[Jlall]

eyhung, on Oct 17 2009, 05:28 PM, said:

Surprisingly, when I simmed the hand, if partner is guaranteed to hold 6 hearts (but never 6-4 shape) with 4-9 HCP, then game is 55%, indicating he should bid 4H!

Where do you get this? This number could easily mean inviting is correct (and does mean that to me). If you want to try to prove that bidding game > inviting you need to prove that game is better than 50 % on hands where partner passes your invite. It's tough/impossible to simulate judgment but I bet if you give partner 4-6 and 6 hearts game is not better than 50 %.


But now simulations are getting dumb since we're just using HCP, so we might as well think about it from a bridge point of view. Bidding 3H on this auction shows a very good hand. Generally a primed out max with 3 hearts. That is what we have. Surely we can respect partner's judgment to be pretty good when we are able to describe our hand very well. The only time bidding game would be better than inviting is if we were not describing our hand that well.

Let's say game was 80%, but partner knew our exact hand if we bid 3H. We would obviously choose to bid 3H right? An extreme example, but you get the point. Partners range is extremely wide right now, ours will be extremely narrow. He should make the final decision.

Also, I would very often, much more than I would over 1D p 1S p 1N, go back to a 5 card heart suit on the auction 1D p 1H p 1N. 1H here is never a singleton, and there are more unbalanced hands with 5 hearts that we can have that don't get to show their second suit.

For instance 1D p 1S p 1N with 5-4 in either red suit we go back to that red suit. After 1D p 1H p 1N we can only go back to the second suit in one case, when it's diamonds. When we have 4-5 in the majors we go back to hearts. This increases the amount of times we will need to rebid our 5 card major, not even factoring in that it's less risky given that partner never has a stiff.

[eyhung]

Jlall, on Oct 17 2009, 03:34 PM, said:

eyhung, on Oct 17 2009, 05:28 PM, said:

Surprisingly, when I simmed the hand, if partner is guaranteed to hold 6 hearts (but never 6-4 shape) with 4-9 HCP, then game is 55%, indicating he should bid 4H!

Where do you get this? This number could easily mean inviting is correct (and does mean that to me).

Justin, you are correct. To rephrase: given a "false dilemma" of bidding 4H and passing 2H, bidding 4H rates to be the better choice. But this ignores the superior third choice, the invite. Obviously inviting is superior to bidding 4H for the reasons you mentioned. I just wanted to highlight to my disbelieving partner that if I held 6, just blasting game was better than passing.

And yes, the rebid of 2H tends to be 5 cards even more than the rebid of 2S. So the concept of "2H guarantees 6" is flawed. But the result that game is percentage opposite 6 hearts was not immediately obvious to me or my partner.

By the way, one nice thing about all this simulation work is that I am building up a library of functions which I can use to capture judgement heuristics and then reuse them for different hands in similar situations. As you can see this library is currently in a primitive state, and I find your comments invaluable to refining my functions.

[Jlall]

Are you really that surprised? You would bid 4H opposite a weak 2 without any qualms probably. Quick tricks + ruffing value + 3 trumps is a great hand opposite a 6 card suit.

[eyhung]

Jlall, on Oct 17 2009, 03:55 PM, said:

Are you really that surprised? You would bid 4H opposite a weak 2 without any qualms probably. Quick tricks + ruffing value + 3 trumps is a great hand opposite a 6 card suit.

You haven't seen some of the weak twos we open.

Just kidding, I'm not that psycho. Point taken.

[fred]

Jlall, on Oct 17 2009, 10:19 PM, said:

So the only real non 5332s here hands with side clubs. Probably 5332s are more common than hands with side clubs (not sure?) so if you pass with most 5332s it seems like 80 % was an overbid.

80% could easily be off - I was really just guessing what the number felt like from experience (versus guessing at a computation). Also, my guess was based on the general class 1m-1M-1NT as opposed to this specific auction. For some reason my brain doesn't give me a sense of what the numbers have been for the various auctions in this class. Maybe I need to play some more hands

I normally play Reverse Flannery so side hearts is not a possibility for me anyways, but agree with your general point that normally in this auction side clubs is the only possibility for non-5332 hands that contain exactly 5 spades.

Also agree that we are close to seeing eye-to-eye on this one.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[fred]

Eugene,

Your results are interesting - thanks.

I would be interested in knowing how much difference it makes if you ignore all constraints on the opponents' hands. If not too much trouble, can you please try that and post the results?

This request is not in any way relevant to the discussion - I am just curious.

I might be confused since I have never used your general methodology before, but I am not sure you are doing this right. I think the idea should be to always give North a hand with exactly 5 spades that would either pass 1NT or bid 2S and see how many spades South has on average under the appropriate constraints.

If I understand what you have said (sorry in advance if I have not) you are including hands for North that contain 4 spades, 6+ spades, and hands with at least enough strength to invite game. To answer this question properly I think you should take these hands out of the mix.

Thanks,

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[eyhung]

fred, on Oct 17 2009, 05:17 PM, said:

Eugene,

Your results are interesting - thanks.

I would be interested in knowing how much difference it makes if you ignore all constraints on the opponents' hands. If not too much trouble, can you please try that and post the results?

This request is not in any way relevant to the discussion - I am just curious.

I might be confused since I have never used your general methodology before, but I am not sure you are doing this right. I think the idea should be to always give North a hand with exactly 5 spades that would either pass 1NT or bid 2S and see how many spades South has on average under the appropriate constraints.

If I understand what you have said (sorry in advance if I have not) you are including hands for North that contain 4 spades, 6+ spades, and hands with at least enough strength to invite game. To answer this question properly I think you should take these hands out of the mix.

Thanks,

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

I actually did remove the opponent constraints out of curiosity :

1 spade = 12.91% (7.47% are 4x1s)
2 spades = 48.69%
3 spades = 38.4%

So it's similar, but slightly less likely to be singleton.

This is including hands where north has 4 spades, 6 spades, gf, weak, etc.

I actually wasn't trying to answer the question of how often we should rebid 2S over a 1NT rebid in this auction, I was just trying to see how often South would have a singleton spade after a real-life auction of 1D 1S 1NT with no enemy bidding. 30:1 felt wrong given that I have also taken to rebidding 1NT with a singleton.

I will be happy to remove game-invitational+ hands, and hands without exactly 5 spades from the mix and report the results.

[fred]

eyhung, on Oct 18 2009, 12:25 AM, said:

30:1 felt wrong given that I have also taken to rebidding 1NT with a singleton.

I am not sure that 30:1 is right either, but there were several things you did that were different than what I did:

- You didn't include (23)44 in your initial simulation

- You included 1444 hands in some of your simulations

- You included hands with 3-card support and small doubletons in some of your simulations

- You made assumptions about the opponents' hands

- Your methodology results in "dependent probabilities" (a good thing!) while mine does not

If you feel kind enough to check my 30:1 number, you should do this:

1) Make South the only hand you constrain and give South 1453 with 12-14 HCP. I get 0.11%.

2) Make South the only hand you constrain and give South 2 or 3 spades, 2 to 4 hearts, 4 or 5 diamonds, 2 to 4 clubs, and 12-14 HCP. I get 3.15%. I now see that I did this wrong as far as the actual problem is concerned since I was including 2254, but I would still be interested in seeing if you get the same %s as me.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[eyhung]

fred, on Oct 17 2009, 05:47 PM, said:

1) Make South the only hand you constrain and give South 1453 with 12-14 HCP. I get 0.11%.

2) Make South the only hand you constrain and give South 2 or 3 spades, 2 to 4 hearts, 4 or 5 diamonds, 2 to 4 clubs, and 12-14 HCP. I get 3.15%. I now see that I did this wrong as far as the actual problem is concerned since I was including 2254, but I would still be interested in seeing if you get the same %s as me.

Generating 1 million hands

1) 0.1154%
2) 3.1474%


-----------------------

Sample size: 1000 hands
North has exactly 5 spades, at most 9 HCP, South has a hand that would open 1D and rebid 1NT (and would have raised on some 3-card spade suits, as described earlier), E/W can have any hand (including those that would bid in real life)

South average spade length = 2.271
1 spade = 132
2 spades = 465
3 spades = 403
4x1 = 84
2S makes but 1NT doesn't = 188
1NT makes but 2S doesn't = 121
Both make = 326
Both go down = 365
Average tricks in spades = 7.498
Average tricks in NT = 6.288

----------------------

Sample size: 1000 hands
Same N/S parameters, but now E/W are operating under the pass_functions described earlier.

Average spade length = 2.23
1 spades = 140
2 spades = 490
3 spades = 370
4x1 = 77
2S makes but 1NT doesn't = 176
1NT makes but 2S doesn't = 110
Both make = 383
Both go down = 331
Average tricks in spades = 7.672
Average tricks in NT = 6.433

-----------------------

Apologies for the smaller sample size, but double-dummy analysis takes longer to run, and we're getting similar numbers (13-14% spade singletons) as the larger sample sizes. Bottom line, 2S seems to be superior to 1NT at IMPs on a 5-bagger even if we rebid 1NT fairly freely with a singleton.

[eyhung]

Note that I didn't put any additional constraints on the North hand such as removing hands that would judge to pass 1NT, such as 9 HCP 5332s.

[655321]

Lobowolf, on Oct 17 2009, 09:27 AM, said:

655321, on Oct 15 2009, 09:46 PM, said:

1NT happily, this is not even an evil to be avoided.

A side benefit is that partner will no longer rebid his unsupported 5 card spade suits.

That's like saying (but with 3 fewer HCP) that you should open 1NT with a singleton regularly, with the benefit that partner will stop transferring into unsupported 5-card spade suits.

Are you deliberately trolling?

Obviously, if your 1NT openings never have 4 or 5 spades, and if they can only have 3 spades in a hand that would not raise partner's 1♠ response, then your comparison would be valid. But what does that have to do with anything?

There is a difference between auctions where partner has denied primary spade support, and those where he has not. You may believe that rebidding spades with 5 is better than passing 1NT, but I still don't see why that should cause you to miss the distinction.

[Lobowolf]

655321, on Oct 17 2009, 11:25 PM, said:

Lobowolf, on Oct 17 2009, 09:27 AM, said:

655321, on Oct 15 2009, 09:46 PM, said:

1NT happily, this is not even an evil to be avoided.

A side benefit is that partner will no longer rebid his unsupported 5 card spade suits.

That's like saying (but with 3 fewer HCP) that you should open 1NT with a singleton regularly, with the benefit that partner will stop transferring into unsupported 5-card spade suits.

Are you deliberately trolling?

Obviously, if your 1NT openings never have 4 or 5 spades, and if they can only have 3 spades in a hand that would not raise partner's 1♠ response, then your comparison would be valid. But what does that have to do with anything?

There is a difference between auctions where partner has denied primary spade support, and those where he has not. You may believe that rebidding spades with 5 is better than passing 1NT, but I still don't see why that should cause you to miss the distinction.

Your original post to which I replied is certainly, to an extent, circular. The point that a "benefit" of rebidding 1NT with a singleton is that partner won't rebid spades with 5 is somewhat specious, because rebidding a 5-card spade suit is, for the most part, primarily a problem because you might have a singleton. If you didn't rebid 1NT with a singleton, you'd WANT partner to rebid his 5-card major suits with a weak hand.

None of which is to dispute the other benefits of the 1NT rebid, which have been enumerated in other posts on this thread.

I agree that my analogy is imperfect in degree, because the 1NT opener may have 4 spades, while the 1NT rebidder does not, and the 1NT rebidder may have raised spades directly with 3 (though depending on your methods, the direct raise with 3 is comparatively rare). However, I have in the past run simulations comparing 1NT and 2♠ contracts where opener has a balanced hand with 2 or 3 spades, and found the 2♠ contract to be superior.

You may believe that passing 1NT is better than rebidding spades with 5, but I don't see why that should cause you to miss the distinction between the set of hands those where the 1NT rebids contain 2 or 3 spades, and those where the 1NT rebids contain 1, 2, or 3 spades.

[hanp]

The following double dummy simulation (using 1000 hands) may be interesting:

South has 12-14 HCP with an 2353, 3253, 3352, 2344, 2443, 3244, 3442, 3444, 1453 and 1444 pattern.

North is 5(332) with 8 HCP.

Neither opponent has 15+ HCP, 6+ hearts or a 7+ suits. West also doesn't have 6 spades. (a very mild condition removing some hands where there would very likely be opposition bidding.)

1NT by south makes on 56% of the hands, 2S by north makes on 54% of the hands.

Further comments: Against 2S the optimal opening lead can often be from any suit, while against 1NT the optimal opening lead is usually from only 1 or 2 suits. This may mean that double dummy analysis favors 2S contracts relative to 1NT contracts (where in practice the opening lead will more often be wrong).

To reflect a different style, I then reran the simulation, requiring now that opener does not have 3-card support and a small doubleton. The results were:

1NT makes 61% and 2S makes 52% of the time.

Again the lead against 1NT is more demanding than the lead against 2S.

As this second style is closer to my own, it seems that I should rebid 2S with 5(332) hands with 8 HCP only rarely when playing IMPs.

[hanp]

When browsing through the hands afterwards I noticed that on hands where 2S makes but 1NT goes down it is very often the case that one of the opponents has a solid overcall. To get more accurate answers I should further restrict the hands of the opponents.

[aguahombre]

hanp, on Oct 18 2009, 06:38 AM, said:

When browsing through the hands afterwards I noticed that on hands where 2S makes but 1NT goes down it is very often the case that one of the opponents has a solid overcall. To get more accurate answers I should further restrict the hands of the opponents.

Alternatively, do you need to throw in the effect of allowing a balance at the 2 level when the 1NT rebid is passed around? And then the effect on the defense if 2S is bid in delay--or the possibility that the opponents play their fit at the 3-level?

It seems to become a question not of whether 1NT or 2S will do better on a million generated hands, but rather of whether you get a better result. I know the simulators tried to eliminate competitive hands from the opposition, but that doesn't stop them from competing --sometimes successfully.

[eyhung]

aguahombre, on Oct 18 2009, 08:12 AM, said:

Alternatively, do you need to throw in the effect of allowing a balance at the 2 level when the 1NT rebid is passed around?  And then the effect on the defense if 2S is bid in delay--or the possibility that the opponents play their fit at the 3-level?

That's a good question, but in practice, I find I infrequently want to balance over 1m - 1M - 1NT when the opponents don't have a fit and I don't have a hand that was worth an action initially. I grant you it's certainly possible, but I don't think it's highly likely.
Responder knows way too much about opener's hand for my comfort. If someone can suggest parameters for bidding over this, it's easy enough for me to discard these hands from the opponent filter as well.

[fred]

eyhung, on Oct 18 2009, 01:35 AM, said:

fred, on Oct 17 2009, 05:47 PM, said:

1) Make South the only hand you constrain and give South 1453 with 12-14 HCP. I get 0.11%.

2) Make South the only hand you constrain and give South 2 or 3 spades, 2 to 4 hearts, 4 or 5 diamonds, 2 to 4 clubs, and 12-14 HCP. I get 3.15%. I now see that I did this wrong as far as the actual problem is concerned since I was including 2254, but I would still be interested in seeing if you get the same %s as me.

Generating 1 million hands

1) 0.1154%
2) 3.1474%

Cool - thanks!

Looks like my program works

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[Jlall]

I think these simulations are proving that using judgment is a good idea with 5 spades when deciding whether or not to bid. Suit quality, shape, honor location, and HCP are all important factors.

[aguahombre]

Jlall, on Oct 18 2009, 12:57 PM, said:

I think these simulations are proving that using judgment is a good idea with 5 spades when deciding whether or not to bid. Suit quality, shape, honor location, and HCP are all important factors.

true enough. And I just noticed that the sims all assumed opening 1D with 4-4m and 3-2 or 2-3M. So people who open 1C with those cannot have 4clubs and 2 or 3 spades when they rebid 1NT. Responder with 5332 (doubleton club) knows the opps have at least 8 clubs and this might affect the decision to remove.