorable. If the double dummy odds are around 50%, then it's a tossup.

Richard Pavlicek did an analysis about the accuracy of double dummy simulations, and for 6 level contracts, the difference in success rates between actual play results and double dummy simulations was around one and a half percent (double dummy better). So not a big difference.

I am very familiar with those results. This is exactly why I specified that this is a case of balanced-opposite-balanced with traditionally insufficient HCP strength. This is an area where DD outperforms single dummy by a large margin. The 1.5 percentage points (not percent) you cite includes 6-suit contracts and is conditional on a dataset where the slam was actually bid at the table. That screens off most of the hands that are comparable to this situation. For 6NT in particular the advantage is 6.97 percentage points in favour of DD (73.48% single dummy versus 80.45% double dummy, though of course most of these digits arent significant), but since this is still conditional on humans bidding the 6NT it is a gross underestimate once we condition on weaker hands, where the contract hinges on a guess more often. This is also reflected in having a double dummy success percentage of well under 80%.

In general I am a fan of using double dummy simulations to gain better insight into bridge. It is a quick and powerful tool, and pretty much the only way to get sufficiently large sample sizes to do statistical analysis. However, it is important to be cautious in interpreting results. Under the exact conditions of this post I think it is naive to rely on double dummy numbers. You ask me "If not double dummy odds, then what???". I think it is better to not insert incorrect numbers, even if we do not have the correct ones.

[Cyberyeti]

smerriman, on 2024-November-16, 00:43, said:

Suppose after hearing 3♥ you bypass 3NT, and plan to bid slam if partner has 12-13, a spade control, appropriate keycards (2, or 1+Q), and can somehow do that without ever ending in 5♥. A simulation has you still only scoring 38% vs signing off in 3NT, simply because of all of the times you're in 4♥ losing to 3NT.

I disagree with this. Remember you're playing odd methods, the field will be in hearts either having passed or opened 1♥ in a strong NT context with partner's hand.

5♥ is not a disaster if 5♥ makes 11 and 3N makes 10 (or it's 12:11). The other issue is that there may be a finesse which you can take trivially in hearts, but opps have a lot of spades or diamonds to cash if it's wrong and you take it in NT.

[Zelandakh]

johnu, on 2024-November-15, 03:12, said:

The problem is whether you can get opener to accept slam with 13 HCP, and decline with 12 HCP.

Using 2♠ as a range ask does offer this possibility. The auction 1NT - 2♠; 3♣ - 4NT is typically played as quantitative asking if Opener has a max max.

[smerriman]

Cyberyeti, on 2024-November-16, 03:46, said:

I disagree with this. Remember you're playing odd methods, the field will be in hearts either having passed or opened 1♥ in a strong NT context with partner's hand.

You're right; I was directly comparing hearts to 3NT rather than considering other tables. But if we're expecting all of the other tables to be in hearts, doesn't that make 3NT more appealing, if there are good odds we're beating everyone in 4♥ even if slam happens to make?

Agreed there are reasons why double dummy saying NT is better than hearts doesn't necessary make it so, but as per the usual discussion that generally averages out.

[Cyberyeti]

smerriman, on 2024-November-16, 14:48, said:

You're right; I was directly comparing hearts to 3NT rather than considering other tables. But if we're expecting all of the other tables to be in hearts, doesn't that make 3NT more appealing, if there are good odds we're beating everyone in 4♥ even if slam happens to make?

Agreed there are reasons why double dummy saying NT is better than hearts doesn't necessary make it so, but as per the usual discussion that generally averages out.

This hand has one of THE holdings that works in favour of NT double dummy, although that reason partially goes away if hearts are going to be trumps unless partner has Kxxxx.

Our auction on this hand would be interesting, P-1♣(4+)-1♠-1N(15-bad 19)-2♣(asking)-2N(17-19 3334) now I suspect 3N hits the table.

[mw64ahw]

Taking another look at this and applying our approach over a strong NT we would have the following responses which could be developed further to find a slam.
1N-2♣ non-promissory 5cM ask?
.. 3♣ exactly 2335 max.
.. 3♥ 5♥ max.
.. 3♠ 2344 w/o ♠ stopper max.
.. 3N 2344 max. with ♠ stopper max.
1N-2♣
2♦-3♣ SI in a minor

[mycroft]

 jillybean, on 2024-November-15, 20:37, said:

Puppet provides responder with a lot of information about openers shape and may be the start of a slam probe. Perhaps only hands with some slam aspirations should use puppet here?

Okay, maybe I'm wrong about 4♠ vs 3NT - but as I have discussed before, after increasing the variance by playing off-field systems, not having a way to "get back to field", whether it's right or not on this hand, increases it more. If you're okay with that, you're okay with that.

But I remember a post on the Other Site that basically said "If you open 5cM in NT regularly, you can play a version of Puppet that doesn't give away 'opener does not have a 4cM' when finding out that partner doesn't have a 5cM; and you can use it routinely, giving away near-zero information to the opponents and finding many of your 5-3 major fits." And it seems right to me.

If there's a better meaning for 3♣ that you currently use, please keep using it ("I don't care what we play 3-level bids as - they never come up. But if we don't discuss them, one will.") If you don't - might still be a good idea.

Maybe not on this hand - I guess, yes, that with this hand, you are only likely to lose tricks in trumps and rarely will play better in spades than NT. But how many 18s do you get, as opposed to the 14s and 15s?

[Zelandakh]

 mycroft, on 2024-November-17, 00:21, said:

"If you open 5cM in NT regularly, you can play a version of Puppet that doesn't give away 'opener does not have a 4cM' when finding out that partner doesn't have a 5cM; and you can use it routinely, giving away near-zero information to the opponents and finding many of your 5-3 major fits." And it seems right to me.

If there's a better meaning for 3♣

Traditional Puppet, 1NT - 2♣; 2D = "no 5M", does not give away whether Opener has a 4 card major if Responder was only interested in 5 and moreover does not give away whether Opener has 4 cards in the other major when Responder has one 4-card major. It has other disadvantages but it certainly covers the bases you are mentioning here.

[jillybean]

 mycroft, on 2024-November-17, 00:21, said:

Okay, maybe I'm wrong about 4♠ vs 3NT - but as I have discussed before, after increasing the variance by playing off-field systems, not having a way to "get back to field", whether it's right or not on this hand, increases it more. If you're okay with that, you're okay with that.

But I remember a post on the Other Site that basically said "If you open 5cM in NT regularly, you can play a version of Puppet that doesn't give away 'opener does not have a 4cM' when finding out that partner doesn't have a 5cM; and you can use it routinely, giving away near-zero information to the opponents and finding many of your 5-3 major fits." And it seems right to me.

If there's a better meaning for 3♣ that you currently use, please keep using it ("I don't care what we play 3-level bids as - they never come up. But if we don't discuss them, one will.") If you don't - might still be a good idea.

Maybe not on this hand - I guess, yes, that with this hand, you are only likely to lose tricks in trumps and rarely will play better in spades than NT. But how many 18s do you get, as opposed to the 14s and 15s?

I may be wrong but wasn't this version of Puppet that conceals openers shape devised by Justin Lall? I was playing this back when I was playing somewhat seriously and I think it was a treatment I picked up here. I will be playing it again, twice a month.

As for variance and off-field systems, I have no real desire to "play the field" to protect my results. While winning once in a while is great I enjoy trying different systems. I add complexities to my card and keep those that work and I can remember. My favorite is 1m (1H) X denying 4 spades.
For a long time, even before I switched to weak/mimi nt, my card has has a line through three level bid responses to 1nt (other than 3♣, puppet)

[johnu]

 DavidKok, on 2024-November-16, 02:05, said:

You ask me "If not double dummy odds, then what???". I think it is better to not insert incorrect numbers, even if we do not have the correct ones.

So if you don't have any numbers, what do you propose to do?

 DavidKok, on 2024-November-16, 02:05, said:

For 6NT in particular the advantage is 6.97 percentage points in favour of DD (73.48% single dummy versus 80.45% double dummy, though of course most of these digits arent significant), but since this is still conditional on humans bidding the 6NT it is a gross underestimate once we condition on weaker hands, where the contract hinges on a guess more often. This is also reflected in having a double dummy success percentage of well under 80%.

Talking about inserting incorrect numbers, your numbers for 6NT are based on results after the opening lead, while my numbers are based on results before the opening lead. The reason your 6NT double dummy result is much better than the 75.76% Pavlicek DD result is that the real life opening leads lose about 5% compared to a double dummy lead. Which is exactly the point I was making that defensive "mistakes" even the odds.

For 6NT specifically, Pavlicek shows a difference of 2.4% between actual and DD results. Certainly a small enough amount that you can make general assumptions based on DD results. But even a 5+% difference is better than nothing. If you want to get pseduo-scientific, you can subtract a percent or 2 from the double dummy results to better approximate what real life results might be.

[DavidKok]

Thank you for the correction, I quoted the wrong table.