[lycier]

Suggestion on new Gib bidding system.

Hello,fred,uday and BBO staff:

Now Gib have made many improvements - from basic to advanced,it is a hard work,not easy !
As we know that Gib also makes many ridiculous bidding sequences,this is a fact.
Why does Gib appear this phenomenon?How to avoid it? I think the main reasons are the current Gib bidding system is not good enough to deal with many complicated hands.In other words,almost of conventions including some bidding standard have been out of fashion,it have been got eliminated.
I think it needs to creat new Gib bidding system - I call it High Version for time being.
Now please allow me to make a rough comparison between Gib 33 version (G33 for short) and High Version (HV for short) at below.

1- Responding principle
G33 - up the line,its defect is it can't display the balance or the unbalance.
HV - walsh resonding,its advantage is it can display the balance or the unbalance,including length of suits.

2- Most important artificial
G33 - NMF : NMF is a great milestone in the history of the bridge,It have developed more excellent extension application - that must be more better than NMF.
HV - Two-way stayman and XYZ.
Here my suggestion is Modified Two-way Stayman by Grant Baze.Please view his point :
Grant Baze said:
Modified Two-way Stayman replaces New Minor Forcing after a 1NT rebid by opener. MTWS is better in terms of sign off sequences, invitational sequences and forcing sequences. Additionally it has a few frills and benefits not seen with NMF. MTWS applies anytime opener bids 1 of a suit , responder bids 1 of a suit, and opener bids 1NT providing responder is not a passed hand. MTWS still applies if there is 1 level competition from the opponents.

3- Bidding standard,for example:
G33 - Soloway jump,its defect is the applied probability is lower,BBOers rarely use it except Gib.
HV - WJSC or WJS,its advantage is the applied probability is higher,most of BBOers often use it.
G33 - 1M -- 3M,showing limited raise
Stangely,for 3th seat or 4th seat,Gib plays 2C Drury,also plays 1M -- 3M limted raise,I am afraid that this is not a normal thing.
HV - On the contrary,it is a preemptive raise,and has been gradually accepted by everyone,this is a fact.

4- More advanced modern 2/1 approach,for example :
Wolff signoff,concealed splinter,snapdragon double,flip-flop and minorwood,etc.
Here note that it is time to improve major raise system at present.

5- The hand evaluation:
G33 - 8421 and total points
8421 has been eliminated on the Gib version 28.
The main defects of Total Points are static and mechanical,it can't show its real values in the process of bidding sequences including competition.So some ridiculous bidding sequences inevitably happened.
HV - A new evaluation method: Value Point (VP for short)
VP :don't add shout suit point except dummy,add length point,distribution point,dummy point and voidness point(=trump length on the dummy),etc.

The main advantage of VP are
1) Dynamic,not only do players increase its values,but decrease its values as well in the process of bidding sequences.
2) Humanized auxiliary,we would say that the stand or fall of evaluation can show real skill of players.This is also the most realistic simulation.

BBO is a great site in the world,its Gib should keep pace with the times.
Keeping pace with the times means that all the theory and work of the Gib system must conform to the times, display great creativity.
Please forgive me if I have a wrong idea.Hoping that BBO can make the most outstanding robot in the world.

Best Wishes

lycier

[Stephen Tu]

They do have to improve the hand evaluation function. But I don't think adding gadgets as you suggest is the way to make GIB better, at all. I play many of the gadgets you list with my human partners, but the reality is that I can bid maybe 99% as well without them. GIB's ridiculous bidding is not a result of not playing these particular gadgets/treatments. Just go back and browse through all the threads you posted in the past few months. What percentage of these would be fixed by adding these gadgets, vs. just being outright bugs? 1 out of 100?

The problem is that the number of possible bidding sequences is really extremely large. And unlike humans, who when encountering a new sequence they haven't seen before, can extrapolate from analogous situations in other sequences what bids ought to mean and which they should choose, computers need rules to tell them what to do for each one. They don't have "common sense". That's why so many sequences, particularly in competition, go awry. So they have to keep on fixing existing rules, and adding new rules to deal with complicated sequences.

Adding gadgets isn't going to solve it.

[lycier]

Hi,Stephen Tu


What you said are two things with me.
In effect,there are many life circles in playing the bridge at bbo,for example,from teams,to groups,clubs,and leagues,etc.However the most important life circles are entertainment and competition.
If the proposition of Gib for mass entertainment,now your option is ok.
If the proposition of Gib for competition,there are many considerations for other purposes,it should be my option.

As for you said "Adding gadgets isn't going to solve it. ".
I can tell you those are two different system - there are many of great differences between the current Gib system and my High Version Gib - including different responding methods and bidding standard,and most of gadgets have upgraded or no longer in use.
It is well known to all,most of gadgets including bidding standard in use of entertainment circle are no longer in use for competition.
Here we don't need to rhetoric.The current Gib system is just a tipical example.

If you have some objections,please you look on JEC,Helene_t etc. - they as experienced expert players never play similar Gib system !!!
Why don't learm from them?
Here,thank you for your kind reply,and again.

lycier

[Stephen Tu]

We all want GIB to be a stronger partner & competitor.

Learning gadgets does not make you a better bidder, when you have major holes on many common auctions. If I am teaching an intermediate/advanced human partner to be a better bidder, I am correcting their basic bidding judgment and frequent errors first. Not teaching them every gadget under the sun when they haven't thoroughly learned basic sequences first. They will play much better when eliminating the basic errors that crop up every 1 out of 8 boards, than learning gadgets that come up once every 200 boards and only affect the final result on those boards 1/10 of the time, effectively obtaining a better final result 1/2000 boards.

Bidding needs strong foundation of not making mistakes on basic sequences. Learning gadgets doesn't fix a shaky foundation.

I will pick a strong player as partner who knows very few gadgets but knows those gadgets cold as well as all basic sequences every single time vs. a player who knows 50 more of the latest conventions but making basic errors every fifth board.

Experienced expert players do not play identical system to GIB, sure. But that is because they already mastered the basics, and can afford to spend time optimizing sequences that come up or matter very rarely. And every single one of these players, if asked by a lessor player whether they need to get their basics solid first, or learn more conventions, would tell them to work on the basics!

BBO has limited development resources. Time spent on adding gadgets is time not spent on eliminating ridiculous sequences. Even if they had more people to work on it, they should still just work on fixing the ridiculous sequences at a faster pace first, rather than devoting one person to adding gadgets.

[manudude03]

When two partners agree on a system, they can make it as elaborate as they like since no-one else has to remember it. When you implement a system by a common engine, everyone who is using that engine needs to remember it. That being said, you could add all the gadgets you like, but it only makes it harder for beginners to cope with them.

Recently you made a number of topics on this forum about server-side issues. Your suggestion is only going to make that problem a lot worse, as well as being generally inconvenient and harder to maintain.

[lycier]

Hi,Stephen Tu,hi manudude03

It seems you have a lot of misunderstanding,this is not really a simple problem about adding some gadgets at the end of the day.
Now there is a simple issue - who prefer up the line to walsh unless Gib insist on it ?
All of us know that walsh responding is more better than "up the line",walsh responding is just a decent option of bidding style,not other.
There are too many of ridiculous bidding sequences caused by " up-the-line" responding, why not play walsh?
Any idea?

lycier

[Stephen Tu]

Point us to a thread that you think playing Walsh would have saved GIB. Compare this to the total number of threads reported.
Seriously what percentage of threads are dealing with auctions that start 1♣-1♦? Go back and find one. And even if you do find a problem where GIB had an up-the-line accident, how many of those could just be fixed just having GIB bid in a sensible manner consistent with up-the-line treatment, vs. one where Walsh clearly leads to a better result?

Compare to the number of threads caused by the issues in "categorizing GIB flaws", the thread started by 1eyedjack. Things like being unable to handle "double because considered too strong to overcall in a suit". Weird "penalty" doubles that only promise 2+ trumps. High level bids in competition that should be based on shape, and might be based on willingness to sacrifice or two-way shots (either we make or it's a good sac), but for GIB promises some absurd 25+, 30+ hcp which leads to subsequent overbidding.

Switching to Walsh solves none of these. It's just not right to prioritize rare problems over common problems. Especially when changing the treatment has a lot of follow-up inference changes, changing is likely to introduce even more bugs. Lets fix the most important bugs we have first. When GIB is solid in all these other areas, then think about changing to Walsh.

[1eyedjack]

lycier, on 2015-April-24, 11:18, said:

There are too many of ridiculous bidding sequences caused by " up-the-line" responding, why not play walsh?
Any idea?

lycier

Perhaps because there are too many fee-paying customers below expert level who do not play Walsh.

I think it fair to say that just about anyone familiar with Walsh will be familiar with non-Walsh. Probably learned the game on non-Walsh before converting. The reverse is less obvious.

There is no excuse for ridiculous sequences caused by up-the-line responding, and in such cases there will invariably be a solution without resorting to Walsh.

Why not program it to play transfer Walsh, to solve all the ridiculous sequences created by straight Walsh?

[lycier]

Hi Jack:

Very good reply.
I am particular grateful to you for your timely help and I finally see the truth.
I believe that this kind of robot will be eliminated sooner or later,very backward.
No matter how to repair,it is very very difficult to get more better skill.


lycier