[dwar0123]

If we can look past what I am sure will be regarded as my retarded 4♦ bid, I still put forth the 4♥ bid is ill advised with any possible hand I could have. It may have been induced in part by the silly description of the 4♦ bid which I suggest should also be changed to reflect something more realistic.

---

Btw, check out what happened to the people who doubled and then bid ♦, which will explain why this result ended up above average.

http://www.bridgebas...ername=dwar0123

[nathan2008]

your fault lol. With GIB, every time if u bid one more times, it means you have more extra values. This hand u must pass 3s because all you have are 2 aces, opps probably have DA and all your D's are useless. What u should do is to adapt yourself to GIB style, not complain about gibs lol. I never complain about gibs.

[nathan2008]

every time if there is a mistake , it is always mine lol. I will tell myself Gib can't understand this bid, then i must avoid doing this next time.... With this i make much progress now!

[pigpenz]

only 6 out of 18 found defence to beat 4♠

[cloa513]

 pigpenz, on 2012-April-11, 09:28, said:

only 6 out of 18 found defence to beat 4♠

And none of them with a sensible lead of spade- given dummy poor support for spades it should be totally obvious from simulations- its just the pointless maximum trick evaluation that probably screws it up.

[barmar]

The book bid with North's hand is actually 5♣. With the strength you've shown, it thinks you should be in game, so the question is just which suit. His hand is worthless as a dummy in ♦, but if he can take some trump tricks then your powerhouse should take care of the rest. The rule says to bid the longest suit.

But then it ran simulations. In its simulations, it's having a hard time finding hands that fit the bidding. It ended up with quite a few hands where your hearts are AKx or AKxx, so 4♥ seemed like a better place than 5♣. In many cases, the opponents took the push to 4♠, and then it expected you to make a forcing pass, and it will then double when it comes around again, and they go for a number.

[dwar0123]

 barmar, on 2012-April-11, 21:06, said:

The book bid with North's hand is actually 5♣. With the strength you've shown, it thinks you should be in game, so the question is just which suit. His hand is worthless as a dummy in ♦, but if he can take some trump tricks then your powerhouse should take care of the rest. The rule says to bid the longest suit.

But then it ran simulations. In its simulations, it's having a hard time finding hands that fit the bidding. It ended up with quite a few hands where your hearts are AKx or AKxx, so 4♥ seemed like a better place than 5♣. In many cases, the opponents took the push to 4♠, and then it expected you to make a forcing pass, and it will then double when it comes around again, and they go for a number.

That is all nice and logical, just as you would expect a computer program to be, but can we adjust something so that it doesn't make such a silly bid?

Some obvious issues.
1. South can't have 4♥, he didn't double.
2. While the hand might be described as 25 points if the intent was to make, a part score or game sac here might be considered by some. It is far more likely south has a very distributional 20 total point hand.
3. Even if the hand is described as 25 points, it is clearly not hcp and blindly pulling partners 25 total point diamond hand into an unsupported 4 card major is not good bridge.

I think stephen tu suggested in another thread that any hand that gets described as 25+ total points is suspect. The gib robot bidding goes off the rails, bearing little resemblance to what a human would do and fairly universally wrong as the cards lie.

The majority of human players got this one wrong and I would wager that of those that got this one right, most got it right only because they knew the gib robot would screw it up and declined to make the normal bid.

Sure I can learn to adapt to the gib robot, but is that what we really want people to do? Can't we also improve the gib robot?

[barmar]

I just tried putting an upper limit on the 2♦ bid, and that did keep North quiet.

This is a common problem throughout the GIB rules. The rules are written with priorities -- the rule for doubling with a strong hand has a higher priority than the rule for overcalling in a suit. So there's an implicit upper limit on the strength of the overcall. But when GIB is figuring out what a bid shows, it doesn't automatically subtract out the meanings of all the higher priority rules that might have matched, but didn't (in many situations there would be hundreds of them -- most of them are irrelevant, but the computer can't tell which); these are the kinds of "what-if" inferences that humans are really good at, but are hard AI. So every rule needs to be explicit about what it shows, and WE have to figure out what's precluded by other rules so we can specify it. Unfortunately, in most cases this hasn't been done. And in many cases, it's not obvious, because low priority rules can match many different situations. To keep from proliferating lots of rules for similar situations, there are many general rules with placeholders for the level and suit, so we can have a rule like "If you're just raising the level one step, you show N more points than you previously showed."

I wish I had a better answer for you, but it's really not practical for us to go through all the rules, adding more specifications like this, or splitting general rules up into more specific rules -- there's too many of them, and not enough hours in the day. So we're just going to keep handling them on a case by case basis.

[Stephen Tu]

 barmar, on 2012-April-12, 00:12, said:

To keep from proliferating lots of rules for similar situations, there are many general rules with placeholders for the level and suit, so we can have a rule like "If you're just raising the level one step, you show N more points than you previously showed."

In competitive auctions, it would be nice if this general rule could be adjusted so that rebidding a suit you've shown previously showed additional cards in that suit rather than additional HCP.

[barmar]

 Stephen Tu, on 2012-April-12, 09:23, said:

In competitive auctions, it would be nice if this general rule could be adjusted so that rebidding a suit you've shown previously showed additional cards in that suit rather than additional HCP.

It does that as well (actually, it increases the suit quality, which generally increases the length). I was just giving examples of the kinds of things general rules can do, not describing a specific rule completely.

Notice that in the auction in the OP, 3♦ shows twice rebiddable, while 4♦ show strong rebiddable.