| ♠ K 7 6 4 3 2 ♥ K 6 2 ♦ A Q ♣ A K |
I open one spade in first seat--pass--one notrump--pass back to me. I don't have many options here. I'm not about to jump shift in a doubleton in the interest of maintaining flexibility. Three notrump looks like a reasonable contract, so I bid it. West leads the deuce of diamonds.
| NORTH ♠ K 7 6 4 3 2 ♥ K 6 2 ♦ A Q ♣ A K | ||
| SOUTH ♠ A ♥ 7 4 3 ♦ K 8 6 3 ♣ Q 10 9 7 3 |
| West | North | East | South |
| 1 ♠ | Pass | 1 NT | |
| Pass | 3 NT | (All pass) |
I have eight tricks, although my entries are tangled up so that I can't cash them all. To come to nine tricks, it seems I will need either the jack of clubs to drop in three rounds or to find the heart ace onside. Perhaps, if the jack of clubs doesn't drop, I can devise some kind of endplay to avoid having to lead up to the king of hearts, but I can work that out later. I rise with the diamond ace. East plays the four. I play the six, since the three would let West know that East's four was his lowest. I cash the ace and king of clubs. East plays eight, four; West plays deuce, jack.
Too bad. I would have better a chance of picking up some IMPs if this hand were harder to make. What are my prospects for overtricks? I have two ways to cash my nine tricks. I could (A) cash the diamond queen, play a spade to my ace, and run cash my minor suit tricks, leaving the spade king stranded; or I could (B) play a spade to the ace, cash clubs, then cross to the diamond queen and cash the spade king, leaving the diamond king stranded.
(A) looks more straight-forward, so I'll try (B). I play a spade--five--ace--eight. It appears spades are three-three, giving West a 3-4-4-2 pattern and giving East 3-3-3-4. I cash two clubs, pitching spades from dummy. West plays the nine and ten of spades as East follows. On the last club, West pitches the jack of spades. I wasn't expecting that. Dummy's spades are now good. Why did I pitch so many of them? I pitch a heart on this trick, then lead a diamond to the queen to cash dummy's three remaining spades. Making five.
| NORTH ♠ K 7 6 4 3 2 ♥ K 6 2 ♦ A Q ♣ A K | ||
| WEST ♠ Q J 10 9 8 ♥ 10 8 ♦ 10 9 7 2 ♣ J 2 | EAST ♠ 5 ♥ A Q J 9 5 ♦ J 5 4 ♣ 8 6 5 4 | |
| SOUTH ♠ A ♥ 7 4 3 ♦ K 8 6 3 ♣ Q 10 9 7 3 |
I guess the six of diamonds at trick one paid off. West would not have pitched all his spades had he known his partner didn't have the diamond king. Of course, he doesn't really need a signal to know that. He should know I have the diamond king because I didn't finesse at trick one. But Jack doesn't seem to draw inferences from declarer's line of play.
How should the opponents defend? This is the position we reached on the play of the last club:
| NORTH ♠ K 7 6 ♥ K 6 2 ♦ Q ♣ -- | ||
| WEST ♠ Q J ♥ 10 8 ♦ 10 9 7 ♣ -- | EAST ♠ -- ♥ A Q J 9 5 ♦ J 5 ♣ -- | |
| SOUTH ♠ -- ♥ 7 4 3 ♦ K 8 3 ♣ 9 |
Could it cost for West to pitch a diamond? Not if I've played sensibly. But if I had jack third of diamonds remaining, it would actually cost the contract. After a diamond pitch I could play a diamond to East's king. East would now have a choice of leading hearts, giving me two tricks in dummy, or leading a diamond, giving me two tricks in my hand.
Could it cost to pitch a heart? Yes. It costs on overtrick on the actual hand. I pitch a spade from dummy, play a diamond to the queen, cash the spade king, and play a heart. East is winkled. He must allow me to score one red king or the other. (That's why I kept dummy's hearts. I knew it looked like a good idea.) The only way the defense can hold me to three is for West to pitch a diamond so that, in the end position, he can overtake his partner's nine of hearts with the ten and return one.
At the other table, North bids four spades over one notrump. Once again we have an example of the difficulties computers face playing bridge. Jack defines three notrump as 18-20 HCP and 5-3-3-2 distribution (though, personally, I think 18 is worth only two notrump). I'm quite sure Jack "knows" that three notrump rates to be a better game than four spades opposite a random one notrump response, so I suspect he bid four spades because, with six spades, three notrump would be a violation of system. If computers are to become expert players, they must learn to distinguish between serious system violations and white lies calculated to improve your chance of reaching the right contract.
East and South pass over four spades, and West, looking at three natural trump tricks, doubles. East leads the five of clubs. Declarer wins and starts cashing minor-suit tricks. He manages to pitch two hearts before West can ruff in, holding his losses to three trump tricks and the ace of hearts. Down one.
Me: +660
Jack: -200
Score on Board 63: +13 IMPs
Total: +157 IMPs
Umm.... +660 vs. -200 is win 13 IMP.
ReplyDeleteOddly, GIB goes overboard on the sanctity of system in the other direction. It seems more than willing to violate system just to play the hand. Or at least it seems that way. But then, perhaps I am projecting the shades of partners past into the situation.
Seriously, earlier versions of GIB would transfer into a 4-card suit when you opened 1NT. As I noted in an earlier post to this thread, Jack on the demo level bid 1C-1NT, 2C with 3=4=2=4. And GIB is indeed reluctant to sit for 3NT, preferring to play in shaky trump fits (often choosing 4M with Jxxxx when I have denied 3-card support for example).
But your point about computers being unable to draw inferences from opponents' plays is very valid. Imagine trying to program that! Assume the software makes decisions based on 100 random deals, and complete double-dummy analysis. Now, for each of those 100 deals, it has to generate 100 random deals for the hypothetical opponent's holding and DD those. So the number of deals to analyze goes from 100 to 10,100, and the time required goes from a few seconds to five minutes. The directors are already hovering. (But I guess you know about that! You wrote one of the funniest pieces I ever read in the Bridge World, where you had a two-way guess where playing either opponent for the key card assumed they had made a complete nullo play earlier.)
Thanks for the scoring correction. I never could keep score. As for how to write a computer program to draw inferences from the play, I think it's even harder than you indicate. Even if computers were infinitely fast, I'm not sure you could write such a program. Declarer's correct play depends on how you would defend in various circumstances. How you would defend depends on how declarer would play. It seems to me any algorithmic method leads to an infinite regression. The way we humans escape this regression is by relying on experience. We know declarer would play a hand certain way not because we can prove it's how he should play but because experience has taught us that's what declarers do. Try this one for example: In notrump, you have A J 5 4 over dummy's 10 3. Partner, who is known to have four, leads the deuce. You win with the ace. Do you return low or the jack? Don't solve it. Just write a set of instructions describing how to solve it.
ReplyDeleteYes, I see your point. The Millennial celebration TYP deals would blow any software's "mind." I think a computer would get your example right, though - it's right to play the Jack unless declarer has exactly Q9, or unless you will need a late entry in the suit. Q9 encompasses original holdings of Q98, Q97, or Q96, so it is 3 of declarer's original 35 possible holdings.
ReplyDeleteSo the need to lead low is highly likely to be driven by the entry situation. This would be a very difficult concept for a computer to cope with.
The point I was trying to make is this: If you conclude that it's right to lead the jack (the apparent percentage play), then, whenever you do lead low in this position, declarer should play the queen from Qxx or from Q9x. But if declarer is going to do that it, then the jack is no longer the percentage play. You should lead low. But if you do that, then hopping with the queen is no longer declarer's percentage play. He should duck. But if declarer's going to do that... How does a computer ever finish its analysis? You'll find a fuller analysis of this combination here: http://sites.google.com/site/psmartinsite/Home/bridge-articles/a-suit-combination
ReplyDelete