[biggerclub]

Is there software where I can enter rules for bidding (i.e., program my own system), and it will play/practice with me?

[Mbodell]

Yes, but it means quite a bit more programming than you might prefer (I.e., not just simple cc fill out). There is dealing functionality built in to bbo where you can control the dealer. There are also software that deal hands (like deal and dealer) so you can prepare your rules and create hands. You can then have robots like GIB or Jack or Bridge Barron bid the other hands, or you can write into your bridge programs which hands bid what.

But for anything past the most simple, it isn't easy to do.

For playing with a partner, partnership bidding on BBO is pretty good. You can make easy adjustments like your side has 23+ points or something so you get good hands to bid. You can also import deals (either that you've programmed in a different program, or that someone else has made for you). I.e., maybe you are practicing your system over 1m-1M-2nt so someone makes a bunch of hands where 95% or 100% of them start like that and then let you practice your continuations.

[helene_t]

Jack has an impressive menu of conventions to chose from and you can also chose things like preempt style. But you can't make your own system.

According to this review (from 2013) there are two programs that allow you to create your own conventions: Blue Chip and Bridge Buff. Both are not very good at it though, as you can read in the review.

[hrothgar]

biggerclub, on 2015-July-21, 19:35, said:

Is there software where I can enter rules for bidding (i.e., program my own system), and it will play/practice with me?

http://www.eclipse.o...lopers/lunasr1a

[blackshoe]

There used to be a program called, iirc, "Bridge Partner" that used AI techniques to "learn" whatever system you might teach it. But the learning process took a long time. The program was around for a couple of years, but I haven't heard anything about it recently.

Edit: it was called "Perfect Partner Bridge".

This post has been edited by blackshoe: 2015-July-22, 12:12

[biggerclub]

helene_t, on 2015-July-22, 05:18, said:

Jack has an impressive menu of conventions to chose from and you can also chose things like preempt style. But you can't make your own system.

According to this review (from 2013) there are two programs that allow you to create your own conventions: Blue Chip and Bridge Buff. Both are not very good at it though, as you can read in the review.

Thanks for the suggestions. Bridge Buff looks like it won't work as I am Windows 8 now. Blue Chip does not seem to fit the purpose either. It was a shot in the dark at best.

/s/ biggerclub, aka the mad scientist of bidding

[biggerclub]

hrothgar, on 2015-July-22, 06:49, said:

http://www.eclipse.o...lopers/lunasr1a

Would probably work but way over my head at the present time.

[biggerclub]

blackshoe, on 2015-July-22, 12:07, said:

There used to be a program called, iirc, "Bridge Partner" that used AI techniques to "learn" whatever system you might teach it. But the learning process took a long time. The program was around for a couple of years, but I haven't heard anything about it recently.

Edit: it was called "Perfect Partner Bridge".

Thank you. I don't think that this is quite what I was thinking either.

[biggerclub]

Mbodell, on 2015-July-22, 01:14, said:

. . . or you can write into your bridge programs which hands bid what. . . .

This is what I am asking . . . how or in what basic (robot) bidding program can I have this flexibility? And how difficult is it to hack existing bidding engines?

(For example: 1C = 17+ HCP any distribution (i.e., all 17+ HCP hands) or 11-16 HCP with 5+ Clubs and Clubs are the longest suit; etc.)

[Mbodell]

biggerclub, on 2015-July-22, 16:48, said:

This is what I am asking . . . how or in what basic (robot) bidding program can I have this flexibility? And how difficult is it to hack existing bidding engines?

(For example: 1C = 17+ HCP any distribution (i.e., all 17+ HCP hands) or 11-16 HCP with 5+ Clubs and Clubs are the longest suit; etc.)

Look at deal. You can see a sample hand and in the comments he gives the auction he's simulating and then the code throws out deals that don't match. The sample will just print out the 4 hands, but you can control what gets printed by the commands and instead of printing the actual hands, print the auction. Then if you wanted to practice have one side (south) print the hand (for you to bid) and the other side print the bids it makes in response. I don't claim the code is super great, but you can see the sorts of simulations I do (usually on leads given auctions, sometimes on contracts given auctions) by browsing.

And again if you use a program like this to deal hands (like the sample hand program given, you can collect those hands in formats where you can then bid them in partnership bidding or at a BBO table with robots. So you can practice the hands that exercise your system - assuming you have a partner or can get the Robots to use your system.

[helene_t]

Mbodell and biggerclub, it looks like you are talking about two different things.

Mbodell talks about software that can generate hands that satisfy a particular criterion, which could for example be "16+ points".

Biggerclub asks for software that can selectthe correct bid for a given hand, given some rules which he has programmed himself.

[Mbodell]

helene_t, on 2015-July-23, 03:57, said:

Mbodell and biggerclub, it looks like you are talking about two different things.

Mbodell talks about software that can generate hands that satisfy a particular criterion, which could for example be "16+ points".

Biggerclub asks for software that can selectthe correct bid for a given hand, given some rules which he has programmed himself.

Right, but if you give the rules to select the program, you can then get the hands that fit them and feed them to something else (including in the feed the bid that you think they fit). If you do that last part, then you have what OP is talking about.

[barmar]

Programming the rules for bidding is incredibly complicated. The part of GIB that implements its bidding rules is over 6,000 lines in an incredibly compact language (not including over 2500 lines of comments explaining what's going on in the code lines).

[Cascade]

barmar, on 2015-July-23, 12:14, said:

Programming the rules for bidding is incredibly complicated. The part of GIB that implements its bidding rules is over 6,000 lines in an incredibly compact language (not including over 2500 lines of comments explaining what's going on in the code lines).

Is there any description of GIBs rules?

[biggerclub]

barmar, on 2015-July-23, 12:14, said:

Programming the rules for bidding is incredibly complicated. The part of GIB that implements its bidding rules is over 6,000 lines in an incredibly compact language (not including over 2500 lines of comments explaining what's going on in the code lines).

It is generally agreed that GIB is horrible. The best there is, but still horrible. I am not trying to make a super bridge bidding robot. Just one that can be tuned to my system. I would love to have the GIB code (6000 lines doesn't sound all that daunting to me . . . ) and just modify/massage that. I assume however, that GIB is proprietary. Maybe not. Could I have it?

[blackshoe]

Doubtful. Very doubtful. Go here.

[barmar]

Just to give an example of what it looks like, here's the code that tells GIB when to pass partner's natural 3NT or quantitative 4NT bids:

; NT game
0 ..  [3-4]N:P:$
1 73d  P ~UFIT<H~P~.+MVAL(PARD,HCP)<=NVALUES(6)-1~&&:BAL :P~NVALUES(6)-1-MVAL(PARD,HCP)~:SBAL
1 52d  P ~UFIT<H~P~.+MVAL(PARD,HCP)<=NVALUES(6)-1~&&:SEMINTYPE :P~NVALUES(6)-1-VAL(PARD,HCP)~:SSEMINTYPE
1 73d  P ~UFIT<H~P~.+VAL(PARD,HCP)<=NVALUES(6)-3~&&:BAL :P~NVALUES(6)-3-VAL(PARD,HCP)~:SBAL
1 52d  P ~UFIT<H~P~.+VAL(PARD,HCP)<=NVALUES(6)-3~&&:SEMINTYPE :P~NVALUES(6)-3-VAL(PARD,HCP)~:SSEMINTYPE
1 23d   P ~UFIT<0~P~.+VAL(PARD,HCP)<=NVALUES(6)-1~ :P~NVALUES(6)-1-VAL(PARD,HCP)~
1 39d  P ~UFIT<H~P~.+VAL(PARD,HCP)<=NVALUES(6)-3~&&:SEMIBAL .
1 27d  P ~UFIT<0~P~.+VAL(PARD,HCP)<=NVALUES(6)-3~ :P~NVALUES(6)-3-VAL(PARD,HCP)~
1 27d  P ~UFIT<H~:M~.+VAL(PARD,LP)<=SVALUES(6)-3~ :m~NVALUES(6)-3-VAL(PARD,LP)~
; with an M fit, the book bid will be to pass 3N with 4-3-3-3 and correct otherwise
; presumably simulations will often overturn this
1 78d  P P~.+MVAL(PARD,HCP)<=NVALUES(6)-1~&&:D......3 :P~NVALUES(6)-1-MVAL(PARD,HCP)~
1 78d  P P~.+VAL(PARD,HCP)<=NVALUES(6)-3~&&:D......3 :P~NVALUES(6)-3-VAL(PARD,HCP)~
;antistupidity prevention of runout 3NT with none-major fit
1 52Xd P NOSLAMPOSS&&NOFREAK2_GP~UFIT<H&&(HOLD(b)+VAL(PARD,b)<9)~ .

[biggerclub]

barmar, on 2015-July-24, 09:09, said:

Just to give an example of what it looks like, here's the code that tells GIB when to pass partner's natural 3NT or quantitative 4NT bids:

; NT game
0 ..  [3-4]N:P:$
1 73d  P ~UFIT<H~P~.+MVAL(PARD,HCP)<=NVALUES(6)-1~&&:BAL :P~NVALUES(6)-1-MVAL(PARD,HCP)~:SBAL
1 52d  P ~UFIT<H~P~.+MVAL(PARD,HCP)<=NVALUES(6)-1~&&:SEMINTYPE :P~NVALUES(6)-1-VAL(PARD,HCP)~:SSEMINTYPE
1 73d  P ~UFIT<H~P~.+VAL(PARD,HCP)<=NVALUES(6)-3~&&:BAL :P~NVALUES(6)-3-VAL(PARD,HCP)~:SBAL
1 52d  P ~UFIT<H~P~.+VAL(PARD,HCP)<=NVALUES(6)-3~&&:SEMINTYPE :P~NVALUES(6)-3-VAL(PARD,HCP)~:SSEMINTYPE
1 23d   P ~UFIT<0~P~.+VAL(PARD,HCP)<=NVALUES(6)-1~ :P~NVALUES(6)-1-VAL(PARD,HCP)~
1 39d  P ~UFIT<H~P~.+VAL(PARD,HCP)<=NVALUES(6)-3~&&:SEMIBAL .
1 27d  P ~UFIT<0~P~.+VAL(PARD,HCP)<=NVALUES(6)-3~ :P~NVALUES(6)-3-VAL(PARD,HCP)~
1 27d  P ~UFIT<H~:M~.+VAL(PARD,LP)<=SVALUES(6)-3~ :m~NVALUES(6)-3-VAL(PARD,LP)~
; with an M fit, the book bid will be to pass 3N with 4-3-3-3 and correct otherwise
; presumably simulations will often overturn this
1 78d  P P~.+MVAL(PARD,HCP)<=NVALUES(6)-1~&&:D......3 :P~NVALUES(6)-1-MVAL(PARD,HCP)~
1 78d  P P~.+VAL(PARD,HCP)<=NVALUES(6)-3~&&:D......3 :P~NVALUES(6)-3-VAL(PARD,HCP)~
;antistupidity prevention of runout 3NT with none-major fit
1 52Xd P NOSLAMPOSS&&NOFREAK2_GP~UFIT<H&&(HOLD(b)+VAL(PARD,b)<9)~ .

So . . . while I would need a glossary to interpret some of the values/terms, it does not seem to be above my bridge or programming abilities. You may be right that I might lose interest as I delve into the complexities of the abyss . . . .

[frisbee]

Thanks for the programming slice, barmar. Now I better understand why GIB is so poor in spots. 6000+ lines of code is small, at least in a compiled language. I do not want to suggest re-starting from scratch (which is usually the most efficient method to improve a program that has been under continuing development for years), but at least try to convert to a more useful programming language.

[barmar]

frisbee, on 2015-August-12, 13:25, said:

Thanks for the programming slice, barmar. Now I better understand why GIB is so poor in spots. 6000+ lines of code is small, at least in a compiled language. I do not want to suggest re-starting from scratch (which is usually the most efficient method to improve a program that has been under continuing development for years), but at least try to convert to a more useful programming language.

6,000 lines of code in a language like C is not very much. But each rule in the GIB bidding system corresponds to something like a dozen or more lines of C code.