[goffster]

This was in the free weekend survivor 2/2. Will have a hand record as they become available.

The bidding, with Robot in first chair, with no opposing bidding.

1C 1H
2H 3D
P!!!!

Robot correctly identifies 3D has a help suit with 4H.
Robot had a balanced hand with 4H.

I guess the description of the bid does not match the logic?
Or there is simply a bug?

Playing with people (much less robots) that pass forcing bids is an infuriating experience
even when it is right to pass.

https://tinyurl.com/yyxtk5af

[smerriman]

This is a long standing bug (I reported it myself over 3 years ago) - bugs just aren't being fixed unfortunately.

[pilowsky]

smerriman, on 2020-August-30, 21:03, said:

This is a long standing bug (I reported it myself over 3 years ago) - bugs just aren't being fixed unfortunately.

Is it possible that the robot was unhappy with your bidding Stephen?
I just gave your hand to the robots and this is how they bid and played it .

PS, I think there is a hyphen in long-standing.

[thepossum]

Interesting discussion. My sympathies are both ways. I would probably have bid a Soloway myslef but surely you cannot pass a game try suit when trumps have been agreed

[cherdano]

A good example why "GIB does fine compared to HUMANs in GIB robot tournaments, hence its standard is fine" arguments are rubbish - one reason an issue like this isn't fixed is that it probably doesn't come up in automated testing by self-play, as GIB probably doesn't bid 3D in this auction. So passing in auctions like this is an advantage for GIB in this comparison!

[goffster]

So are they no longer fixing gib bugs at all?

[pilowsky]

goffster, on 2020-August-31, 15:39, said:

So are they no longer fixing gib bugs at all?

Perhaps they are. Unfortunately, there is one thing GIB can't fix: people that post auctions without the complete hand or information - twice.
I went back to the teaching table and ran Stephens hand through it in a couple of different ways. This is what happened. Every way I looked at it the robots get to and make 6♠=.
It made no difference whether or not 3♦ or 3♥ is bid.
This is the Prime Area with advanced robots so I'm not sure if that makes a difference.
There is a tendency on this forum to characterise any failure on the part of GIB to achieve a desired result as a 'bug'.
This is because any error by a human is characterised as a 'mistake'.
Clearly, the ability of the human to achieve the desired result is relative to their current ability.
So is the ability of GIB. GIB can have 'a bad day' as well. You might have a non-advanced robot. You might get GIB when it has no time to simulate or when there is a power outage or all sorts of other things. ***** does happen to machines as well as humans.
That's why it took so long for computers to beat people at Chess and Go.

[goffster]

I did modify this post with a url.

I think it is clear to anyone that this is a bug.

This is not about achieving the optimal result.

There needs to be some sort of fallback logic in GIB when it
is in a 100% forcing situation where it does not pass.

i.e. I always rebid any agreed suit. I bid NT. I bid a new suit. I rebid a suit.
Anything except pass.

[terpdrew]

I'd never seen this bug before. But it is alive and well. Just a moment ago...

Similar auction with GIB passing a 100% forcing bid (game try).

First seat, vul vs non vul, ROBOT opens 1C with: A, 7632, Q10642, AK9

I bid 1H, Robot bids 2H, I bid 2S (forcing)... and Robot passes.

This is an obvious glitch. GIB itself shows it as a forcing bid. It's pretty straightforward to program a robot to NEVER pass what it characterizes as a forcing bid.

It's ridiculous this has been going on so long. First mention here that I saw was 2013. Should be an easy fix.