The Aces on Bridge: Thursday, June 6th, 2013
Real stupidity beats artificial intelligence every time.
Terry Pratchett
West | North |
---|---|
East-West | ♠ A 3 ♥ 7 6 ♦ Q 8 7 4 ♣ A K Q 6 3 |
West | East |
---|---|
♠ J 10 9 7 5 ♥ J 9 3 2 ♦ — ♣ J 9 8 4 |
♠ K 8 6 4 ♥ 10 ♦ J 9 6 5 ♣ 10 7 5 2 |
South |
---|
♠ Q 2 ♥ A K Q 8 5 4 ♦ A K 10 3 2 ♣ — |
South | West | North | East |
---|---|---|---|
2♣* | Dbl. | 3♠ | |
4♠ | Pass | 4 NT | Pass |
6♦ | Pass | 7♦ | All pass |
*Weak, both majors
♠J
The humiliation of losing to a machine would never befall a top-class bridge player, would it? Playing bridge had previously seemed to be an insurmountable task for artificial intelligence. Until a decade ago, no program had proved capable of offering more than mediocre opposition to a decent player. But things may be about to change. Onno Eskes of the Netherlands has written in IMP magazine about how in the right circumstances computers can defeat humans. Over the next few months, I will be running some deals that reflect their strengths.
In seven diamonds (reached after West had shown the majors and a weak hand), two expert Dutch declarers went wrong. They won the spade lead and led a diamond to the ace, West discarding a spade. With hearts not breaking, declarer could not take advantage of the favorable lie of the clubs and went one down — for a flat board.
Eskes decided to present the hand to GIB, a bridge program developed by Matt Ginsberg, which can process the information about West’s hand plus the opening lead.
After 30 seconds it produced the spade ace, then, after a pause, it played the club ace. The computer discarded the spade queen and ruffed a small club in hand. Next came the diamond ace (discovering the bad trump split), a diamond to the queen, the two top clubs and the master club six. East ruffed, South overruffed, and he could now draw the last trump and ruff his last heart in dummy.
You are worth a second call, but the danger of bidding three clubs is obvious; no one likes playing in a 5-1 fit. The way to get spades into play would be to double. By contrast, a call of two no-trump is not natural but shows minors, with longer clubs of course. Your sequence suggests this precise pattern, plus extras.
BID WITH THE ACES
♠ A 3 ♥ 7 6 ♦ Q 8 7 4 ♣ A K Q 6 3 |
South | West | North | East |
---|---|---|---|
1♥ | |||
2♣ | 2♥ | Pass | Pass |
? |
Hi Bobby,
Interesting but isn’t ruffing a club at T2 a better shot, as West might have 5521 or 5431 shape. Then cash the DA getting the news, spade to ace and start running clubs. I agree that 7-1 clubs are unlikely (but so are 4-0 trumps) but west is known to have 9+ cards in majors so perhaps not so unlikely, especially if he has 10 cds in majors and has fortunately not led a heart from 5.
Unless D are 4-0 or hearts 5-0 the hand is easy. The computer line gets the extra chance of a 44 club break but walks into an overruff when west has 1 club. The alternative C ruff at T2 gains in both cases.
Regards,
Iain
Hi Iain
The spade Ace was played at trick 1.
Regards
Yasser
Hi Yasser,
Thanks I’d just realised my goof so my ine should be SA, club ruff, DA, D to Q, CA and so on. Brain frazzled as I had a tire blow out on the way to work this morning! Low speed though.
Regards,
Iain
Hi Iain and Yasser,
Yes, Iain your mind seems one step up on the computer (not an easy feat) and yes Yasser, someone needs to stand guard at the gate to make us all better (especially me when I goof), but do you two agree with me, that it appears the computer programming by Matt Ginsberg, is likely (at least sometimes) done on a results basis with advance knowledge of all four hands.
Since I know so little by how computers are programmed, I am wildly guessing to the above statement, but what Iain has taught us on this hand, is just a safer way to play it, only unimportant, because of the specific hand layout.
However, ever since I, long ago, saw the unique and futuristic movie, “2001, A Space Odyssey” I have been very respectful of computers for fear of Hal’s retaliation.
Hi Bobby,
I would guess that the program only sees the cards revealed to date (and the bidding inferences) or life would be too easy. Although the program’s line could be improved a bit, I suspect Matt Ginsberg has come up with something quite impressive. Computers haven’t replaced chess players (although the best are better via sheer number-crunching) but can greastly aid human players. I think something similar may happen at bridge.
Regards,
Iain