Free Weekly Instant Tournament - October 6 - Board 1

Board 1
Neither vulnerable

♠ 9 4 3  ♥ A K Q 6 4 2  ♦ A 6 2  ♣ A

Two passes to me. I open with one heart, LHO passes, and partner raises to two hearts. I'm worth driving to game, but it's not clear which game to drive to. I have eight tricks off the top, so one trick in partner's hand, assuming it's fast enough, is all I need for three notrump. Finding ten tricks in hearts may be harder. 

At IMPs, I would certainly bid three notrump, since it rates to be the safer game. But at matchpoints, the prospect of a spade or diamond ruff in dummy tempts me to try four hearts.

Players often get this backwards, spurning a major to play three notrump at matchpoints but not at IMPs. At matchpoints, it's better to play the major if it takes only one trick more than notrump. At IMPs, the major must take two tricks more (ten versus eight) before the major is better. So, in general, you should choose the major more often at matchpoints than at IMPs.

I bid four hearts, everyone passes, and LHO leads the seven of hearts.

	NORTH Robot ♠ Q 6 5 ♥ J 9 8 ♦ K 8 7 ♣ J 10 9 8
	SOUTH Phillip ♠ 9 4 3 ♥ A K Q 6 4 2 ♦ A 6 2 ♣ A

West	North	East	South
Robot	Robot	Robot	Phillip
	Pass	Pass	1 ♥
Pass	2 ♥	Pass	4 ♥
(All pass)

Three notrump is clearly the right spot at IMPs. Unless the opponents can take five spade tricks off the top, three notrump is cold, and four hearts is tenuous. It's not clear what the right spot at matchpoints is.

It's unlikely West led a trump holding the ace-king of spades, so I will need to find my tenth trick in the club suit. I can take two ruffing finesses, but I need three dummy entries to do so. This means I need to reach dummy twice in hearts. Does that work? I can unblock the club ace, lead a heart to dummy, and lead the jack of clubs, ruffing if East covers and pitching a spade if he doesn't. Then another heart to dummy for the second club finesse. Finally, I return to dummy with the diamond king to take my pitch.

Yes, it works, assuming East has at least one club honor. The question is, do I have two trump entries? If I play low on this trick and RHO covers with the ten, then yes, I do. If the ten does not appear, I will need to decide whether to play for two-two hearts or to finesse West for the ten.

I play the heart eight. East covers with the ten. Problem solved.

I win in my hand with the king, cash the club ace--four--eight--deuce and play the deuce of hearts to dummy. West pitches the diamond four. So East made an error in covering. I would go down had he withheld the ten.

I lead the club nine. East covers with the queen, and I ruff. West follows with the six.

I lead another heart to dummy, lead the club ten, pitching a spade, and lose the trick to West's king. Eventually I pitch my diamond loser on the good club. Making four.

	NORTH Robot ♠ Q 6 5 ♥ J 9 8 ♦ K 8 7 ♣ J 10 9 8
WEST Robot ♠ A 8 2 ♥ 7 ♦ Q J 4 3 ♣ K 7 6 5 4		EAST Robot ♠ K J 10 7 ♥ 10 5 3 ♦ 10 9 5 ♣ Q 3 2
	SOUTH Phillip ♠ 9 4 3 ♥ A K Q 6 4 2 ♦ A 6 2 ♣ A

Plus 420 is worth 93%. No one was in three notrump. Most of the field was in four hearts going down. As I've observed before, the field in these Instant Tournaments is terrified of not drawing trump. Any time you must delay drawing trump, you get a ridiculously good result.

East could defeat this contract by playing a low heart at trick one. This is the kind of play a human is more likely to find than a robot. Robots decide what to do by generating random deals. If a robot doesn't happen to come across a deal where a particular play matters, it has no reason to find that play. 

Humans, on the other hand, can sometimes follow general principles without needing to construct a full deal. That's not always a good idea. Sometimes general principles fail. But here this approach works just fine. You can see that if you play low, you deprive declarer of a late dummy entry. That's often a good idea even if you can't see how it might help. So playing low should be routine for an expert human.

Robots would play better if they started with their deal-generating approach, then followed general principles if they thought their play didn't matter. But where should these general principles come from? A few years ago, the state of the art required deriving rules by interviewing human experts. This was always an unreliable process, since experts have often internalized their expertise and aren't consciously aware of why they do things. Now, with the advancements in generative AI, the technology exists to allow robots to generate their own rules. Using this approach, I suspect they will find rules we aren't even aware of. If anyone succeeds in doing this, we will see an astonishing improvement in how robots play.

Postscript:

In the above write-up, I stated "I'm worth driving to game" without comment, since I didn't think that would be open to question. But when I posted the hand to the Bridge People group on Facebook, the majority of respondents simply made a game try. So I suppose I should justify my statement.

The lazy way to decide what the hand is worth is simply to count points. You have 19 total points: 17 HCP plus one for the fifth heart and one for the sixth. Marshall Miles, in How to Win at Duplicate Bridge, suggested that, when partner raises your suit, you should add two points for holding six cards in your suit, two more for holding seven, one for a side singleton, and two for a side void. I've found this method of valuation to be quite accurate. Under this method, the hand is worth three additional points, bringing the total to 22. One normally makes a game try with 17 or 18 and drives to game with 19. So 22 is more than enough to bid game.

A more accurate way to decide what the hand is worth is to give partner a perfect minimum, say, two kings, three trumps, and a useful doubleton. If that gives you a virtually laydown game, then you should invite. If partner accepts with a maximum, perhaps the perfect minimum with be included in that maximum. If not, then perhaps game won't be cold but will still have decent chances. With this hand, all you need for game is  king doubleton of diamonds. That's a full king less than a game try by this method, so the hand is worth driving to game.