[pescetom]

Would you choose the same lead as the robot?

the board

I can see that if one cannot understand explanations and is also certain that opponent must have Jxxx the chosen lead has sense. But...

[johnu]

In GIB 2/1, 3NT is a strong balanced 25-27 HCP, not a gambling 3NT with an outside stopper.

That being said, if North has 25-27 HCP, East has 12 HCP, then West has at most a king, and as little as a jack (or nothing) (same for South). It's hard to see how ♥J is going to be the winning lead. If you are going to beat 3NT, the only likely source of tricks is clubs, or that declarer just doesn't have 9 tricks. A quick simulation of 100 deals showed that 3NT would be down 92% on the lead of a high club (sometimes N/S have 4/5 clubs to the jack so there are only 3 top club tricks but there are still only 8 winners available, otherwise clubs run). Without a high club lead, N/S can usually scramble 9+ tricks about 80% of the time. So GIB's ♥J lead makes no sense.

I would expect a lead poll to be 100% for a high club lead (I would lead the Queen asking for the unblock of the jack).

[smerriman]

While I get similar percentages to johnu, they're not the most relevant ones when it comes to what lead GIB makes. Over the 500 deals I ran, my sim says that a heart was only better than a club once, but was equal about 10% of the time. Given dealer.php struggles to sim hands that match the 3NT bid in the first place (after running for 5 seconds, it only finds about 100 in 10 million attempts, and GIB plays much faster than this), it's not entirely impossible that in the small number of matches it found, all of them were the 'equals' case and thus it chose a lead at random.

But, more likely, it's probably just GIB being GIB again, where nobody knows why it's broken.

[johnu]

smerriman, on 2022-May-29, 20:16, said:

Given dealer.php struggles to sim hands that match the 3NT bid in the first place (after running for 5 seconds, it only finds about 100 in 10 million attempts, and GIB plays much faster than this), it's not entirely impossible that in the small number of matches it found, all of them were the 'equals' case and thus it chose a lead at random.

Matching hands are very rare indeed. That's why I only did 100 hands. I stopped and reduced the number when I saw how slowly hands were being found (Dealmaster Pro). I didn't expect the results to show anything except the overwhelming superiority of the high club lead rather than being a close decision.

[pescetom]

smerriman, on 2022-May-29, 20:16, said:

Given dealer.php struggles to sim hands that match the 3NT bid in the first place (after running for 5 seconds, it only finds about 100 in 10 million attempts, and GIB plays much faster than this), it's not entirely impossible that in the small number of matches it found, all of them were the 'equals' case and thus it chose a lead at random.

Yes that is logical and something I forgot to figure in.

I guess my underlying questions were more:
1) does GIB assume 3NT = NAT Bal 25+ as johnu affirms (logical, but not disclosed in the System Notes)?
2) should it blindly trust human-human opponents to agree the same (and never deviate either) ?

[johnu]

pescetom, on 2022-May-30, 12:50, said:

1) does GIB assume 3NT = NAT Bal 25+ as johnu affirms (logical, but not disclosed in the System Notes)?

There have been other posts about the meaning of 3NT in the past, but I confirmed this in a free GIB game (and GIB partner raised to 7NT with an opening bid).

[pigpenz]

johnu, on 2022-May-29, 17:10, said:

In GIB 2/1, 3NT is a strong balanced 25-27 HCP, not a gambling 3NT with an outside stopper.

That being said, if North has 25-27 HCP, East has 12 HCP, then West has at most a king, and as little as a jack (or nothing) (same for South). It's hard to see how ♥J is going to be the winning lead. If you are going to beat 3NT, the only likely source of tricks is clubs, or that declarer just doesn't have 9 tricks. A quick simulation of 100 deals showed that 3NT would be down 92% on the lead of a high club (sometimes N/S have 4/5 clubs to the jack so there are only 3 top club tricks but there are still only 8 winners available, otherwise clubs run). Without a high club lead, N/S can usually scramble 9+ tricks about 80% of the time. So GIB's ♥J lead makes no sense.

I would expect a lead poll to be 100% for a high club lead (I would lead the Queen asking for the unblock of the jack).

John I thought A lead also asks for unblock or count, Q does ask for J but i thought only in context of holding KQt9