[aguahombre]

gordontd, on 2015-July-14, 03:43, said:

Maybe even first choice, if you held the card.

He meant "A" as an article, not Ace.... a low heart.

[Mbodell]

For some results, here was how the panel in MSC voted/rated:

Quote

Action  Score  Panel Votes  % of Solvers
S4       100       9              23
ST        80       6              24
D3        80       7              25
H3        60       4              19
C3        40       1               2
CA        20       0               3
other      0       0               4

At the time of this posts the bbf voting is scattered with multiple votes for both spades, diamond 3, and club A and single votes for others.

Based on my simulation (code here, you can check my assumptions on this auction restrictions) I got this result for simulation (imp score is compared to the ♣A lead):

Quote

ST lead yeilds 101528 tricks (10.1528), sets 2456 times (24.56%) with imps of -4963 (-0.4963).
S4 lead yeilds 101519 tricks (10.1519), sets 2461 times (24.61%) with imps of -4899 (-0.4899).
HK lead yeilds 103731 tricks (10.3731), sets 1934 times (19.34%) with imps of -13168 (-1.3168).
H8 lead yeilds 101785 tricks (10.1785), sets 2378 times (23.78%) with imps of -5991 (-0.5991).
H3 lead yeilds 101769 tricks (10.1769), sets 2388 times (23.88%) with imps of -5869 (-0.5869).
DJ lead yeilds 101784 tricks (10.1784), sets 2469 times (24.69%) with imps of -4909 (-0.4909).
D4 lead yeilds 101192 tricks (10.1192), sets 2671 times (26.71%) with imps of -2073 (-0.2073).
D3 lead yeilds 101192 tricks (10.1192), sets 2671 times (26.71%) with imps of -2073 (-0.2073).
CA lead yeilds 100077 tricks (10.0077), sets 2756 times (27.56%) with imps of 0 (0.0).
CT lead yeilds 103017 tricks (10.3017), sets 2186 times (21.86%) with imps of -9173 (-0.9173).
C8 lead yeilds 102777 tricks (10.2777), sets 2274 times (22.74%) with imps of -7936 (-0.7936).
C4 lead yeilds 102680 tricks (10.268), sets 2302 times (23.02%) with imps of -7533 (-0.7533).
C3 lead yeilds 102680 tricks (10.268), sets 2302 times (23.02%) with imps of -7533 (-0.7533).

That is, the DD best lead is the ♣A (unsurprisingly since DD favors the A always knowing the exact right continuation). The second best choice was the ♦3/4. Leading a trump was about as good as leading the ♦J, and both were better than a heart or low club. The really horrible lead was the ♥K, and the ♣T was also pretty bad.

I added the "second best choice" to the version of the MSC poll I did last month, partially because I knew the A was going to be the DD favorite this month, and wanted to do the "second best choice" this month without making that 100% obvious that the A was going to be DD best. I'm still trying to figure out how to sim next months MSC, since the auction is very messy, and much of the problem's purpose, I think, is to figure out what people think the various sequence means.

[mikeh]

Mbodell, on 2015-July-16, 13:32, said:

For some results, here was how the panel in MSC voted/rated:



At the time of this posts the bbf voting is scattered with multiple votes for both spades, diamond 3, and club A and single votes for others.

Based on my simulation (code here, you can check my assumptions on this auction restrictions) I got this result for simulation (imp score is compared to the ♣A lead):



That is, the DD best lead is the ♣A (unsurprisingly since DD favors the A always knowing the exact right continuation). The second best choice was the ♦3/4. Leading a trump was about as good as leading the ♦J, and both were better than a heart or low club. The really horrible lead was the ♥K, and the ♣T was also pretty bad.

I added the "second best choice" to the version of the MSC poll I did last month, partially because I knew the A was going to be the DD favorite this month, and wanted to do the "second best choice" this month without making that 100% obvious that the A was going to be DD best. I'm still trying to figure out how to sim next months MSC, since the auction is very messy, and much of the problem's purpose, I think, is to figure out what people think the various sequence means.

I suspect that I misread your code, which wouldn't surprise me at all. However, maybe you can clarify the shape constraints you imposed on the 1NT opening bid. FWIW, I would have limited the hand to 4 spades, 2-3 hearts, and 2-4 in each minor, plus (if I could include this) any 5332 with 5 spades, and of course opener is to have 16-17 hcp.

I am not at all sure what your constraints were, but what info I was able to discern causes me to think there may be a problem. I doubt, for example, that many good players would hold 4=2=2=5 shape for 1NT, so if your sim allows that, then IMO the results will be skewed, and the same is true if you allow 4 hearts for opener and so on. I see you ruled out that latter hand, but you had responder using stayman with 7 hcp, and inviting, and I don't think that would be popular.

Edit: double dummy analysis would, I think, distort reality. Consider dummy holding AJxx or KJxx and partner Qxx in trump. Declarer would almost surely lose a trump trick in real life, unless we led a trump, but playing double dummy would lead the J through partner. Thus your sim over-states the effectiveness of the spade lead by assuming that it doesn't blow the suit ever. Similar issues may impact other choices, but certainly this issue really weakens the value of the sim, IMO.

[Mbodell]

mikeh, on 2015-July-16, 23:37, said:

I suspect that I misread your code, which wouldn't surprise me at all. However, maybe you can clarify the shape constraints you imposed on the 1NT opening bid. FWIW, I would have limited the hand to 4 spades, 2-3 hearts, and 2-4 in each minor, plus (if I could include this) any 5332 with 5 spades, and of course opener is to have 16-17 hcp.

I am not at all sure what your constraints were, but what info I was able to discern causes me to think there may be a problem. I doubt, for example, that many good players would hold 4=2=2=5 shape for 1NT, so if your sim allows that, then IMO the results will be skewed, and the same is true if you allow 4 hearts for opener and so on. I see you ruled out that latter hand, but you had responder using stayman with 7 hcp, and inviting, and I don't think that would be popular.

Edit: double dummy analysis would, I think, distort reality. Consider dummy holding AJxx or KJxx and partner Qxx in trump. Declarer would almost surely lose a trump trick in real life, unless we led a trump, but playing double dummy would lead the J through partner. Thus your sim over-states the effectiveness of the spade lead by assuming that it doesn't blow the suit ever. Similar issues may impact other choices, but certainly this issue really weakens the value of the sim, IMO.

I generally assume that 4333, 4432, 5332 (including 5M), 5m422 (I.e., not when the 5 is a major), and 6m322 are all possible 1nt shapes (before the rest of the auction comes about). If it is 5M332 and a max, (17 hcp), then I disallow it assuming players would open 1M and upgrade. It is true that not everyone opens all these shapes (In my experience I think the frequency is open 1nt on 5M332 > 5m422 > 6m332 > 5M422 > 5431 > 4441. I didn't include the last 3 at all, but did include the first 3.) Obviously, for many folks it depends on the quality of the suits and fragments as well. I admit I'm biased a bit to include the hands I'd open on (which include nearly all the 5m422 hands as well as the majority of the 6m322 hands) because I sometimes sim auctions I was in. Of course on this auction when we bid 2♠, we have 4 or 5 spades, so the 6m322 can't happen, but 4=2=2=5 and 4=2=5=2 can happen. 5=3=3=2, 5=3=2=3, 5=2=3=3, 4=3=3=3, 4=3=4=2, 4=3=2=4. 4=2=5=2, 4=2=2=5, 4=2=4=3, 4=2=3=4 are the 1nt shapes possible given the auction here (not 4=4=3=2 or 4=4=2=3 because I assumed BWS of 2♥ with both majors). In terms of HCP, with the last 7 shapes opener has 16-17. With the 5(332) opener has 15 or 16 (I assumed 17 would have opened 1♠ and that even with only 15, you'd accept with 5 card spades - double dummy you certainly should opposite my invite range below - even with the invite occasionally having 7 hcp).

For the 7 point hands, I only had those if you had 4 hearts and 5 spades and 7 points for the invite opposite 1nt (I agree not everyone would do that, but some would be able to try to improve the part score and upgrade to the invite when they hit the 9+ card trump fit). For the 8 hcp invites I had you accept any time you had 4+ spades (and a shape that uses stayman) except exactly 4=3=3=3. Invites also include 9 hcp hands, but not 10 (since 10 would game force).

I would invite anyone else to use the assumptions that make sense to them, I expect the DD results will be fairly similar in their ordering of the leads, but perhaps not. It is true that DD gets certain things wrong about guessable suits and the like. I don't think this hand is especially bad for DD analysis, though.

[mikeh]

Mbodell, on 2015-July-17, 01:21, said:

I generally assume that 4333, 4432, 5332 (including 5M), 5m422 (I.e., not when the 5 is a major), and 6m322 are all possible 1nt shapes (before the rest of the auction comes about). If it is 5M332 and a max, (17 hcp), then I disallow it assuming players would open 1M and upgrade. It is true that not everyone opens all these shapes (In my experience I think the frequency is open 1nt on 5M332 > 5m422 > 6m332 > 5M422 > 5431 > 4441. I didn't include the last 3 at all, but did include the first 3.) Obviously, for many folks it depends on the quality of the suits and fragments as well. I admit I'm biased a bit to include the hands I'd open on (which include nearly all the 5m422 hands as well as the majority of the 6m322 hands) because I sometimes sim auctions I was in. Of course on this auction when we bid 2♠, we have 4 or 5 spades, so the 6m322 can't happen, but 4=2=2=5 and 4=2=5=2 can happen. 5=3=3=2, 5=3=2=3, 5=2=3=3, 4=3=3=3, 4=3=4=2, 4=3=2=4. 4=2=5=2, 4=2=2=5, 4=2=4=3, 4=2=3=4 are the 1nt shapes possible given the auction here (not 4=4=3=2 or 4=4=2=3 because I assumed BWS of 2♥ with both majors). In terms of HCP, with the last 7 shapes opener has 16-17. With the 5(332) opener has 15 or 16 (I assumed 17 would have opened 1♠ and that even with only 15, you'd accept with 5 card spades - double dummy you certainly should opposite my invite range below - even with the invite occasionally having 7 hcp).

For the 7 point hands, I only had those if you had 4 hearts and 5 spades and 7 points for the invite opposite 1nt (I agree not everyone would do that, but some would be able to try to improve the part score and upgrade to the invite when they hit the 9+ card trump fit). For the 8 hcp invites I had you accept any time you had 4+ spades (and a shape that uses stayman) except exactly 4=3=3=3. Invites also include 9 hcp hands, but not 10 (since 10 would game force).

I would invite anyone else to use the assumptions that make sense to them, I expect the DD results will be fairly similar in their ordering of the leads, but perhaps not. It is true that DD gets certain things wrong about guessable suits and the like. I don't think this hand is especially bad for DD analysis, though.

You and I definitely differ on 5422 hands that have no rebid problem. I think it to be bad bridge to open 1N with such hands, especially at imps where 5m (not to mention minor suit slams or moysian major suit games) are to be preferred over failing 3N contracts. Also, when your sim assumes that the spade lead will virtually never blow a trick, because declarer will always play us for 10x whatever suit we lead, and the results of all non heart K leads are within a couple of percentage points, then you have lost me completely. The implications of double dummy play must have at least a few percentage point implications, thus rendering the results you obtained basically meaningless IMO. This is not an attempt to salvage respect for my choice of leads, btw I have long viewed double dummy analysis as seriously flawed as a means of assessing this sort of problem.