[steve2005]

Hand
This was in a challenge match normal robots.
South opens 1♦ promising at least 3 cards in the suit.
Later South jumps in ♦ and Gib treats as a splinter in support of spades with 3♠ and now 1♦. So this is definitely a bug.
4♦ probably not wise bid but isn't the point.

After the splinter bidding gets out of control but again isn't relevant

Also interesting West Gib with A♠ cant find a double. Guess it was afraid it's 2 aces wasn't enough against 7NT? East surely also has a double. Mercy rule ?!

[johnu]

Instead of thinking 2 steps in advance, GIB can't think 2 1 step(s) in the past

1♦ shows 3(+) diamonds
2♥ shows 5(+) diamonds

What kind of program doesn't look at the previous round(s) of bidding for context on what the next round of bidding should be?

[smerriman]

johnu, on 2018-June-21, 22:58, said:

What kind of program doesn't look at the previous round(s) of bidding for context on what the next round of bidding should be?

My guess is that 4♦ was never considered to have a natural meaning, and there is a rule that says jumps to the 4 level, unless defined otherwise, are splinters. Not too hard to see how/why a program might have that rule.

You could change that to 'a splinter unless you've bid the suit naturally', but adding that rule in itself would be entirely pointless without defining what the jump actually means. (Ie, saying it's a splinter is no more helpful/unhelpful than saying nothing at all, like most other descriptions of "impossible jumps").

So the question really is, what should the description of 4♦ be? 7-4 distribution or something like that? It's a pretty rare one.

[steve2005]

This hand should pattern out with 3♥ showing 6-5.
So 4♦ should probably be something like 7-4 or 8-4.
Is unusual enough if was undefined wouldn't be a big deal.

But you can not have a bid show 1 card or 0-1 card when previous bidding has shown 3+ cards in the suit and really this bidding has shown 5+ cards.

This mistake seem to indicated something I suspected, Gib may not always be considering previous bidding when it makes a current decision. This may explain some of Gib's horrid bidding mistakes.

[steve2005]

smerriman, on 2018-June-22, 01:31, said:

My guess is that 4♦ was never considered to have a natural meaning, and there is a rule that says jumps to the 4 level, unless defined otherwise, are splinters. Not too hard to see how/why a program might have that rule.

Not an excuse. I presume Gib has 1♦-1♠-4♦ showing 6♦4♠.

[virgosrock]

steve2005, on 2018-June-22, 04:50, said:

Not an excuse. I presume Gib has 1♦-1♠-4♦ showing 6♦4♠.

Seems like most replies assume "human" logic and reasoning.
GUBBO is not "human".
I blame South for 1d-1s-2h and the jump to 4D.
Also, not saying GUBBO is right.

[johnu]

smerriman, on 2018-June-22, 01:31, said:

My guess is that 4♦ was never considered to have a natural meaning, and there is a rule that says jumps to the 4 level, unless defined otherwise, are splinters. Not too hard to see how/why a program might have that rule.

You could change that to 'a splinter unless you've bid the suit naturally', but adding that rule in itself would be entirely pointless without defining what the jump actually means. (Ie, saying it's a splinter is no more helpful/unhelpful than saying nothing at all, like most other descriptions of "impossible jumps").

There should be a decision matrix to decide if a bid was a splinter (or anything else). It's absolutely unbelievable that checking if the suit was bid naturally isn't the first condition to test so you could be right that jumps to the 4 level are splinters. That doesn't make the programming any less unbelievable.

smerriman, on 2018-June-22, 01:31, said:

You could change that to 'a splinter unless you've bid the suit naturally', but adding that rule in itself would be entirely pointless without defining what the jump actually means. (Ie, saying it's a splinter is no more helpful/unhelpful than saying nothing at all, like most other descriptions of "impossible jumps").

So the question really is, what should the description of 4♦ be? 7-4 distribution or something like that? It's a pretty rare one.

Maybe when in doubt, consider a bid natural? And if GIB is "confused" about a sequence, look back at the previous bidding to find a bid that is correctly defined which GIB seems to do a lot of (e.g. 11-21 points 5+ ♠ repeated for many spade bids in an auction)

[wbartley]

GIB already passes when it doesn't have a definition for a bid. So, if there is a rule that says jumps to the 4 level are splinter bids, change it to say, jumps to the 4 level are splinter bids unless the bidder has previously bid the suit naturally. You don't have to define 4♦. Just don't misdefine it.

[steve2005]

virgosrock, on 2018-June-22, 08:37, said:

Seems like most replies assume "human" logic and reasoning

Nobody is using Human logic.
We are using Gib's own definitions
1♦=3+♦,
2♥=5+♦
4♦=1♦

[jdonn]

Will have them look