GIB consistency

[jonottawa]

I don't care so much about GIB mistakes. I make plenty of them myself. If GIB could wonder it would doubtless wonder why I am so bad at keeping track of spots or counting out a hand.

 

But what I find quite annoying is when GIB makes a mistake against everyone else but me. Take this wonderful hand:

 

All red IMPs

 

You hold:

 

6 64 AKQ AKQT964

 

You open 1♣ in 3rd seat, pard bids 1♥ and you bid ...

 

2♣???

 

Oh wait, no, it's jonottawa, so I'll reverse. (Here are a few other examples.)

 

I suspect this bug has something to do with the slider that adjusts how long GIB is given to 'think'. Rather than a slider, I'd like to see 4 or 5 options and the boards compared against other people who chose the same option. That way GIB's play would (presumably) be more consistent across the field.

 

(And yes, I realize that from time to time I benefit from this behavior, but I think those of us who give GIB a little time to think tend to get burned in the long run (because it improves the play of 2 opps and only 1 partner.))

 

On another note, I'd like the table option of playing all hands for our side when playing with GIB (like the current default option in some tournaments.)

[Bbradley62]

Fun hands! I hope you reported all those bugs... some of them are pretty ridiculous. Edited September 27, 2012 by Bbradley62

[Stephen Tu]

I suspect this bug has something to do with the slider that adjusts how long GIB is given to 'think'. Rather than a slider, I'd like to see 4 or 5 options and the boards compared against other people who chose the same option. That way GIB's play would (presumably) be more consistent across the field.

 

As has been stated before, there are 3 versions of GIBs being played on BBO. You are using the oldest version, with the windows standalone app client, running GIB on your own machine. Adjustable thinking time, which is good, but older bidding db with whatever bugs there were the last time it was updated, years ago. Then there are the basic bots and advanced bots accessed via the web flash clients (and smartphone apps), which spawn GIB processes on BBO's server farm. The basic bots are cheap but they have been lobotomized with the older weaker DD-only playing algorithm, playing on very fast settings, and I think less (no?) simulation in the bidding, so their bidding judgment and play is markedly poorer than the advanced bots or the win client (in cases where there isn't a bidding bug fixed in later version but unfixed in win client). But since they are cheapest ($1/week), that's what the vast majority use.

 

Ideally one would only compare against people using the exact same software and settings, if you wanted a fair measure. But there may not be enough people still using the older client or advanced bots to get a decent # of comparisons.

 

So if you want to be compared against "most other people", sit south and play against basic bots using the web client. Or play tourneys where the bots are advanced and equalized.

 

I personally prefer the more challenging game vs. the advanced bots. I use web version for fewer bidding bugs. Pay for them by playing the robot races.

[jonottawa]

As has been stated before, there are 3 versions of GIBs being played on BBO. You are using the oldest version, with the windows standalone app client, running GIB on your own machine. Adjustable thinking time, which is good, but older bidding db with whatever bugs there were the last time it was updated, years ago. Then there are the basic bots and advanced bots accessed via the web flash clients (and smartphone apps), which spawn GIB processes on BBO's server farm. The basic bots are cheap but they have been lobotomized with the older weaker DD-only playing algorithm, playing on very fast settings, and I think less (no?) simulation in the bidding, so their bidding judgment and play is markedly poorer than the advanced bots or the win client (in cases where there isn't a bidding bug fixed in later version but unfixed in win client). But since they are cheapest ($1/week), that's what the vast majority use.

 

Ideally one would only compare against people using the exact same software and settings, if you wanted a fair measure. But there may not be enough people still using the older client or advanced bots to get a decent # of comparisons.

 

So if you want to be compared against "most other people", sit south and play against basic bots using the web client. Or play tourneys where the bots are advanced and equalized.

 

I personally prefer the more challenging game vs. the advanced bots. I use web version for fewer bidding bugs. Pay for them by playing the robot races.

 

Thanks for the info. I vaguely knew about different GIBs but you've summarized precisely what I wanted to know. It's odd (that my GIB is the 'dumbest' of the lot) because it seems to me that at least 2/3 of the time when my GIB opps do something unusual (relative to other GIBs) that it is successful, but that may just be my natural pessimism.

[Stephen Tu]

Thanks for the info. I vaguely knew about different GIBs but you've summarized precisely what I wanted to know. It's odd (that my GIB is the 'dumbest' of the lot) because it seems to me that at least 2/3 of the time when my GIB opps do something unusual (relative to other GIBs) that it is successful, but that may just be my natural pessimism.

 

No, your GIB isn't dumbest. It's got the oldest bidding database so more bugs of that nature. But its play and bidding judgment if your slider is set reasonably should be quite a bit better than the "basic" bots, if it's not hitting some egregiously bad definition in the database. Basic bots are the worst overall.

[cloa513]

The link via "examples" says nothing- the variation is all due human stuffups.

[georgi]

The case with rebidding simply 2♣ is fixed.

 

Most of the fixes directly reflect on the Basic Robot, Advance Robot has some capability to avoid them unless it's general bug.