Richard Willey is looking for participants to provide bidding data for a data mining project. The link is:
https://bridgewinners.com/article/view/help-needed-with-data-mining-project/ and a description of what he is trying to do follows. To me, it sounded like an interesting project, so I have provided my bidding responses. In order to participate, you probably will have to register with Bridge Winners, but registration is free (though they do seek donations in another portion of the web site).
Richard welcomes ALL levels of players to participate.
"I am attempting to train a classifier that can sort bridge players into one of a small number of distinct sets based on how they answer 5-10 bidding questions.
My hope / belief is that players on online bridge sites might be able to use this sort of algorithm to identify other players that they might be compatible with. (Or, if nothing else, people might find this amusing.)
In order to do this, I need a dataset that I can play with. At a high level, I need to determine:
Can sort players into a useful number of sets?
Does set membership remain stable if I make small permutations in the questions being used?
Can I create an accurate classifier that only uses a small subset of the questions?
And, for this, I need a bunch of data…
I am hoping that folks here on Bridge Winners might be willing to lend a hand by answering 50 or so bidding questions and sending me the answers. If you are willing to participate, please answer the bidding problems on the next page. Ideally, I’d like to get 100 or so people to take part.
(If you are willing to participate, could you please reply to this post with your name? It will be useful to see how many arms that I need to twist in order to get enough participants.)
Regretfully, this is an “all or nothing” type situation. The clustering algorithms that I am working with need a complete dataset to be able to process the data. So if you don’t answer all 50 questions, none of your answers count. Note: It’s fine to answer “abstain” to individual questions if you have a serious objection, but please do this as a deliberate protest regarding the bidding decision and not because you don’t have time to think about stuff.
All of these questions were taken from past issues of The Bridge World. When you answer, please presume that you are playing Bridge World Standard. Jeff Rubens and The Bridge World were gracious enough provide permission for these problems to be used and posted here on Bridge Winners. I’m very thankful for their support. If anyone would like to show their appreciation, they can always use new subscribers. (Plus, it’s a good magazine. There’s a reason that I’m using their questions.)
Special thanks to Cam French and David Levin for helping to select hands!
FWIW, this is a completely “open” project, by which I mean I will publish:
The data set
The code that I am using to generate results
The trained classifier
Anyone who is interested in using / adopting the techniques is more than welcome to do so. (My suspicion is that whatever comes from this project might need to be extended. For example, if the ACBL want to adopt something like this for their own online partnership desk, they might want to create their own set of questions based on some other bidding system and train a different classifier that would be more useful for their player base.)
For anyone taking the quiz, please feel free to send me your answers either via a private message here on BW.
All that I need is data in a format like the following:
Q1. 7N
Q2. 3D
The following is very important:
If you are entering a bid, please code it using a number and a letter so:
One Diamond would be listed as "1D"
One No Trump as "3N"
If you are choosing to double or redouble please represent this as X or XX.
If you are choosing to pass, please type "Pass".
If you are choosing to abstain, please type "Abstain".
Thanks!"
