[uresh]

I have started doing some analysis, using a hand generation program and Bo Haglund's double dummy solver, of some common bidding scenarios. My aim is to quantify and advance our collective knowledge in the field. I am looking to answer questions like, for instance, should we use stayman over 1N with a 4-3-5-1 and a 5-count, at match points or at IMPs?

Before I invest too much time in this, I wanted to check if anyone has heard of similar published work? I have seen Bird and Anthias's excellent books on leads based on simulation, but nothing so far on bidding.

I would also like to know if there is interest in this type of analysis. If so, I would welcome questions and scenarios to analyze.

Finally, would the BBO forums be a good place to publish the results? Any other thoughts or ideas?

[Siegmund]

Yes, double dummy analysis can be used to inform bidding decisions. It is easier when you have a fixed system and you want to ask, which is the best bid on a given type of hand. (4-3-5-1 5-count in response to 1NT? Yes, use Stayman. Also 3-4-4-2, 3-3-6-1, and several other shapes you might not expect.) Choosing among several meanings for a bid is hard - if you make one response better maybe you make another worse. Any experiment that requires letting two hands vary at once rather than just one is also hard.


BBO is certainly one place you can make people aware of your findings.

The analyses I have done in the past I posted on my own website. Of course I am very lazy about finding time to write up the results nicely after I run an experiment so I have not posted much in awhile http://taigabridge.net/articles/dd/

[WesleyC]

Welcome to the forums uresh!

I've got experience running simulations - usually to try and shed some light on a hand someone else has posted where a consensus hasn't been reached.

As far as general simulations go, there is quite a lot of stuff scattered around:

Thomas Andrew has done quite a lot including a "ridiculously detailed look at 3-5pt hands opposite a strong 2NT opener".

Richard Pavlicek has a huge amount of fascinating stuff on his website, although his site isn't the easiest to navigate.

You'll also find some more bits and pieces by searching through the BBForum archives.

[uresh]

Thanks for the replies so far, and keep them coming :-)
Siegmund, I am glad I looked at your articles. Looks like we both started with some of the same questions, but you have done a lot more work than I so far.
Many of your conclusions validate my independent work, but I reached a very different conclusion on what to do with balanced 9-counts without a 4-card major over a 15-17 NT.
My analysis, along similar lines as yours, indicates that it is better at all forms of scoring and vulnerability to invite with these hands unless you have a 5332 shape, in which case you should bid game directly.
I will try to write up the analysis and share with you.

[Siegmund]

Quote

Many of your conclusions validate my independent work, but I reached a very different conclusion on what to do with balanced 9-counts without a 4-card major over a 15-17 NT.<br style="color: rgb(28, 40, 55); font-size: 13px; line-height: 19.5px; background-color: rgb(248, 248, 248);">My analysis, along similar lines as yours, indicates that it is better at all forms of scoring and vulnerability to invite with these hands unless you have a 5332 shape, in which case you should bid game directly.<br style="color: rgb(28, 40, 55); font-size: 13px; line-height: 19.5px; background-color: rgb(248, 248, 248);">I will try to write up the analysis and share with you.

I had to go back and look at my article again. With 3343 or 3244 shape it is pretty much a tie, at matchpoints or NV imps, whether to invite or not, if your rule is "pass with 15, go with 16." If you have a more precise way to decide whether to accept an invitation, that will make the invitation more appealing. If you used slightly different restrictions on opener's hand, that can make a difference too. I would not argue either way if someone wanted to invite with 3343 or 3244 hands and 9HCP. (But from a system design standpoint, if I am going to pass all 8s, and bid 3NT on all 10s and the 3352 9s, I think I have better things to do with the 2NT rebid than use it for an invitation.)


Vulnerable at IMPs, I would be surprised if it was right to stop in 2NT with 15 opposite 9. Usually with 24 it is better to be in 3NT than in 2NT (maybe not better to 3NT than 1NT - but once you have invited that isn't an option.)

[etha]

I think the first place to start is hand evaluation.
just try comparing the standard point count and A=6 K=4 q=2 j=1 t=0.5 9= 0.25 normalised to 40 for example.

If this makes a massive difference then you should use it for future sims.

[Siegmund]

Thomas Andrews has some articles about best point counts for suits and for notrump (they are not the same; HCPs are sort of halfway in between.) "Using it for future sims" isn't really an option if our task is to evaluate responses to a standard 15-17 notrump -- though, for instance, opening with HCP and using a specialized point count to decide whether to accept an invitation is.

[helene_t]

Just promoting my analysis of hand evaluation methods for suit contracts: http://www.bridgebas...s-investigated/

For notrump contracts I found something very close to 4321.

[uresh]

I misspoke in my previous post. When vulnerable at IMPs, we should bid game directly with any 9-count (and no major interest). The invite is better at matchpoints or non-vul at IMPs with 4333 or 4432, and at matchpoints even with 5332. The assumptions are that opener accepts with any 16 or 17 count (as you have noted, conservative acceptance fares worse), and RHO does not have a hand suitable for action over 1N. I have attached below the detailed analysis for 4432 shapes. Can share the analysis for the other shapes if needed.

Analysis for 4432 9-counts: Bidding 3N vs Inviting
--------------------------------------------------
The only hands that matter are when opener has a 15-count. On all others, you will be in game either way. (There is a slight benefit of concealing opener's point range when responder bids 3N directly; hard to quantify.)
In a simulation of 10,000 hands, when responder had a 9-count 4432, opener had 4364 times. Of these, the hands made 9 or more tricks 1475 times, 8 tricks 1733 times, 7 tricks 839 times, and 6 or fewer 317 times. At BAM or match-ponits, you gain when the hands make 9 tricks, lose in the other three, for a net loss of 1414 points in BAM, or 0.14 pts per deal. Vulnerable at IMPs, you gain 10 IMPs when making 9+ tricks, but lose 6 IMPs when making 8 tricks and 3 when making fewer than 8, for a net gain of 884 IMPs, or 0.09 IMPs per deal. Non-vul at IMPs, the gain for making game is 6 IMPs, and the losses are 5 IMPs when making 8 tricks and 2 IMPs when making fewer, for a net loss of 2127 IMPs, or 2.1 IMPs per deal.

[Mbodell]

If you find more conclusions I'm always interested in reading. My stuff is extremely not well organized, but I have some simulation stuff at here. I encourage everyone simulating things to always share their code so others can benefit from it, also check your work and your assumptions.

The work I'm most interested, if anyone has a good method coded up, would be combining the simulation scripts with single dummy studies. I have flavors of GIB, BridgeBaron, and Jack but have never gotten around to figuring out how to problematically make the deal code call single dummy solvers.

[Cascade]

Mbodell, on 2015-September-18, 02:26, said:

The work I'm most interested, if anyone has a good method coded up, would be combining the simulation scripts with single dummy studies. I have flavors of GIB, BridgeBaron, and Jack but have never gotten around to figuring out how to problematically make the deal code call single dummy solvers.

I too would be very interested in single dummy simulations.

I have done some single dummy simulations using GIB to analyse opening leads but I did that just by getting GIB to play the hands single dummy rather than using a solver.

I have done many double dummy analyses over the years. One does need to be careful as sometimes double dummy does not give good real bridge results. Nevertheless I feel that my judgement and system has improved based on double dummy results.

Whatever the failings of double dummy, there is a lot of information in analysing more than a life times worth of hands in one particular situation and recognising patterns in how to bid. It is also great when something mildly unusual and perhaps contrary to standard practice or how you previously played comes up at the table and following the indicated method from your simulations leads to a good result.