An example hand -- and I'm sure BBO could easily find many more -- is given here (click)
My problem/request is that GIB should be able to differentiate between good Texas bids and bad ones. If I (South) unilaterally invoke RKCB, it becomes a lottery.
If, instead, GIB could show uninspiring hands thru Texas (1NT-4♥; 4♠) and invitational/good hands thru Jacoby (1NT-2♥; 2♠-4♠) that would make life much easier for South to guess when to try for slam.
In the example hand above, 4♠+2 was only 33%.
Page 1 of 1
Robot's Texas Transfers Any way to refine definitions?
#2
Posted 2017-July-09, 08:07
The robots do play that 1NT-2H-2S-4S is a slam try. It took you having a super-max (and probably worth an upgrade) to make 6.
Wayne Somerville
#3
Posted 2017-July-09, 08:47
I think I knew that robots do play 1NT-2H-2S-4S as a slam try. And my example may not have been best to make my point; it was just the most recent I came across.
My point is that perhaps GIB developers can evaluate whether the "slam-try" logic works effectively. I have a sense that bots often use Texas transfer and then pass. As a hypothesis, perhaps the GIB simulation presumes a 2-card fit with 1NT opener -- which is why it overwhelmingly uses Texas instead of Jacoby & game.
My point is that perhaps GIB developers can evaluate whether the "slam-try" logic works effectively. I have a sense that bots often use Texas transfer and then pass. As a hypothesis, perhaps the GIB simulation presumes a 2-card fit with 1NT opener -- which is why it overwhelmingly uses Texas instead of Jacoby & game.
#4
Posted 2017-July-09, 10:10
The hand given is 6-4 in the majors. So Stayman then if 2♦ jump to 3 as Smolen then if 3N-4M shows a 6-4 slam try.
Without 6-4 in another suit you can transfer and bid second suit. You could also self-splinter (not sure if Gib plays that.
With 6331 unless your playing self-splinters stuck with Jacoby transfer then jumping to 4M showing a mild slam try.
South African transfers allow what you want 4C=H, 4D=S then bidding the suit in between says opener likes the suit. Not many people play this.
Without 6-4 in another suit you can transfer and bid second suit. You could also self-splinter (not sure if Gib plays that.
With 6331 unless your playing self-splinters stuck with Jacoby transfer then jumping to 4M showing a mild slam try.
South African transfers allow what you want 4C=H, 4D=S then bidding the suit in between says opener likes the suit. Not many people play this.
Sarcasm is a state of mind
#5
Posted 2017-July-09, 13:47
manudude03, on 2017-July-09, 08:07, said:
The robots do play that 1NT-2H-2S-4S is a slam try. It took you having a super-max (and probably worth an upgrade) to make 6.
There are several other types of 1NT hands that may make slam worthwhile.
1) Nothing wasted in diamonds
2) ♦AK and nothing wasted in clubs
3) Make up your own example
Of course if you do have a super max, you don't have a chance to show the super max except by going past game. If you make a unilateral slam try, you deserve to find a dummy where even 4 is in doubt.
#6
Posted 2017-July-09, 13:50
shyams, on 2017-July-09, 08:47, said:
My point is that perhaps GIB developers can evaluate whether the "slam-try" logic works effectively. I have a sense that bots often use Texas transfer and then pass. As a hypothesis, perhaps the GIB simulation presumes a 2-card fit with 1NT opener -- which is why it overwhelmingly uses Texas instead of Jacoby & game.
If GIB uses simulations (IMO, this is a "book" bid based on fixed criteria), then it would model partner having a range of fit, from 2 to 5 card support.
Page 1 of 1