Argine Vs GIB
#1
Posted 2019-May-02, 15:38
Argine (FunBridge)
GIB (BBO)
I think both are rather good, but which is the best?
/Bjorn
#2
Posted 2019-May-03, 04:44
#3
Posted 2019-May-03, 05:35
As I understand, entrants to the World Computer-bridge Championship install their software on regulation machines at the playing site (that of the given year's World [human] Bridge Championships).
A skim of the Funbridge website seems to indicate that it's available only at a server.
Have you tried challenges against Argine, not so easy to play against.
#4
Posted 2019-May-03, 08:00
#5
Posted 2019-May-03, 08:55
etha, on 2019-May-03, 04:44, said:
That is clear but probably not the point - we don't really need a world computer bridge champion to play with on-line.
For BBO-like purposes, Argine is almost acceptable whereas GIB is not.
In everything except slam bidding it is better than the average human.
#6
Posted 2019-May-03, 09:27
pescetom, on 2019-May-03, 08:55, said:
For BBO-like purposes, Argine is almost acceptable whereas GIB is not.
In everything except slam bidding it is better than the average human.
I don't know about "average human", but GIB does pretty well competing against BBO players. We've had it play the hands in daylong tournaments, and it averages in the high 50's.
#7
Posted 2019-May-03, 11:51
No not true
Argine could play WC
We put Argine in once but it’s not the same goal : Argine is programmed to play with million of different partners
Others computers are programmed to win the championship that’s mean overcall and preempt all the time!
Another view is that Argine is based on Wbridge5, comment from Jerome:
In the past Argine was build on Wbridge5 but it was a nightmare for our client so we change it. Now Argine is based on MBA
/Bjorn
#8
Posted 2019-May-03, 13:32
etha, on 2019-May-03, 04:44, said:
Based on software reviews, I'm guessing that if they entered the tournament, they wouldn't be the worst programs in the field. I'm not saying they would be close the best, maybe middle or lower middle of the pack if I had to guess. There is no minimum level of bridge competency to enter.
That being said, it's very expensive to compete in the computer championships. Competitors have to physically travel (probably on an international flight) to the site, and pay for several days of hotels and meals during the tournament.
Most companies enter their programs for bragging rights. If they win or do well, they can include those results in advertising and hopefully increase sales. At the worst, they get a better idea how well their product performs compared to the competition.
In GIB's case, BBO still has an old version available for sale, but why would anybody buy a 20 year old program? They still make patches, but the main purpose of GIB (and Argine) these days is for online robot play. It really doesn't matter if GIB (or Argine) is any good or not. If you want to play online and compare scores against other players, you don't have a choice.
#9
Posted 2019-May-04, 05:12
I joined funbridge and so far can see no way to run a match but I'll have a look when I get more time. At least they seem to be actively trying to improve their engine.
#10
Posted 2019-May-04, 08:57
Try a challenge Against Argine, she is not easy to play against.
#11
Posted 2019-May-05, 03:42
#12
Posted 2019-May-05, 09:11
Hah I think I found a way to switch off the 2nt response to stayman maybe this is the best online multi player robot after all. Still no idea how strong it is though i'll try to fake some matches where just one of the players is another robot. I may be able to do this and maybe gives some indication of strength. Robots generally are really really bad at playing with other robots that aren't themselves.
I seem to have found a new insane method I can't remove though. 1!c (1!d) dbl = exactly 4 hearts. If you disable this it turns off all negative dbls. Again no way to report this.
#13
Posted 2019-May-06, 02:41
#14
Posted 2019-May-06, 02:57
#15
Posted 2019-May-06, 03:27
Kt954
qjt8
j2
k4.
p-(1nt)- ?
You are forced by gib to play capp. no alternative. Qplus thinks 2♦ is 1.3 imps better than pass. GIB thinks this shows 5-5 so we'll see if that causes a problem.
2♦ (2nt) (3c) (3nt)
#16
Posted 2019-May-06, 03:57
kq64
akt4
t632
4
favourable.
1♦ p p dbl (2♣) (2♥) (4♥)
do we dbl? qplus thinks it will gain 3.1 imps so he dbls.
Gib of course in insane and bids 4spades.
Lucky we had some spades I guess. Amusingly it only went one off and 4!h dbled might make.
#17
Posted 2019-May-06, 04:34
aq
kqt94
a63
kj5
qplus thinks 2nt is 0.5 imp better than 1♥. So I'll make that bid.
#18
Posted 2019-May-06, 04:59
p (1♦)
unfavourable.
qt975
6
kq42
k42 is this worth 1♠? qplus thinks bidding is 0.1 imps worse than pass so pass.
p (1h) (1nt) and qplus bids 2!s quite surprisingly. dbl p p p
#19
Posted 2019-May-06, 05:30
akq862
at84
j9
2
p-?
favourable. qplus thinks 0.2 imps better to bid 4♠ than 1♠.
#20
Posted 2019-May-06, 05:35
all
7
8652
kjt86
ak8
Is this an opening bid ? qplus thinks yes by 0.2 imps