Simulations

[Rossoneri]

Apologies if someone has asked this before or there is info in some other post, I couldn't really find any good info on this.

 

I've seen simulations being talked about in quite a number of posts. So my question is, how do you guys usually set one up, what programs do you use etc.

 

Wanted to test a specific hand, but so far I haven't found any good ways of doing so. Tried using Richard Pavlicek's Deal Finder, but problem is I still have to sift through the double dummy analysis myself....

 

Thanks.

[Stephen Tu]

I've been using dealer by van Stavaren et al, which hooks into GIB, a commercial bridge playing program that will also spit out DD analysis.

 

Bo Haglund recently released a freeware deal simulator based on his double dummy DLL & Thomas Andrew's Deal program. I think if starting I'd probably use this one since it doesn't require a GIB copy.

[helene_t]

I use R for most purposes. This is a general-purpose statistical package so is serves a different purpose than the bridge-specific software. For making deals with specific properties I use Deal which is fast, flexible and platform-independent. For double-dummy analysis I don't have any software so I query the GIB DD database.

[mikeh]

I use Dealmaster Pro: a program I've had for quite a few years. However, I wouldn't use any software for double-dummy analysis (unless constructing a problem) because I don't analyze bidding or lines of play on a double-dummy basis. Double-dummy analysis will mislead you if you use it to establish the effectiveness of bidding or of lines of play.....

[helene_t]

Bo Haglund recently released a freeware deal simulator based on his double dummy DLL & Thomas Andrew's Deal program.  I think if starting I'd probably use this one since it doesn't require a GIB copy.

Great, that was what I was looking for!

 

It is available for Mac and Windows. I just managed to get it running on Linux. You need to do the following:

 

- Make sure that python-dev and python-distutils-extra are installed

- Qualify the paths for the header files referred in pydds.c

- Run python setup.py install

 

However, I wouldn't use any software for double-dummy analysis (unless constructing a problem) because I don't analyze bidding or lines of play on a double-dummy basis.

But if you're just aware of that limitation, DD analysis is still better than nothing, isn't it? You could still use it for scanning for situations in which a particular bidding treatment leads to absurd results, for example.

 

I once saw an article in Bridge World about when to bid 3N in spite of a 5-3 fit in a major, based on a single dummy analysis (using GIB). In that article, influence from the bidding on the opening lead and defense was not accounted for, according to the author (I don't quite understand that, GIB will always use inference from the auction when defending, won't it?). Anyway I suppose that could be taken into account as well.

[Rossoneri]

I use R for most purposes. This is a general-purpose statistical package so is serves a different purpose than the bridge-specific software. For making deals with specific properties I use Deal which is fast, flexible and platform-independent. For double-dummy analysis I don't have any software so I query the GIB DD database.

How do you use R to do that? (Maybe I should learn since I am starting university next year!)

 

mikeh: I agree that DD doesn't take in a lot of factors...but just wanted to get a general idea of where the deal is heading...

[ArcLight]

I use Deal Master Pro available from Ed Marzo. www.dealmaster.com

 

It does the following

 

A] You can enter one or more exact hands.

You can specify up to 20 "screens" of constraints of rthe other hands such as

 

North:

1. Not 15-17 balanced

AND

2. not 20-21 Balanced

AND

3. Less than 22 HCP + Distrubutional points* (you can define the distributional points by suit, using different valuations for each suit)

AND

4. at least 13 HCP+Distributional points

 

South

1. Balanced

AND

2. 8-10 HCP

AND

3. No 4 card major

 

 

b] You enter the # of simulations

You can inspect each one and reject any if you like

 

3] Double Dummy Analyzer - Deep Finesse (this is included for free)

Enter the contrscat you are interested in, such as NT by S, or 2♠♥♦♣ by East (whatever)

It will tell you the results for each contract

Ex:

1NT S -2 5%, -1 10%, making 43%, +1 33% +2 9%, as well as the opponents suit contract

 

You can inspect those deals that interest you.

This lets you look at WHY did 6♠ go down? Aha, the long hearts were not very useful.

 

 

c] It also tells you the best lead against each contract

 

Plus a lot of other stuff.

Its VERY easy to use

 

Its not as powerful as a programming language like simulator like Deal 3.0 but its easier to use.

 

Making Marty Bergens Rule of 20 is rather complex, and is probably easier in Deal 3.0.

 

The owner (Ed Marzo) is a very pleasant person, and I'm quite pelased with the product.

[Mbodell]

I use Deal 3.0 and if you have a programming background it is very easy to use. I didn't know tcl before using deal but it was easy for me to pick up (CS major who programs in C/C++ for a living, YMMV). I've been able to simulate a whole bunch of different situations. I like that I can simulate a large number of hands, even with very specific programmatic restrictions, and like that I can count things like "how often does this line work versus this line" by inspecting all the hands on each deal.

[helene_t]

I use R for most purposes. This is a general-purpose statistical package so is serves a different purpose than the bridge-specific software. For making deals with specific properties I use Deal which is fast, flexible and platform-independent. For double-dummy analysis I don't have any software so I query the GIB DD database.

How do you use R to do that? (Maybe I should learn since I am starting university next year!)

 

mikeh: I agree that DD doesn't take in a lot of factors...but just wanted to get a general idea of where the deal is heading...

You can get the script I used to fit the Bermuda Bowl Round Robin IMP results here. The assumptions I made have been discussed in this thread.

 

To simulate from a linear model of the form mymodel=lm(y~x):

 

simulated.y = rnorm(length(x),mymodel$coefficients,sd(mymodel$residuals)

 

This may not look very impressive but the cool thing is that the R user community has implemented just about all statistical models that repeatedly come up in scientific litterature, and the R software itself has a wide range of options for validating the models and producing graphical output.

 

If you have access to MATLAB software at the university it will probably be better suited for bridge-related (and other) simulation tasks.

[Rossoneri]

I happen to have Matlab.