GiB suggestions

[sflicht]

Hi,

 

1) Can the GiB algorithm be open-sourced? I don't know whether BBO has purchased the exclusive right to use and distribute the software from Ginsburg (or whoever). If so, it seems to me that it doesn't lose any value as a money-making tool for BBO if the algorithms were made public. After all, BBO will always have control over what robots are allowed to run on their servers.

 

2) A few thoughts on GiB's bidding system:

 

2a) Don't jump in response to a takeout X without extra length in the jump suit. Instead, make a forcing bid such as a cuebid. This frees up the takeout X in many scenarios to be slightly more flexible with regard to major suit length than the FD specification of the bid requires.

 

2b) Right now GiB has a somewhat unfortunate habit of making natural jumps to the 5 level in a major suit fit as a generic slam try. This is unscientific in the extreme, especially since it's rather unclear what GiB "wants" us to go to slam with. Even if this bid remains (instead of having GiB bid keycard more often -- I assume this bid comes up most often when GiB thinks 4NT is quantitative) it would be very helpful to have a precisely defined meaning; e.g. "Go to slam with 2 trump honors."

 

2c) Capp doesn't make a lot of sense versus strong 1NT. The penalty X almost never comes up.

 

3) Is there an intrinsic obstacle to making GiB's system customizable? I've been experimenting with python scripts which can parse a standardized file format for bidding conventions and bid accordingly. (I used XML, which is easy to manipulate in python, to define specific restrictions which must be satisfied in order for a giving bidding sequence / continuation to be permitted.) I suspect GiB must have some sort of system for doing the same installed; otherwise changes to GiB's bidding such as the recent innovation of "no transfers to 4 card majors" (!) would require changing a hard-coded bidding algorithm, which seems painful.

 

So if GiB *IS* customizable, why not open it up to users? It would be very helpful for practicing more interesting bidding methods.

 

-sflicht

[barmar]

I don't agree with 2a. That's not how most people bid, why should GIB be different?

[orlam]

How about teaching GIB to raise in competition? That seems about 15 times as important as all the issues you raise combined. (I have seen passes with 5-5, one of these suits being the one partner overcalled, including two kings in the long suits.)

[fred]

1) Not going to happen. Sorry, but I don't want to get into a discussion as to why.

 

2) Thanks for the suggestions. Our GIB programmer's (Ari's) list of things to do is more or less infinite, but part of our general strategy for a while has been to focus on areas that our users care about. I will send him a copy of your suggestions.

 

3) The bidding system that GIB plays is customizable to some extent (actually to a fairly large extent). There are several factors that come into play in terms of why we have not opened up this possibility for our users. For now one particular reasons is all that really matters: currently GIB rentals are only available on the Windows client and development of the Windows client is currently (and likely to remain) frozen.

 

Fred Gitelman

Bridge Base Inc.

www.bridgebase.com

[peachy]

I don't agree with 2a. That's not how most people bid, why should GIB be different?

Absolutely agree.