This article was originally published in  vol. 8, nr. 4, June 1997.

The GIB Zone (3) 
  by Onno Eskes

First try these three problem deals.
 
1. W/both K 5 
A Q 5 
A K 3 2 
8 7 5 3
- -------
e A 8 6 3 
8 3 2 
8 6 5 4 
9 6
 
 WEST 
  21) 
  Pass 
  Pass 
 
 1) Flannery, 
    11-15 hcp, 
  5and 4 
NORTH --- 
Double 
Pass 
Pass
EAST ---- 
Pass 
Double
SOUTH 
3 
Pass
West plays §A-K-Q and East follows suit with the 2, 8, and J (high-low even). Can the hand still be made?
 
 
2. E/NS Q 6 5 
Q J 10 6 2 

A Q 9 4
- -------
e A K 10 9 2 
A 5 
A K J 
10 6 3
 
----  West leads the 9 against your contract of 6. How do you play the hand?
 
 
3. E/NS A J 10 8 6 2 

K Q 6 2 
Q 10 2
- -------
e 7 2 

A J 9 8 7 5 
A 6 5 4
 
 WEST--  
  - 
  4 
  5 
  Pass
NORTH -- 
- 
5 
5 
Pass
EAST ---- 
1 
Pass 
Double 
Pass
SOUTH 
2 
Pass 
6
Nicely bid by North (Jean Harings) who called 5 rather than 4 in anticipation of a 5-bid, and subsequently completed the picture with 5. Once East doubles, you decide to run to 6. West tables the Q. Plan your play.
 

Deep Blue

Deep Blue has defeated world champion chess, Gary Kasparov, in an official head-to-head encounter. In IMP we continue exploring the limits of GIB, the bridge program that's expected to follow in Deep Blue's footsteps by beating world champions within the next five years.

1. The first problem hand is virtually an open book following West's Flannery opening, the penalty double, and East's signals:
 
 

1. W/both K 5 
A Q 5 
A K 3 2 
8 7 5 3
10 9 4 2 
K J 9 6 4 

A K Q 10
Q J 7 
10 7 
Q J 10 9 7 
J 4 2
e A 8 6 3 
8 3 2 
8 6 5 4 
9 6
 
--- South plays 3 doubled. After West has cashed two club tricks it's not clear how you can prevent East from taking three trump tricks. But for GIB these types of problems are a cinch. He ruffs the third round of clubs, takes the heart finesse, and ruffs a spade in dummy. The A is cashed followed by the last club from North, in this position:  
 
1. W/both

A K 3 
8
10 
K J 9 

10


Q J 10 9 7 
-
e

8 6 5 
-
 
--- East ruffs high to prevent the 8 from becoming the ninth trick, while declarer discards the 8. East is now forced to exit with a high diamond, taken by North's ace. Again, East is forced to play a high trump on the lead of dummy's heart, and now he is endplayed, having to lead from 9-7 into the leader's K-3 opposite 8-6.  
 
 
2. E/NS Q 6 5 
Q J 10 6 2 

A Q 9 4
J 8 3 
9 3 
10 8 6 5 3 
8 5 2
7 4 
K 8 7 4 
Q 9 4 2 
K J 7
e A K 10 9 2 
A 5 
A K J 
10 6 3
 
----  West leads the 9 against your contract of 6. How do you play the hand? 
This second deal was submitted by Dick Ippel. The winning line in 6 was missed at the table in the Haarlem premier division. The 9 was covered by the ten while East ducked, after which declarer put his money on the double finesse in clubs. Dick predicted GIB would take the superior approach of refusing to cover the lead, and so it turned out: 9 taken by the ace, pull trumps, set up the hearts for three pitches. Should the hearts break worse than four-two, then the club finesse remains as a last resort. Case closed. Next? 
Not quite. The machine is set straight by Erik Kirchhoff. GIB had been guilty of a typically human trait: he didn't take account of a possible bad trump split. The Kirchhoff plan is to play a high heart from dummy, followed by another heart to the ace. If East-West follow suit to the second round, then four trumps to the Jack with West is the only remaining problem, with the club finesse in reserve. But GIB would have gone down, even with that finesse on. If West trumps the second heart, then, like GIB, we must rely on the club hook. If it turns out that West has ruffed from four trumps (i.e. GIB goes down), then Erik will fall back on a diamond finesse for his contract. A fascinating hand.

This problem does, however, show up one of the whims of GIB. A coincidence? I had noticed something similar before. It concerned the infamous hand played by Jimmy Cayne from this year's Vanderbilt final, and reported by Larry Cohen in the previous issue of IMP.
 

- A K 10 
A J 
A K J 10 6 5 
A 6
- -------------
e 5 3 
K Q 7 6 3 
9 4 
K Q 10 7
 
----  It's almost incredible that someone playing in the finals of one of the toughest events in the world managed to screw up this standard combination in diamonds. Finessing twice is four times as likely to succeed as playing a top diamond first (any small singleton over dummy versus singleton queen). "Piece of cake for GIB," I'm thinking, but when I submit the problem, GIB, too, starts by cashing the ace. I jack up the time allotment for 'tanking', but to no avail: again the ace. I next give him the maximum time available, and this time he first plays the king rather than the ace. Sorely disappointed, I e-mail Matt Ginsberg, GIB's creator, "Please review your safety plays, and teach GIB a lesson as well!" Pretty soon I receive a reply - a crash course in GIB. 
The Delphic Oracle: how does GIB work?

The heart of GIB is a superfast double dummy solver. He only gets to see declarer's hand and dummy, and then he generates hundreds of random hands for the defenders at breakneck speed. For each of these opposing hands he determines which card is the best one to play. Whichever card is correct most often is then played. Simple, elegant, and with amazingly good results.

Yet, the method is not quite foolproof. Take, for example, the 6-contract, with the trumps Q65 opposite  AK1092. GIB plays 6 and doesn't seem to worry about a trump-loser. "That's right," Ginsberg explains, "because GIB never has a trump-loser (if it's avoidable)!" GIB plays the hand, double dummy, a couple of hundred times, and each time a possible Jxxx is picked up without fail. And without a trump-loser one can indeed refuse the heart finesse at trick one at no cost. Only when GIB actually gets around to playing trumps does he discover that playing single dummy is rather more difficult than double dummy.

Jimmy Cayne's 7-contract is a similar case. GIB consistently starts with A, not just to cater to singleton Q behind the A-K-J, but also to take care of doubleton queen behind! Keep that in mind that GIB plays a hand several hundred times double dummy. Each time there's doubleton queen behind, he picks it up, and consequently, starting with the ace has the best chance of success as far as he's concerned. Only in the second instance is he forced to choose between Qx behind or Qxx in front, and only then does he realize that he must take the finesse (while also realizing only then that he should have discounted the chance of finding Qx in the first place!).

Disillusion
I reckoned I had played the deal below rather well. Following an elegant bidding sequence I was declarer in 6.
 

3. E/NS A J 10 8 6 2 

K Q 6 2 
Q 10 2

Q J 9 7 6 3 2 
10 3 
J 9 7
Q 9 5 4 
A K 10 8 4 

K 8 3
e 7 2 

A J 9 8 7 5 
A 6 5 4
 
 WEST 
 
- 
  4 
  5 
  Pass 
 

Lead: Q

NORTH -- 
Harings 
- 
5 
5 
Pass
EAST ---- 
- 
1 
Pass 
Double 
Pass
SOUTH 
Eskes 
2 
Pass 
6
The spade suit has to be set up for a few club pitches, and if it's divided 3-2 there'll be no problem. But if the spades are 4-1 (which is not unlikely, considering the double) it would seem there are insufficient entries after you have pulled trumps. Therefore, I ruffed the opening heart lead, played the A, and when West's king dropped, I crossed to hand in trumps in order to play a small spade toward the J-10. If West ruffs, the queen can be ruffed out later, and the high spades cashed, as there are still two diamond entries remaining. Should West decline to ruff, as happened at the table, then East will score his queen, but a third spade can be ruffed high in hand, after which the spades are high and the trumps may be pulled.

"Not bad," I thought to myself. Would GIB find this line also? I give him the hand: heart lead ruffed, king of diamonds, diamond to the ace. "Aha, at last that infernal machine plays more asinine than I do," I notice with glee. GIB next plays a small spade, West the king, and GIB?.......small from dummy!

That's how simple it really was.
-

 Do you have an interesting problem in declarer play which you would like to submit to GIB? 

Send the hand (including all details) to Bridge Magazine IMP, or directly to Onno Eskes: eskeso@xs4all.nl 


  To IMP Bridge home page