[lorserker]

hrothgar, on 2024-November-22, 09:29, said:

Thanks very much for posting how well GIB / Ben do in the various events

If possible, it would be interesting to see the Standard Deviation in addition to the means.

To me, the most interesting result is how much worse GIB does in the Zenith daylong than any other event...

Are the conditions of contest significantly different?

Zenith has 120 instances.
If we look at how the 120 MP-percentages of the instances are distributed, the standard deviation is about 7.
The score of an instance could be anywhere between 40% and 70%, but is mostly in the low 50s. And the distribution is a bit skewed towards higher scores.

The reason why GIB is a bit weaker on Zenith is because GIB's strength is declarer play and GIB's weakness is bidding, especially competitive bidding.
In Zenith there is less declaring and more competitive bidding (because the boards are random, not best hand) and this disadvantages GIB.

In my opinion, when users get an impression of how robots play, subjectively declarer play doesn't count as much. Because in a typical best hand tournament the robot rarely declares.
What the user is most exposed to is bidding with the robot, and some defense. With bidding being the weak spot (very much for BasicGIB), the impression is that the robot is much worse than it is.

[barmar]

While GIB may have some flaws in its logic, where it excels compared to humans is in its memory. How many times have you gotten a poor result because you (or partner) didn't notice someone show out in a suit, couldn't remember which spot cards were played, or forgot something from the auction? Computers never make these kinds of silly mistakes.

I think I've even read examples of such mistakes in high-level tournament reports, although they're obviously very rare among champions. But they're not infrequent among average players.

Just getting these basics right all the time gives the robots a significant edge.

[benellis58]

GIB doesn't revoke, doesn't lead out of turn, and doesn't forget cards. Those are its good points. Everything else is wretched. Its system is pathetic, its leads are horrible, its logic is non-existent, its defence is disgustingly bad, and even its declarer play isn't very good. Furthermore, most of its definitions are poorly written, often illogical, often incorrect, and often ridiculously inconsistent. GIB should have been replaced or at least improved long ago. The ineptitude of the GIB robots removes a huge amount of pleasure that BBO otherwise provides.

[mycroft]

Again, then why, when GIB plays with GIB in a club field, do they regularly come third or fourth with a 54% game (even playing against people whose system they can't understand)?

I treat GIB as I do the games I play (have to play?) as fill-in director with life Cs who know how they play bridge, do not want to learn anything (*), and will not read my signals anyway. At least some of which, frankly, you are going to get with any pickup partner, even ones much better than you. My job is to make their life as easy as possible, playing their way as much as I can so they can play their best so I can get the best score possible(**).

I actually hate GIB on defence. If there's ever a situation where "I'm good unless they find the switch", if GIB's on lead, I'm going down at least twice as often as if it's a human. Sure and I intensely dislike defending with GIB, too, but still.

If you don't like robot games, Don't Play Them™.

Oh, and if you're upset with no advancement, go to the BEN page and sign up. Maybe that's better for you!

I don't like the Permanent Pickup Pool on BBO because it is concentrated with players nobody would voluntarily partner (due to skill, inflexibility, attitude, ethics, ...) because the ones who can, get regular partners and get out of there. So I don't play pickup on BBO (unless the partner is vouched by someone I do know).

I try not to play against pickups on BBO, either, because I so rarely play the One True Bidding System, and not playing OTBS annoys pickups no end, so the average lifespan of a partnership is 3-4 hands.

That does limit my use of the product (as does the fact that one of my regular partners hates playing online). I don't gripe about it, because it is a great solution for many many people who *don't have my issues* with that; and they help support the parts of BBO that I like.

(*) You don't want to try teaching them anything, not during that pickup game, either. All that will happen is that they think so hard on that thing they don't understand that the rest of their bridge game will go downhill. I'm talking about the weeks and weeks before where others are pointing out the holes and they wall up.
(**)I know, wrong order. Edits. Why do I seem to be channelling Simon so much this week? Third time? Fourth? If you don't know how to play with Mrs. Guggenheim, you don't know how to play with (or against) GIB.

[pescetom]

mycroft, on 2025-January-19, 12:21, said:

Again, then why, when GIB plays with GIB in a club field, do they regularly come third or fourth with a 54% game (even playing against people whose system they can't understand)?

Yet again, I dispute this claim in a club field where the dominant system is even moderately different from GIB's system.
My experience in a (not very strong) club field playing mainly Italian variants of 2/1 is that GIB-GIB initially scored 58% but then slid down to 45% or slightly less.
Some of that difference is due to GIB grossly misunderstanding 2♦ Multi, but only about a quarter of our pairs played that.
Perhaps more is that the players did not understand the often misleading explanations offered.
I am curious to know what would happen in a club in Poland

[jm02130]

The robots' pulling obvious penalty doubles has always been infuriating, as has their tendency to assume captaincy with weak hands in high level auctions. But they seem to have become even worse recently, and I wonder if programming changes were made.

[barmar]

jm02130, on 2025-January-24, 13:29, said:

The robots' pulling obvious penalty doubles has always been infuriating, as has their tendency to assume captaincy with weak hands in high level auctions. But they seem to have become even worse recently, and I wonder if programming changes were made.

No changes have been made to GIB's bidding in years (not counting the addition of the SAYC system). We recently made a few minor tweaks to its play (see the announcement of GIB v41).

[benellis58]

I found this comment by jm02130 to be quite interesting: "...they seem to have become even worse recently". I have said many times to my friends, who all agree that the GIB robots are stunningly incompetent, that it seems the robots are getting even worse, but that in reality they probably are not. It only seems that they are because they are so utterly wretched that they manage to commit new, creative atrocities on a regular basis, while at the same time continuing to annoy users with all their longlasting and foolish habits. Surely there are better robots than the horrendous GIB ones, so it would be nice if BBO somehow managed to provide its customers with better ones. Sadly, BEN does not seem to be the solution, but whatever the solution is, it's long past the time to send GIB to a smoldering garbage pile.

[GlobalClub]

The results seem pretty consistent. How about in Goulash tourneys?

[GlobalClub]

barmar, on 2025-January-30, 19:06, said:

No changes have been made to GIB's bidding in years (not counting the addition of the SAYC system). We recently made a few minor tweaks to its play (see the announcement of GIB v41).

Sad. Very Sad.

Couldn't an AI that read all your archived boards develop a bidding strategy you can put in a Rust program to work better and faster?
Can't you interest some PhD student?

[lorserker]

GlobalClub, on 2025-February-10, 01:57, said:

Couldn't an AI that read all your archived boards develop a bidding strategy

That is Ben.
I have enabled it for you so you can try it. (find Ben & Friends under free tournaments)

[harikannan]

If i am not wrong, i read somewhere that it is possible to play 'friends challenge' with twooneben, the AI. When i try, the challenge gets listed -- but does not open for play. What should i do?

[diana_eva]

I split the benellis58 comments into a dedicated GIB bashing thread.

Let's keep this one for that it was meant, which is benchmarks of various robots used on BBO.

[benellis58]

Thank you, Diana. That was an excellent decision and will greatly simplify things for all concerned.