[nigel_k]

bluecalm, on 2012-February-13, 17:38, said:

OK, I will start. How often 3NT makes opposite 6-7 balanced with maximum 4-4 in majors (the hand which passes 1NT opener but drives o game opposite 18-19).

1st hand:

3NT makes 46%

2nd hand:

3NT makes 35%

Which is not really surprising if you did a lot of those simulations before.
In short:



This is correct.
Shape is overrated (when it comes to 1NT openings), you should never downgrade just because you have 4-3-3-3 and probably you shouldn't upgrade because of 5-3-3-2. What counts are pc, spots and aces. (A's are underrated and two tens and two nines are worth almost 1pc while 3 tens are probably a bit more than 1pc).
Some people did a lot of research with dd simulations and real hand results to arrive at those conclusions, I won't repeat this discussion here.
I can just say, that I am paying attention to this stuff and imo this upgrading algorithm:
a)don't worry about shape
b)add 1PC for 2aces and 2tens or 3tens or 2tens/2nines/one ace
c)downgrade only awful hands without any spots

Is imo better than w/e people at w/e level come up with.

I will try to replicate your simulation results and I may well be wrong about that specific pair of hands. But the rest of your post seems clearly wrong. You are basically saying:

a) 4333 and 5332 should be treated as if they are the same

b) There is no difference between having one zero aces and zero tens and having one aces and two tens, regardless of the rest of the hand.

c) Far more hands will qualify for an upgrade than a downgrade. This means you are not playing a 15-17 NT any more. The strength range is lower because you are putting in 14 HCP hands and taking out 17 HCP hands, but hardly ever putting in 18 HCP hands or taking out 15 HCP hands. The whole point about upgrading and downgrading is that partner can make the right decision when they assume your hand is in the nominated range. And partner is also allowed to be aggressive.

[JLOGIC]

4333 = 5332 is a common result of simulations, this has been replicated many times on RGB. It is possible human judgement is wrong, but the real answer imo is that double dummy they always make the right lead, so having a doubleton makes it more likely they have a long suit they can attack successfully. Gravity stoppers FTW. I do not think that 4333 = 5332 holds true in real life.

The other problem of course is that when you open 1N you do not know you're going to play NT. 5332 will be better for a suit contract if you have a fit since you have a ruffing value.

[bluecalm]

Quote

4333 and 5332 should be treated as if they are the same

I believe this is correct. Counter intuitive it might be but if there is any difference it's so small that it could be ignored.

Quote

There is no difference between having one zero aces and zero tens and having one aces and two tens, regardless of the rest of the hand.

I wrote that two tens are worth almost 1hcp. I mean, there is difference but somehow difference is small. You have to draw the line somewhere and while A532 K52 AQ5 J52 is cleary not 1NT opening, AT32 KT2 AQ5 J52 might still not be 1NT opening while AT32 KT2 AQ5 JT2 is slam dunk 1NT opening if you play classical (which means partner will pass awful 8's and invite with decent 8's).

Quote

Far more hands will qualify for an upgrade than a downgrade. This means you are not playing a 15-17 NT any more.

No it doesn't mean it.
By 15-17 I mean what is generally expected and how people react to it which is driving to game with every 10 and inviting with many (most ?) 8's while passing all <8hcp.

Quote

The whole point about upgrading and downgrading is that partner can make the right decision when they assume your hand is in the nominated range

The whole point is to open hand which are worth 15-17hcp.
To me hand with 15hcp is worth 15hcp unless there are some glaring weaknesses in it while many good 14's (ie having a lot of spots) are worth 15hcp. Your interpretation is that 15hcp is not really 15hcp unless it has its share fair of positive additions like shape or spots.
What I want is to count T's and 9's for something and adjust value of A's a bit.
I mean, if A32 AQ2 AJ2 532 is not 15-17hcp 1NT opening in your book maybe we don't agree what "hcp" stands for...

[nigel_k]

4333 and 5332 are not the same. Saying they are is not just counter intuitive, it is totally at odds with having played bridge for any length of time and observed how cards take tricks.

My point about the two tens is that the rest of the hand matters. Sometimes the honour structure apart from any tens will be such that one ten will put the hand into a higher category. Sometimes it will be such that you need two tens or three tens for that. By disregarding the rest of the hand you have forced yourself into the conclusion that a hand with one ace and two tens is no better than another hand with the same HCP but no aces and no tens.

On the third point, I agree that you are still opening hands that you view as worth 15-17 HCP. But there will be upgraded 14s in there but no downgraded 18s, and good 17s will be taken out but bad 15s remain in. This means that when you open 1NT the frequency of hands at the low end is increased and the frequency of hands at the high end is decreased. So the expected strength is weaker than a vanilla 15-17 1NT opening would be.

[JLOGIC]

I like it nigel, you appeal to simulations, then when simulations prove you wrong you say they are counter to playing bridge for any length of time and point out their flaws. Either way you win!

A point is a point, it's kind of like how a pawn is a pawn in chess. These points that we have are not bad. I still maintain that if you downgrade this hand, you will downgrade most 4333 18s. Maybe you do but I think that is a bad strategy.

Also re simulations and computers, it is important to note how much computers changed backgammon, a lot of things that were years of conventional wisdom/common knowledge were proven wrongs, things as basic as opening moves. For instance, using 6-4 to make the 2 point is not bad at all, even though it was considered horrendous. Likewise, "computer moves" changed chess, though not as much as they did for backgammon. It is possible that the simulations are correct, but I agree with you that they are likely flawed when it comes to 4333 vs 5332. Still, you are the one who brought up simulations.

For my money I would always want the example hand over your example hand. But you example hand is not really relevant to whether or not this hand is above average for 4333 18 counts as it is not 4333.

[bluecalm]

I can just add, for the record that almost never downgrading 18's and often upgrading 14's is exactly what the best players in this game do and that is true even for very disciplined openers.
For example Lauria-Versace always played 15-17 NT which is like gold standard for classical as they rarely pass 8 count opposite it, they never open 1NT with 5M and very rarely with 6m or 2-2-(5-4).
In whole vugraph history they opened 1NT 529 times. 2 times with 18hcp and 88 times with 14.
You can do that for any other world class pair and you will see the same thing: a lot of upgrades, often off shape weaker openings but those 4-3-3-3 outside the upper range never land in 1NT.

[gnasher]

JLOGIC, on 2012-February-14, 13:51, said:

4333 = 5332 is a common result of simulations, this has been replicated many times on RGB. It is possible human judgement is wrong, but the real answer imo is that double dummy they always make the right lead, so having a doubleton makes it more likely they have a long suit they can attack successfully. Gravity stoppers FTW. I do not think that 4333 = 5332 holds true in real life.

Another factor is that the more balanced you are the more options you have, so there is more scope for the double-dummy solver to gain an unfair advantage. For example:

  KJ432
  A10
The double-dummy solver gains against Qxx-xxx compared with single dummy.

 KJ43
 A102
The double-dummy solver gains against Qxx-xxx and Qx-xxxx.

Rather than a simulation, I'd be more interested in an analysis of real-life deals with human declarers and defenders. Such a study would need quite careful design, though. You couldn't for example, consider only deals that were played in 3NT, because you'd get selection bias caused by people upgrading 5332s and downgrading 4333s. On the other hand, someone playing in 1NT at IMPs may not make nine tricks when he can. Hence you'd probably have to use matchpoint deals only.

[Zelandakh]

gnasher, on 2012-February-15, 07:19, said:

Rather than a simulation, I'd be more interested in an analysis of real-life deals with human declarers and defenders.

I thought I read about a French analysis that reached the conclusion that a 5 card suit was worth 0.4 hcp. It is a long time ago though so I cannot remember any details, or even if I am right about it!

[bluecalm]

Quote

You couldn't for example, consider only deals that were played in 3NT, because you'd get selection bias caused by people upgrading 5332s and downgrading 4333s. On the other hand, someone playing in 1NT at IMPs may not make nine tricks when he can. Hence you'd probably have to use matchpoint deals only.

You could use imp hands and how often the contract made opposite various pc counts as metric.
I am not sure how big variance is for such thing. If I am bored later today I can do that on vugraph hands.
Before I do though, let's think what would constitute good argument as I am done with doing exactly what my opponent in this discussion asked for and then having him back off instead of admitting defeat
For example if I pull all 5-3-3-2's 14count which opened 1NT and got raised to 3NT directly by hand with 9-12hcp and compare it to all 4-3-3-3's 15count which opened 1NT and got raised to 3NT by the same range of hands, does it prove anything if 4-3-3-3's 15 count make the game more often ? Or maybe I need to add some features to 14hcp 5-3-3-2's ?
Any other ideas ?

Quote

I thought I read about a French analysis that reached the conclusion that a 5 card suit was worth 0.4 hcp.

I think 0.4 is reasonable. I don't think its that much for now but I wouldn't exclude the possibility.

[gnasher]

bluecalm, on 2012-February-15, 08:23, said:

For example if I pull all 5-3-3-2's 14count which opened 1NT and got raised to 3NT directly by hand with 9-12hcp and compare it to all 4-3-3-3's 15count which opened 1NT and got raised to 3NT by the same range of hands, does it prove anything if 4-3-3-3's 15 count make the game more often ?

All that would tell us is that people undervalue 4333s or overvalue 5332s. We wouldn't know whether there was no difference between the shapes, or there was a difference but people tended to overrate it.

[nigel_k]

JLOGIC, on 2012-February-14, 22:44, said:

I like it nigel, you appeal to simulations, then when simulations prove you wrong you say they are counter to playing bridge for any length of time and point out their flaws. Either way you win!

A point is a point, it's kind of like how a pawn is a pawn in chess. These points that we have are not bad. I still maintain that if you downgrade this hand, you will downgrade most 4333 18s. Maybe you do but I think that is a bad strategy.

It is true that if I was comparing two hands that were basically identical except that one was 4333 and the other was 5332, then I wouldn't bother with a simulation to see if there was a difference. And if a simulation showed no difference I would assume the results were wrong even if I couldn't immediately figure out why.

But neither of the hands I suggested comparing in a simulation was 5332 so that weakens your criticism quite a lot. If simulations show that 4333 and 4432 are nearly the same when playing in NT I have no problem with that.

Also, there are plenty of ways a 4333 18 can be better. It could have a ten, or even a nine, or a three card sequence such as KQJx or AKJx. Most 4333 18s have one of those.