[USViking]

From tournament #1066 Robot Duplicate- MP 2011-11-20, 13:14, Deal 6:

http://tinyurl.com/76n89zp

The GIB 5♠ bid is explained as “4+ S; 5- 8421 HCP in S; 13+ total points”.

Is the 13+ total points a bug? I assume the 8421 treatment apples only to ♠,
but even if it applies to the whole hand with ♥K = 4 there are only 11 points,
unless the void is upgraded to 5.

Also, I wonder if some refinement might be possible in cases like this to convey
the information that GIB has 0 or maybe 0-1 points in the trump suit.

[Bbradley62]

This doesn't look like a bug to me... how many human Norths wouldn't investigate slam after the 4♠ call? Presumably, the description says 13+ simply because North is committing to slam opposite 19+ total points, therefore he "must have" 13 (or something similar to this).

It's unfortunate that South has so many wasted points. Can expert players tell us how they would have bid these hands?

[USViking]

Bbradley62, on 2011-November-20, 16:51, said:

This doesn't look like a bug to me... how many human Norths wouldn't investigate slam after the 4♠ call? Presumably, the description says 13+ simply because North is committing to slam opposite 19+ total points, therefore he "must have" 13 (or something similar to this).

It's unfortunate that South has so many wasted points. Can expert players tell us how they would have bid these hands?

If 8421 applies only to the Spade suit then the information conveyed by the 5♠ bid is off
by a whopping 3 points at the 5 level. That is like saying you have one King when you have zero Kings.
I will go out on a limb and express doubt that stars like Gitelmoss, Grall and Meckwell make a practice
of deceiving their partners by the value of one King when probing for a slam!

[Bbradley62]

USViking, on 2011-November-20, 15:40, said:

Is the 13+ total points a bug? I assume the 8421 treatment apples only to ♠, but even if it applies to the whole hand with ♥K = 4 there are only 11 points, unless the void is upgraded to 5.

USViking, on 2011-November-20, 18:25, said:

If 8421 applies only to the Spade suit then the information conveyed by the 5♠ bid is off by a whopping 3 points at the 5 level.

What in the world are you talking about? "5- 8421 HCP in S" means "5 or fewer points in spades, counting A as 8, etc." North has 0 points in spades, which is certainly consistent with "5 or fewer".

[USViking]

Bbradley62, on 2011-November-20, 18:49, said:

What in the world are you talking about? "5- 8421 HCP in S" means "5 or fewer points in spades, counting A as 8, etc." North has 0 points in spades, which is certainly consistent with "5 or fewer".

I have presented the issues with perfect clarity, and if you have not grasped them
then that is your problem.

[Bbradley62]

USViking, on 2011-November-20, 19:31, said:

I have presented the issues with perfect clarity, and if you have not grasped them then that is your problem.

The part that says "8421 points" and the part that says "13+ total points" have nothing to do with each other.

[USViking]

Bbradley62, on 2011-November-20, 19:38, said:

The part that says "8421 points" and the part that says "13+ total points" have nothing to do with each other.

If your reading comprehension skills were been up to the task it would be obvious to you
that I covered this, with perfect clarity.

Now, you have been unable to add value to this thread in three replies. Please consider
stepping aside, and letting someone else have the next go at it, OK?

[cloa513]

Bbradley62, on 2011-November-20, 16:51, said:

This doesn't look like a bug to me... how many human Norths wouldn't investigate slam after the 4♠ call? Presumably, the description says 13+ simply because North is committing to slam opposite 19+ total points, therefore he "must have" 13 (or something similar to this).

It's unfortunate that South has so many wasted points. Can expert players tell us how they would have bid these hands?

North should cue clubs, then South can easily subside in 5 ♠, if North had a stronger hand without fantastic distribution then 4NT or 5♠ is a better bid. The club void could have gold for South.

It wasn't the worst contract to end in anyway- just needs a 2-2 break in spades. Its the explanation that's wonky and the lack of other options than 5♠ (how uninformative can the bidding be).

[barmar]

The book bid with North's hand is Pass. In simulations it considers 5♣ and 5♠. In my 11 simulation runs at IMP scoring, the results were 2x5♣, 6x5♠, 3xPass. At MP, it was 3x5♣, 5x5♠, 3xPass.

Since the bid came from simulations, not the bidding DB, the explanation doesn't match the hand (as explained in the documentation we posted last week on understanding GIB explanations).

And the slam is a little better than cloa513 says. It makes with trumps 2-2 or singleton honor in West; when you see the honor drop on the first round, come back to hand and finesse due to restricted choice. Although this means that you go down on one of the 2-2 breaks (when West has QJ).

[calm01]

There seem to be three problems in the Robots bidding here:

- when GIB mentions it use of 8421 rather than 4321 I always underbid my hand as it is appears to be the Robots way of saying "I do not know how to value this hand properly and I tend to overbid in these situations.",

a) when GIB mentions total points in practice it seems that it is the Robots way of saying "I am in punt mode.",

b) when a human partner raises a major to the five level it is meaningless unless there is a prior agreement - I always go with partners viewpoint which has included:

* bid six if your trumps are better than you have promised,
* I have the trumps covered - is your playing strength better than you promised,
* grand slam enquiry - bid a side suit feature (as agreed),
* if you have a lower suit void bid it or 5N for a higher suit void so I can judge if we have wasted values for the grand,
* ...

when GIB employs a raise of a major to the five level I have yet to devise any consistent pattern.

But you may know better.

[USViking]

barmar, on 2011-November-20, 20:26, said:

The book bid with North's hand is Pass. In simulations it considers 5♣ and 5♠. In my 11 simulation runs at IMP scoring, the results were 2x5♣, 6x5♠, 3xPass. At MP, it was 3x5♣, 5x5♠, 3xPass.

Since the bid came from simulations, not the bidding DB, the explanation doesn't match the hand (as explained in the documentation we posted last week on understanding GIB explanations).

And the slam is a little better than cloa513 says. It makes with trumps 2-2 or singleton honor in West; when you see the honor drop on the first round, come back to hand and finesse due to restricted choice. Although this means that you go down on one of the 2-2 breaks (when West has QJ).

Thank you for the reply.

When either book or simulation produce a bid one whole King's worth worse
than promised I think serious thought should be given to making a change.

I am not sure what, but something.

[barmar]

Humans know that they should re-evaluate their hands when they've found a fit: add points for extra trumps, increase the value of short suits, double fits are great. But GIB's hand evaluation is static. it depends on simulations to overcome this, that's its replacement for the judgement that humans use.

North's hand only has 10 total points according to basic evaluation, but with two 9-card fits and a void, it has enormous playing strength, and that's what the simulations tell it. It's worth at least 13 points in this context.

[USViking]

barmar, on 2011-November-20, 22:27, said:

Humans know that they should re-evaluate their hands when they've found a fit: add points for extra trumps, increase the value of short suits, double fits are great. But GIB's hand evaluation is static. it depends on simulations to overcome this, that's its replacement for the judgement that humans use.

North's hand only has 10 total points according to basic evaluation, but with two 9-card fits and a void, it has enormous playing strength, and that's what the simulations tell it. It's worth at least 13 points in this context.

You are the expert, and I am not being sarcastic. If you feel the extras are worth a king
then you are right.

However, the hand has glaring drawbacks, such as it is aceless, and the trump suit is 10-high.
I am curious to know how those drawbacks are discounted.

I guess I should at this point admit I was going to bid slam over any move by GIB past game.

On the other hand I am not ever going to make a bid worth three points less than the mouseover
explanation. So far I am quite sure I have never deliberately bid with more than one point less.
Since the best part of GIB's game seems to be its declarer play maybe I will up my lying to two
points' worth, if it is clear that GIB will be playing it.

[barmar]

I never claimed to be a bridge expert. I'm an expert programmer, but I'm just a decent and enthusiastic bridge player.

The drawbacks aren't accounted for directly. It does simulations. It deals a bunch of hands consistent with the bidding so far, and then sees what would happen if it made each of the bids that it's considering. Then it chooses the bid with the highest expected value.

It's impractical for us to try to teach GIB all the clues and heuristics that bridge players use when evaluating and re-evaluating hands. First, we would have to reprogram it to make all these adjustments, and then we would have to update hundreds of bidding rules to take these changes into account. Instead, we depend on the simulations to fill in the gaps.

Am I saying that what we've got is perfect, or even great? No, GIB is not an expert bridge player (despite the claim Ginsberg made in the paper he published when he wrote it). We're constantly tweaking the bidding rules, to fix glitches. But we're not likely to make any major redesigns to the program itself any time soon. As I get a better understanding of how it works, I'm going to make some adjustments here and there, but the overall approach it uses is not likely to change. The way GIB is designed, it's not likely to ever become a great bridge program. It's like remodeling a home: you might repaint it, add an addition, or redo the kitchen, but a tract house will never become a mansion without being torn down and replaced.