Jump to content

GIB could play better?


Recommended Posts

As far as I know, GIB finished the 12th in the par contest 1998, in Lille, France.

That's 8 years ago. Tried a few session of money bridge, but the GIBs I messed with

at BBO are clearly not of this calibre. I have a couple of questions:

 

1. Is the GIB running on BBO server a full-featured one?

2. What is the setting of the sample space? Would it be possible to use larger

sample space setting?

 

Thanks

Link to comment
Share on other sites

It is full featured. How it performs is mostly a function of how long we're willing to give it to think per hand (which in turn affects how long you have to wait for it).

 

How long we give it is a function (at money bridge tables) of whether the host clicks slow/med/fast on robot-speed. In money bridge tourneys, we give it, IIRC, something like 30 or 40 seconds per seat ( human has one, robot has the other 3). The speed of the rented GIBS ( the $1/day, $3/week thing) is a per-user setting and is not related to the two types of server-side gibs (mb. mbt)

 

Gib has some special settings for par contests, we don't use those.

Link to comment
Share on other sites

As I understand, GIB choose what card to play based on the result of monte carlo simulation. I think there's parameter, the size of sample space, setting how many random hands GIB should analysis before making its play.

 

What is the size of the sample space right now? is it possible to set the sample space larger? say greater than 100?

Link to comment
Share on other sites

Without touching src, i believe you can only control the # of hands in card play (not, say, in picking a bid). I use the defaults now; I've seen GIB go belly up once in a while when the # was set "too high", so I'm playing it safe. Additionally, perhaps it is smart enough to use larger numbers when it has more time? not clear.
Link to comment
Share on other sites

The number of deal needed to make a good descision depends on the amount of information available to the GIBs.

e.g. bidding is 1NT - all pass.

So the amount of possible distributions at the first trick is very big. The chance that the sample deals GIB generates differ much form the actual distribution is much greater than those after a more detailed bidding sequence.

So the opening lead and the first few tricks of a GIB declarer/defender can be very bad. With each trick played the possible amount of distribution shriks rapidly, while the number of samples that can be processed (in the given time) increases.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...