[Scarabin]

Thanks Barmar,MrAce, and Antrax, for your insights. I guess we have pretty well established that we should not expect a rational approach from a robot. That said I think my peeve with robots is that I want to feel I have played against a human expert, someone who plays like me only better, and not against an idiot savant. I think that is the main point and how well it plays and if it cheats are secondary in importance.

Having said that I still want to verify Ginsberg's claims about how well GIB played the BridgeMaster deals, and to discover if GIB has weaknesses related to certain types of hands: my hypothesis would be these could include combining chances and choosing between alternative plays.

The possible alternative of trying to reproduce and examine how GIB's methods operate is just too daunting and probably a waste of time since I do not want to produce a random simulation.

Thanks again, I will report results.

[Antrax]

I think your comments make a lot of sense. GIB's design goal may have been misguided, IMO. It's designed to play Bridge as well as possible, and I think it probably does a good job at that. But if you're building a robot to play with humans, maybe the perception of playing well, or even not being anti-field by making the same mistakes a human would make, would be more important.

[Cthulhu D]

 barmar, on 2012-February-11, 22:20, said:

Unfortunately, the description language doesn't allow for OR, so we can't say "4+ S OR (3+ S AND (1- H OR 1- D))", which is closer to the actual meaning of 1♣ 1♠ 2♠.

Wow, that's a major issue! Can that be added in as a language extension? It would seem to make the description language much more effective, and if you could specify the percentage for each total clause (so say 95% 4+ S OR 5+ 3 w/side suit singleton) you could much more accuracately inform GIB's simulations.

[barmar]

 Cthulhu D, on 2012-February-22, 22:44, said:

Wow, that's a major issue! Can that be added in as a language extension? It would seem to make the description language much more effective, and if you could specify the percentage for each total clause (so say 95% 4+ S OR 5+ 3 w/side suit singleton) you could much more accuracately inform GIB's simulations.

It's not just the language, it's the way the program represents the information internally. It only has a single HCP range, total point range, suit length range, etc. for each hand.

[Cthulhu D]

Ugh, and I imagine changing the data structures is close to impossible. Intresting oversight. So I'm guessing GIB could never play Swedish club. Interesting

[Scarabin]

 barmar, on 2012-February-22, 23:11, said:

It's not just the language, it's the way the program represents the information internally. It only has a single HCP range, total point range, suit length range, etc. for each hand.

Some programs allow several alternative meanings for each bid, by having the same bid with different priorities, could GIB not be modified to allow this? It would not provide Cthulhu D's percentages but it would provide more flexibility?

[barmar]

 Scarabin, on 2012-September-30, 02:46, said:

Some programs allow several alternative meanings for each bid, by having the same bid with different priorities, could GIB not be modified to allow this? It would not provide Cthulhu D's percentages but it would provide more flexibility?

Anything is possible, if you have enough time and expertise available.

First, we'd have to redesign the way the program represents what a player has shown about their hand, so that multiple possibilities are allowed.

Then, we'd have to change the way this information is used by the program. Currently, a GIB bidding rule is written as something like "if the auction is like X, and partner has shown Y, and I have Z, then this rule applies and it shows A" (not in this type of language, it's actually something closer to regular expressions). If we add multiple meanings to a bid, we have to figure out how to write rules that can deal with this (how do you deal with "might have Y" rather than "promised Y"?). And we'd have to go through hundreds of rules, updating the ones that are affected by this change (the rules are very hard to read, so this is not trivial).

Then we have to figure out how to use it when doing simulations.

I think that if it's doable at all, it would be several months of work -- about the amount of time it took Fred to implement the mobile app. Not that we actually made a concrete decision between the two options, but if we were considering both of them at the same time I think we would have chosen the mobile app. Maybe if GIB were our main product something like this could rise high enough on the priority list, but the way things are, and with the resources we have, the most we're likely to do is small, incremental tweaks, not major redesigns.

[Scarabin]

 barmar, on 2012-September-30, 10:33, said:

Anything is possible, if you have enough time and expertise available.

Elision.

I think that if it's doable at all, it would be several months of work -- about the amount of time it took Fred to implement the mobile app. Not that we actually made a concrete decision between the two options, but if we were considering both of them at the same time I think we would have chosen the mobile app. Maybe if GIB were our main product something like this could rise high enough on the priority list, but the way things are, and with the resources we have, the most we're likely to do is small, incremental tweaks, not major redesigns.

Thanks. Good to hear from you.

Obviously we have completely different aims. You are concerned with real life and have to produce worthwhile solutions, and have to accept performance may be satisfactory rather than perfect.

I, on the other hand can play with aiming at perfect solutions and have to accept these may never be complete.
Part of my fantasy is to improve on every existing program, and that's why I treasure singularities like Oxford bridge allowing for options or Wbridge5 covering Vingte signals.

Slainte,

Ian