search

anntzer / redeal

A reimplementation of Thomas Andrews' Deal in Python.
Other
63 stars 41 forks source link

about opening_lead.py in example #18

Closed spade14 closed 5 years ago

spade14 commented 5 years ago

Hi I think the library is very helpful,thanks a lot.But I would like to confirm about the OpeningLeadSim

Actually I modified the part OpeningLeadSim(Simulation), adding print(self.payoff.table) in the final() part, and run redeal opening_lead.py -n10 to see the IMP difference between leads. I wonder the data in self.payoff.table is the IMP difference between different leads,If Yes,I think the distribution of the IMP difference is kind of weird. there are many 2 and -2 in it, with no +1 -1(usually caused by an overtrick) or +10 -10( the 3NT made or went down).

Would you check if for me? Thank you

anntzer commented 5 years ago

Oops, I got the sign wrong in OpeningLeadSim... I was scoring tricks for the defense, which was always down a bunch and thus scored n*50 (hence the 2 imps difference). That's very embarrassing, and fixed now as of master. Can you have a look?

spade14 commented 5 years ago

I have checked that,the distribution became normal.Thank you. Noticing SQ had a all plus column (I believe that is the best lead?),I think now the result matrix is transposed, which means the column is your lead choice and row is the lead you compare with.

anntzer commented 5 years ago

I fixed the transpose now too :) (I confirmed it with some even more skewed deals where the correct lead is obvious, e.g. AKQxx xxx xxx xx) thanks again.