GIB's suit play

[TimG]

In slam with a trump suit of JT62 opposite AK74 (there was one outside loser, but transportation to both hands) GIB started with the Jack and, when next hand played low, went up with the Ace, dropping the stiff Queen.

 

Is this a case of thinking that 2nd hand will always cover with the Queen, so GIB changed course and "hoped" for Qx offside?

 

Tim

[barmar]

I've been wonder about things like this myself. I've noticed that GIB is a strong believer in covering honors, which is very helpful when you have a two-way finesse. With AJT opposite Kxx, lead the Jack, and if GIB doesn't cover, go up and finesse on the way back; this seems to work at least 90% of the time.

 

In general, when programming computers to play games, they have to assume that the opponent will use the same playing strategy that they use. So if GIB's algorithm says that covering is usually right, and the opponent doesn't cover, it has to "assume" that the opponent doesn't have the Queen.

[hotShot]

Be careful not to over-interpret GIBs actions.

GIB has no plan and no principles it follows.

 

It just generates a set of deals that fit his and dummies holding and the restrictions defined during bidding. It solves all those double dummy and looks for the best card to play in most of these deals.

 

Obviously GIBs simulations will be closer to the actual deal the more is known about distribution and honors. GIB plays best when everybody is bidding so that much is known about all hands from the start.

During the first tricks GIBs DD analysis takes much time so GIB can't check many deals and the randomly generated deals may be far away from the one played. This can lead to plays/bids that appear irrational.

During the last tricks GIB can handle lots of simulated deals and many facts are already known. The picture that GIB gets from his simulations is much closer to the actual deal and GIBs play gets close to perfect.

[barmar]

I realize it's not really "planning". But double dummy analysis assumes that the opponents will also play optimally. If that includes covering an honor, and the honor isn't covered, doesn't GIB update the restrictions to include the "fact" that there's no honor there to cover with?

[PetteriLem]

I dont want to repeat myself, but I remind you of an occaison, when GIB succesfully finessed a suit. The contract would have been set, if the finesse failed. Later GIB refused to finesse again to make the contract, but tried another line. I suspect that GIB blindly makes simulations without memory and good heuristics. It would interesting to know situations, when GIB actually places a certain card and later plays accordingly against the odds.

[PetteriLem]

I cant emphasize that it is one the most terrible moments I have had with GIB. Watch it playing a small slam, finesse a critical suit and succeed. Then later not to finesse again and fail miserably. Has anyone of you ever changed your mind like that? I admit, I cant think of situation where I would tell my team mates my brilliant play, which would mimic GIB's antics.

[barmar]

It's done that to me, too.

 

One of the things that GIB is missing is the notion of "if there's only one lie of the cards that allows the contract to make, play for it." It simply does its simulations, and takes the line with the highest expected payoff.