[Bbradley62]

I think these Norths need to re-examine their hands...

---

---

[Bbradley62]

At the first table... if 2♥ is "limit raise or better", and North is an unpassed hand, shouldn't the total points be 10+ rather than 10-12?

[barmar]

The book bid in both cases is the competitive raise. However, since we're promising 4-card support with this bid, the rule for raising with only 3 requires a simulation before it can be chosen.

In the first auction, in my 11 simulation runs, I got 8 Passes, 2 3♣, and 1 2♥. Then, for some reason when it was simulating what to do over 2NT, Pass wasn't even one of the options it considered. It was choosing between 3♣, 3♦ (the book bid), and 3NT. On most hands, it didn't matter which it bid, it ended up in 3NT (since it gave South 15-17 HCP, and North had "promised" 10+, so South will always bid game); there was one hand where it mattered, and 3NT went down the least, so that was the choice it went with.

In the second auction, I got 8 3♦, 2 Passes, and 1 2♥. North's length in the majors makes it more likely that South has 5+ ♦, and almost certain that it's at least 4. The quality of the ♦ support is probably what swung the sims in the direction of raising, rather than 2♥ as in the first auction.

[Bbradley62]

barmar, on 2012-June-12, 13:37, said:

The book bid in both cases is the competitive raise. However, since we're promising 4-card support with this bid, the rule for raising with only 3 requires a simulation before it can be chosen.

GIB should not be allowed the possibility of raising on 3-card support if partner might also only have three.

barmar, on 2012-June-12, 13:37, said:

In the first auction, in my 11 simulation runs, I got 8 Passes, 2 3♣, and 1 2♥.

So, simulations allow for making a bid that shows 3-5 total points more than GIB actually has? I think this is bad.

barmar, on 2012-June-12, 13:37, said:

Then, for some reason when it was simulating what to do over 2NT, Pass wasn't even one of the options it considered. It was choosing between 3♣, 3♦ (the book bid), and 3NT. On most hands, it didn't matter which it bid, it ended up in 3NT (since it gave South 15-17 HCP, and North had "promised" 10+, so South will always bid game); there was one hand where it mattered, and 3NT went down the least, so that was the choice it went with.

Why does 2NT show 15-17 instead of 12-14?

barmar, on 2012-June-12, 13:37, said:

In the second auction, I got 8 3♦, 2 Passes, and 1 2♥. North's length in the majors makes it more likely that South has 5+ ♦, and almost certain that it's at least 4. The quality of the ♦ support is probably what swung the sims in the direction of raising, rather than 2♥ as in the first auction.

I don't understand your differentiation of "raising, rather than 2♥ as in the first auction". 2♥ in the first auction is raising partner's opening minor.

[cloa513]

barmar, on 2012-June-12, 13:37, said:

The book bid in both cases is the competitive raise. However, since we're promising 4-card support with this bid, the rule for raising with only 3 requires a simulation before it can be chosen.

In the first auction, in my 11 simulation runs, I got 8 Passes, 2 3♣, and 1 2♥. Then, for some reason when it was simulating what to do over 2NT, Pass wasn't even one of the options it considered. It was choosing between 3♣, 3♦ (the book bid), and 3NT. On most hands, it didn't matter which it bid, it ended up in 3NT (since it gave South 15-17 HCP, and North had "promised" 10+, so South will always bid game); there was one hand where it mattered, and 3NT went down the least, so that was the choice it went with.

In the second auction, I got 8 3♦, 2 Passes, and 1 2♥. North's length in the majors makes it more likely that South has 5+ ♦, and almost certain that it's at least 4. The quality of the ♦ support is probably what swung the sims in the direction of raising, rather than 2♥ as in the first auction.

That's why I say the forward chain logic of GIB is ludicrous- it cannot possibly know what is in partner's or opponents` hands with almost any amount of bidding so the chains just lead it to the wrong place. It should always use bookbids, if not available then a simulation with that bid as the final contract and simple DD analysis will allow it get the biggest possible sample of hands. I see it simulated 11 hands out of millions.

[Bbradley62]

cloa513, on 2012-June-12, 15:26, said:

I see it simulated 11 hands out of millions.

No, I think you misunderstand this. When Barmar re-runs hands to see where simulations might lead, his program runs "a bunch of simulations" 11 separate times; Barmar therefore reports a single decision that was based on each of those 11 "bunches". I don't recall ever hearing how many hands are "dealt" in each "bunch".