Robots overrate distribution for NT contracts

[CoreyCole]

On two or three occasions in the last month, I have been delicately searching for the best contract, only to have the robot jump directly to 6NT with a clearly inappropriate hand. For example, yesterday the robot had Q1052 / AQ63 / KQ532 / VOID and I had A7 / J98 / A94 / AKQ54. The auction (I dealt) went 1C - 1D - 2NT - 3H - 3NT - 6NT.

 

While the robot can rightfully get excited about my 2NT call with his distributional hand, 3NT is a signoff suggesting a misfit. If the robot is going to jump to slam, the call should be 6D, showing an extremely distributional hand, rather than 6NT. But 4D (clearly forcing) is probably better, as it gives me a chance to show delayed support for one of his suits. 6NT is unilateral and can't be right.

 

The only pairs to play this in 3NT lied about their points by rebidding 1NT over 1D, a terrible call in my opinion.

 

In general, I find that the robots regularly count distributional points in No Trump auctions. This seems like an evaluation error. You can add some long suit points for No Trump, but the robots should also deduct for very short suits (singletons and voids) to end up with a more reasonable total. We could easily have been off two Aces - and possibly a runnable club suit - on the auction I gave.

[BunnyGo]

In the future providing the link helps the programmers check for bugs. You can do it by going to my hands.

[georgi]

Here the points which GIB promise are not TP but HCP, so no extra points added.

 

4♦ should promise 6+♦ as GIB already promised 5+♦ with 3♥.

 

4NT after 3NT is 12-13HCP, 6NT as displayed for GIB is 13-17HCP, so it's about to pick between 4♦,4NT,6NT. Probably Pass too, but 3NT is not sign-off, just you're explaining your hand. Denying 4th majors ( the explanation needs to be corrected though for 3NT ) is the only info 3NT should provide.

 

The lack of two aces seems like issue for this sequence although GIB had 14HCP which makes 32+HCP for sure for the pair.

[barmar]

In the future providing the link helps the programmers check for bugs. You can do it by going to my hands.

Since all we need is the robot's hand and the auction, the information he provided is just as good as a link to the actual hand. Although a link ensures that he doesn't make a mistake when transcribing.

[barmar]

The lack of two aces seems like issue for this sequence although GIB had 14HCP which makes 32+HCP for sure for the pair.

GIB only had 13 HCP.

 

However, GIB thinks that 31 HCP is enough for 6NT. And while it knows how to answer Gerber, it never uses it itself -- with enough points, it just crosses its silicon fingers and blasts.

 

And if my partner jumped to 6♦, I'd expect a better suit than KQxxx. GIB thinks this shows 6+.

[xxhong]

Well, the total HCP requirement for 6NT is wrong, which has been reported again and again. If no fit is discovered, gib should bid 6NT with 33 HCPs, not 31. Of course, strong suits can make that requirement lower, which is a hand evaluation question, but it is just a nonsense and losing bridge to push every 31 HCP misfit hand to 6NT.

 

Here the points which GIB promise are not TP but HCP, so no extra points added.

 

4♦ should promise 6+♦ as GIB already promised 5+♦ with 3♥.

 

4NT after 3NT is 12-13HCP, 6NT as displayed for GIB is 13-17HCP, so it's about to pick between 4♦,4NT,6NT. Probably Pass too, but 3NT is not sign-off, just you're explaining your hand. Denying 4th majors ( the explanation needs to be corrected though for 3NT ) is the only info 3NT should provide.

 

The lack of two aces seems like issue for this sequence although GIB had 14HCP which makes 32+HCP for sure for the pair.

[barmar]

I asked Uday and Fred about this. They tried bumping up the 6NT minimum a while ago, but it had undesirable side effects. The bidding rules are a complex system, with many inter-dependencies that are not easy to detect.