[mrdct]

keylime, on Jul 17 2005, 06:32 PM, said:

With regards to the CPU issue, agreed with Dave - unless you have it basically cranked at maximum it becomes a drag to get done.

I assume that when you set GIB to play faster, it will generally play worse as it hasn't had time to fully calculate all of the possible eventualities.

Is all the GIB processing done centrally on the BBO server? Maybe some sort of distributed computing model could be used to share the GIB processing amongst the machines of the human players at the table.

I would like to be able to get four GIB players playing against eachother. Even better would be to run five or six tables of GIB players against eachother at varying speeds to produce a datum.

I also assume that GIB has no way telling what system its opponents are playing which kind of makes it a bit unfair. Perhaps when a GIB player is at the table results will only be meaningful if everyone is playing SAYC.

[fred]

mrdct, on Jul 18 2005, 03:19 AM, said:

keylime, on Jul 17 2005, 06:32 PM, said:

With regards to the CPU issue, agreed with Dave - unless you have it basically cranked at maximum it becomes a drag to get done.

I assume that when you set GIB to play faster, it will generally play worse as it hasn't had time to fully calculate all of the possible eventualities.

Is all the GIB processing done centrally on the BBO server? Maybe some sort of distributed computing model could be used to share the GIB processing amongst the machines of the human players at the table.

I would like to be able to get four GIB players playing against eachother. Even better would be to run five or six tables of GIB players against eachother at varying speeds to produce a datum.

I also assume that GIB has no way telling what system its opponents are playing which kind of makes it a bit unfair. Perhaps when a GIB player is at the table results will only be meaningful if everyone is playing SAYC.

You are correct. The faster GIB plays, the worse it plays.

GIB currently runs on your PC - not on our server.

A distributed model (ie having GIB use the resources of several computers to help it solve the same problem) would likely require a massive rewrite of GIB itself. Though Matt Ginsberg might well find this to be an interesting challenge, I seriously doubt that he will have the time to do this during the next couple of years. I also doubt that there is another person on planet earth who would have a chance to successfully complete such a project.

However, it is not at all unlikely that we will have a server-based version of GIB running by the end of the summer. In my opinion, this facility will have a few useful applications, but the sort of things that you can currently do with GIB and BBO will likely continue to work via a client-based version of GIB. Uday is more hopeful than I am about the usefulness of a server-based GIB.

You can have 4 GIBs playing against each other. Start a total points, rubber, or teaching table and you can do just that.

GIB knows the basic structure of various bidding systems as well as dozens of conventions. It also has facilities whereby the user explain his bids to his GIB opponents. I have not had a chance to make an interface between BBO and these aspects of GIB yet, but this will be near the top of my list of things to do when I get back from the ACBL Nationals in early August.

While I am away, if you have any more questions about GIB, you might try asking Stephen Pickett (sfbp on BBO). Stephen has consider experience writing software that interfaces with GIB. He was very generous with his advice to me when I was getting started on the BBO-GIB programming (thanks!).

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[Deanrover]

Has this software re-write made it easier to incorporate other bots into BBO? The increased competition would be useful on many levels.

[barmar]

fred, on Jul 17 2005, 01:10 PM, said:

Another interesting possibility involves tables with 2 humans, each playing with a GIB as their partner. If we disallowed kibitzers at such a table it would be nearly impossible to cheat. We might even let people play rubber bridge for real money under these cicumstances.

I think this would only be true if you implement server-based GIB. As long as GIB is running on the same computer as yourself, it would be possible to hack into the communication and see its cards.

[fred]

Deanrover, on Jul 18 2005, 04:43 AM, said:

Has this software re-write made it easier to incorporate other bots into BBO? The increased competition would be useful on many levels.

The robot interface that I created is not GIB-specific. It would not take much work to make it possible for other bridge-playing engines to work with BBO.

No way we are going to get involved in this until at least we have had time to evaluate the potential of GIB on BBO (both in terms of how much the membership uses it and how much revenue it generates).

Besides that, Bridge Base Inc, has a strong business connection with GIB and you might say that I have an emotional attachment with this program (I have been a friend of Matt Ginsberg for years and I have watched GIB "grow up").

Even if GIB is the only robot we offer, I think our members will be well-served. There is only one other computer bridge program that comes close to GIB in terms of its quality of play (a Dutch program named Jack).

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[scoob]

someone else earlier in this thread - i forget who said:

this is awesome

[DenisO]

fred, on Jul 18 2005, 11:19 PM, said:

Even if GIB is the only robot we offer, I think our members will be well-served. There is only one other computer bridge program that comes close to GIB in terms of its quality of play (a Dutch program named Jack).

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

I think you are being a fit unfair to Jack here. GIB is still a great program but it's been two years since any real development was done. In that time Jack has been continuously improved and most people who have both programs, including long-time GIB fans from the GIB discuss group, now consider Jack to be the better program; and it has been World Computer champion for 5 years in a row, which includes a period when GIB was active.

It would be marvellous if this initiative from BBO can generate sufficient interest/revenue to persuade Matt to start work again on GIB. GIB seems to be the only program with the potential to knock Jack off the No.1 spot.

Denis

[DenisO]

fred, on Jul 17 2005, 11:16 PM, said:

During today's vugraph Walddk would occasionally ask GIB what it would have bid or played. I personally found this to be interesting. He was careful to emphasize the difference between GIB's analysis (which is single dummy) and Deep Finesse's analysis (which is double dummy).

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

I like this feature but it only seems to work during live vugraph. When I open an existing lin file on my PC there is no GIB button even though I have GIB installed. In contrast the DF button will work - I have DF installed as well but I'm not sure if that's why it works.

Denis

[Gerben42]

1) Agree with Denis0. Jack is really the strongest program around nowadays, I think few people doubt that.

2) I really really like the new GIB feature I must say! Totally wicked

I think it is the right way that GIB uses the user's computer to calculate his move, this will be much faster than have some central computer deal with requests. Running some server based GIBs is possible of course, for some special situations, but running several will be detrimental for the playing strength.

[Pooh]

GIB SYSTEM DETAILS

Is there a Schedule of the System which GIB uses in BBO Robot Mode?

I discovered the other day that we were playing against Inverted Minors

It seems to be some form of SAYC ... but "tweaked".

Thanks

Pooh

[uday]

When FG returns (he is at a tourney) he will address the gib cc issue.

[Brandal]

uday, on Jul 25 2005, 07:39 AM, said:

When FG returns (he is at a tourney) he will address the gib cc issue.

Maybe I should meantion this here too:

I played in main with a human partner and
opps were two very fine GIB's.

I alerted my artificial bids and explained,my
LHO GIB alerted his Pass,happened every time.

They explained in full if I clicked on their bids,
all sorts of possible explanations

[slothy]

I was so taken by the bridge skills of GIB E - GIB W is arrogant! - that i am seriously considering asking her on a virtual date....


....i got her on web-cam and her mother-board is awesome!!

[rayos]

Resistance is futile

All your base belong to us!

[EarlPurple]

When will I be allowed to partner GIB in a tourney?

When can we have an "indy" tourney where every pair contains a human and a GIB (except maybe one pair of 2 GIBs)

How about a regular indy where GIB enters, so we see how well he does?

[Limey_p]

I smell a "Turing test" coming up. If chat is disabled, can we detect a GIB as pard or op?

*This* may be worth a trial from an ambitious tourney operator!

AP

[Deanrover]

Gib would fail the Turing Test due to her bizarre bids. Jack passes it as it plays sayc/2o1 type bidding systems tonnes better than Gib

[helene_t]

Even Jack fails the Turing test, IMHO. The kind of mistakes he makes are characteristic: never micounts, but is easily trapped by psyches, even by elementary falsearding.

[asdfg2k]

Has anybody a list of the system/conventions that GIB plays?

[Gerben42]

Quote

Even Jack fails the Turing test, IMHO. The kind of mistakes he makes are characteristic: never micounts, but is easily trapped by psyches, even by elementary falsearding.

So are some club players. If you find examples of this, perhaps report these to the Jack team so they can work on it.