[Bbradley62]

These were back-to-back hands in a GIB tourney.

---

When I bid 3N over nmf, I am almost certainly 3235. Why doesn't GIB leave me in 3NT? At one table, the auction was 1N-2C-2D-3S*-3N and GIB left it there, even though that GIB had more to fear in the club suit than my GIB did. Or was my GIB trying to get to 6♦?

---

Why doesn't GIB use nmf here like he did in the first? I guess I could have bid 3♠ over 3♥, but I was just trying to sign off.

[barmar]

Notice also that the description of 3NT includes "2-3 H, 2-4 S". Shouldn't it actually be "2 H, 2-3 S", since you didn't bid 3♥ or 3♠, which would show 3 and 4 cards in the respective suits. And in practice it should be 3=2=3=5 as you say, since with 2=2=4=5 you'd reverse rather than bidding 2NT, and 3=2=4=4 would open 1♦.

[cloa513]

Bbradley62, on 2011-February-14, 23:50, said:

These were back-to-back hands in a GIB tourney.

---

When I bid 3N over nmf, I am almost certainly 3235. Why doesn't GIB leave me in 3NT? At one table, the auction was 1N-2C-2D-3S*-3N and GIB left it there, even though that GIB had more to fear in the club suit than my GIB did. Or was my GIB trying to get to 6♦?

---

Why doesn't GIB use nmf here like he did in the first? I guess I could have bid 3♠ over 3♥, but I was just trying to sign off.

Better logically to have responded 2C and 1S on those hands followed by 3NT- although examples have come of GIB pressing on to 4H anyway.

[Bbradley62]

cloa513, on 2011-February-15, 03:00, said:

Better logically to have responded 2C and 1S on those hands followed by 3NT- although examples have come of GIB pressing on to 4H anyway.

I can see responding 1S on the second one, although I'd expect to have trouble defining my strength. But, it's absurd to respond 2C on the first hand.

[xxhong]

The basic algorithm of gib bidding is not to play NT when possible. That's due to two reason I think. One, the original meadowlark bidding database is very badly designed, which always require a few HCP more in your range to bid 3NT. Two, gib does double dummy analysis on a very limited sample size. Therefore, it often finds 3NT difficult after the perfect defense and gib rarely sets the constraints of the sample generating algorithm correct. This really needs a lot of work and I don't think bbo has the resource to improve it by a lot. Also, this is restricted by the computer performance. Of course, an ideal solution would be to buy a lot of high performance computer clusters and doing both double dummy analysis and single dummy analysis. Still, for now, I think there are ways to improve. The algorithm should be set to disallow removing from 3NT in most situations for non-freakish hands.