[RossSCann]

I read Frank Stewart's column and bidding question every day. Recently he gave an answer I was sure was wrong. I happen to have been writing a bridge teaching program as a retirement hobby for the past 20 years. I realized I could test the issue with my program. Given the specified hand for South, I could generate a large number of random distributions for the cards in the other three hands to see the outcome for each of two different bids by South.
Here is the case:
South holds S's J 10 6 5 2 H's Q J 10 9 D's 7 C's K 4 2
Dealer N opens 1 D, S responds 1 S, N now bids 1 NT: Question is what S bids next. Frank said no question, bid 2 H's
I would pass in an instant with my weak 7 HCPs.
So I ran 500 random cases with my program and the results were not even close.
Pass had 337 tops
2 H's 121 "
42 flat

I sent Frank the results and he cordially declined to believe my analysis. I offered to send him the 500 hands but he doesn't have time to look at them.
My bigger question now is hasn't anybody else ever created an analytical tool like this to get definitive answers on bidding questons???
The bridge bulletins, columns etc. are loaded with such bidding questions. Are pundits just guessing the anwers??
Does anyone here know of such a tool?? Does anyone know of an example where such a bidding question was answered by actually looking at a large sample of random hands relevant to the bids in question and going through the probably bidding and play?
You can look at my program at www.bridgegame.com if you like.
I hope my questions ring some bells here. This seems to me to be an important theoretical issue.

Here is more detail from a rerun.
Results for 500 random cases

Dealer N opens 1 D and S responds 1 S, N now bids 1 Nt: Key bids are 2 H's or Pass


Total Score Top Boards
Key bid 1 2H -5294 135
Key bid 2 Pass 4964 332

Flat boards 33



Distribution of High Card Points For North - Number of Hands
11 12 13 14 15 16 17
0 118 484 258 121 20 0

Distribution of Imps Won - Number of Boards
Imps 2H Pass
1 31 9
2 28 3
3 8 32
4 0 0
5 2 71
6 3 43
7 5 67
8 5 34
9 2 36
10 25 12
11 21 11
12 4 2
13 1 0
14 0 0

Total 774 2099

Contracts Reached After Bid Choice One 2H
Declarer Bid Suit Made Cases
North 3 Notrump -1 46
North 3 Notrump -2 41
South 3 Heart 3 35
North 3 Notrump 3 33
South 3 Heart -2 28
South 3 Heart -1 23
South 4 Heart -2 22
North 2 Notrump -1 21
South 2 Spade 2 19
North 3 Notrump -4 15
South 3 Heart -3 15
North 3 Notrump -3 15
North 3 Diamond -2 14
North 2 Notrump 2 13
South 4 Heart -1 12
South 4 Heart -3 12
North 2 Notrump -2 11
South 3 Heart 4 11
South 4 Heart 4 11
North 3 Club -2 10
North 3 Club 3 9
North 3 Diamond -4 9
North 3 Diamond -1 8
South 2 Spade -1 7
South 2 Spade -2 7
North 2 Notrump -3 6
South 2 Spade 3 6
North 3 Notrump 4 5
North 3 Diamond -3 5
South 3 Heart -4 5
North 2 Notrump 4 4
North 2 Notrump 3 3
North 3 Club 4 3
North 3 Club -1 2
North 3 Diamond 3 2
North 3 Notrump -5 2
North 3 Diamond -6 2
South 3 Heart 5 1
North 3 Diamond 4 1
North 2 Notrump -5 1
North 3 Club -3 1
North 3 Diamond -5 1
South 2 Spade -3 1
South 4 Heart -4 1
North 3 Club -4 1

Contracts Reached After Bid Choice Two Pass
Declarer Bid Suit Made Cases
North 1 Notrump 1 197
North 1 Notrump 2 112
North 1 Notrump -1 69
North 1 Notrump 3 62
North 1 Notrump -2 32
North 1 Notrump 4 13
North 1 Notrump -3 8
North 1 Notrump -4 5
North 1 Notrump -5 1
North 1 Notrump 5 1

Distribution of North Hands

Number of
Cases Spades Hearts Diamonds Clubs
80 3 3 4 3
75 2 3 5 3
69 2 4 4 3
51 2 3 4 4
37 2 4 5 2
30 3 3 5 2
25 3 2 5 3
22 3 4 4 2
21 3 2 4 4
20 1 4 4 4
14 2 2 5 4
12 2 2 4 5
10 4 3 4 2
8 1 4 5 3
6 4 4 3 2
5 1 3 5 4
5 4 2 4 3
4 1 3 4 5
4 4 2 5 2
2 1 2 5 5

[blackshoe]

I don't think 500 cases is a large enough sample set. Run it for 5000 and see what you get.

[lalldonn]

blackshoe, on 2012-July-06, 18:34, said:

I don't think 500 cases is a large enough sample set. Run it for 5000 and see what you get.

Obviously that's not the problem here. What do "random hands" mean? What are the constraints on each of the other players? That is where the problems are in this analysis, guaranteed.

[Stephen Tu]

What exactly do you mean by "tops", do you just mean NT outscores the major on a MP basis? Did you take into account that North is taking preference to 2 spades fairly often, not passing 2 hearts? Are you comparing mp score of the resulting major partial vs. NT? How are you scoring the hand, do you have a double-dummy engine you wrote or got from somewhere else?

What were your parameters for opener rebidding 1nt? (does north raise to 2s with wk nt with three spades and outside small doubleton rather than bid 1nt? Does north ever bid 1nt with a singleton spade?) These kind of things can affect the outcome of your sim.

Lots of us use dealing programs and double dummy analysis to figure out whether borderline decisions are correct or not, in situations which are unclear. Less common for columnists. Generally bridge columnists tend to make their recommendations based on their personal experience of what works or not, or their knowledge of what other experts tend to do, which also tends to be based on what works or not.

It's pretty automatic for people to bid 2 hearts on this sort of hand. If you feel people are wrong, that passing 1nt is better, then present your results. But you have to give more complete data. Is this MP or IMPs? What were parameters for the opposite hand? Are you sure you are scoring each hand vs. the correct major partial (i.e. if opp hand is 3-3 in majors, you want to score 2s vs. 1nt, not 2h vs. 1nt)?

[lalldonn]

Stephen Tu, on 2012-July-06, 18:40, said:

Did you take into account that often North is taking preference to 2 spades fairly often, not passing 2 hearts?

I'll bet that was a problem too lol

[blackshoe]

lalldonn, on 2012-July-06, 18:38, said:

Obviously that's not the problem here. What do "random hands" mean? What are the constraints on each of the other players? That is where the problems are in this analysis, guaranteed.

Yeah, you're right. I was going to address that, but... the cat distracted me. That's my story, and I'm sticking to it!

[Stephen Tu]

FWIW I ran my own sim with the hand in question using Hans van stavaren Dealer / GIB. Gave North 12-14 HCP, 4333/4432/5332 with 4+ diamonds, 2-3 spades. So the "never raise on 3, don't rebid 1nt with stiff" scenario. No constraints set on opponents. Assumed played in hearts if opener has 4, spades otherwise. Major partial took more tricks than NT something like 82% of the time over 2000 deals.

So I am inclined to believe that original poster made some sort of error(s) setting up his analysis. Most probably North always passing 2h as the biggest factor, since his post didn't mention 2s contracts?

RossSCann, on 2012-July-06, 18:16, said:

You can look at my program at www.bridgegame.com if you like.

Bored and curious, I downloaded this. I tried out the NT play tutorial test. I kind of dislike pooh-poohing on something you've obviously spent quite a bit of time on, but IMO this program needs quite a lot more work, and much better bridge analysis, if the quality of the rest of the program is similar which I assume is likely.

Example: diamond suit is something like AJ8x in dummy opposite K7x in hand. Contract 3nt uncontested, LHO leads 2 of diamonds. You have 5 fast tricks outside, but your club suit is xx opposite Qxx, so you'd obviously prefer 4 tricks from this combo without losing the lead. I of course take what I think is the automatic line, play low from dummy, planning deep finesse 2nd round, playing LHO for T9xx/Q9xx/QTxx/t9x/qtx/q9x. This works for me; LHO has t9xx in the problem. I then click on "recommended line". The program says that I am supposed to play the JACK first round ??? calling it a "free finesse"??? It holds itself to 3 tricks in this suit, establishes an entry to RHO in the play, and then LHO helps declarer out by cashing CAK from AKT over declarer's Q, instead of just exiting a heart to partner for the set.

Other NT tutorial test hands were also of highly dubious quality, perhaps too complicated for teaching hands (no clear guaranteed lines), and recommending lines that to me were clearly inferior, attacking suits in wrong order, not preserving entries, not maximizing chances, also relying on defensive errors to make.

In Bridge Master, if you take suboptimal line, opps defeat you, even if they have to swap cards so that finesses fail both ways. In this one, program recommends inferior line, then defense helps you out? It's like Bridge Master written by a low intermediate bridge player who can't see much better lines that are available.

I randomly chose the slam tutorial, it recommended 2nt-3d-3h-4c = Gerber with a 0-5-4-4 hand, LOL. Sorry Ross, I think you need to get a LOT better at bridge before you attempt to teach other people how to play. Try some of the bridgebase.com products yourself, you'll see how far you need to improve, and how below standard your current effort is. Also many excellent books from www.baronbarclay.com are available, and you can post questions here as well, for stuff not easily looked up in books and opinions on particular hands.

[the hog]

2H is obvious. Pass is a beginner's call.

[lalldonn]

I tell you what Ross, I'm no Frank Stewart but run it for 50 hands and post the results here, then send them to me and I'll look at them. jdonn@hotmail.com

[Cyberyeti]

It depends on your parameters, if opener bids 2♠ on some of the most suitable 3 card holdings, it will swing it back considerably.

Also if you pass 1N on some of the hands where partner has 6 or 7 cards in the majors, one of the opps will reopen in a minor and you'll need to work out what happens next.

[RossSCann]

lalldonn, on 2012-July-06, 18:38, said:

Obviously that's not the problem here. What do "random hands" mean? What are the constraints on each of the other players? That is where the problems are in this analysis, guaranteed.

By "random" I mean hands for W N & E of equal probablitiy of occurence that agree with the bidding. i.e points and card length. So the 500 hands are therefore representative of hands that would occur in actual play. You see the distribution of HCP's above for N. Also we know N doesn't have 4 spades or 6 fiamonds, etc.
Neither E or W has a hand that warrants and overcall.

[RossSCann]

Stephen Tu, on 2012-July-06, 21:57, said:

FWIW I ran my own sim with the hand in question using Hans van stavaren Dealer / GIB. Gave North 12-14 HCP, 4333/4432/5332 with 4+ diamonds, 2-3 spades. So the "never raise on 3, don't rebid 1nt with stiff" scenario. No constraints set on opponents. Assumed played in hearts if opener has 4, spades otherwise. Major partial took more tricks than NT something like 82% of the time over 2000 deals.

So I am inclined to believe that original poster made some sort of error(s) setting up his analysis. Most probably North always passing 2h as the biggest factor, since his post didn't mention 2s contracts?



Bored and curious, I downloaded this. I tried out the NT play tutorial test. I kind of dislike pooh-poohing on something you've obviously spent quite a bit of time on, but IMO this program needs quite a lot more work, and much better bridge analysis, if the quality of the rest of the program is similar which I assume is likely.

Example: diamond suit is something like AJ8x in dummy opposite K7x in hand. Contract 3nt uncontested, LHO leads 2 of diamonds. You have 5 fast tricks outside, but your club suit is xx opposite Qxx, so you'd obviously prefer 4 tricks from this combo without losing the lead. I of course take what I think is the automatic line, play low from dummy, planning deep finesse 2nd round, playing LHO for T9xx/Q9xx/QTxx/t9x/qtx/q9x. This works for me; LHO has t9xx in the problem. I then click on "recommended line". The program says that I am supposed to play the JACK first round ??? calling it a "free finesse"??? It holds itself to 3 tricks in this suit, establishes an entry to RHO in the play, and then LHO helps declarer out by cashing CAK from AKT over declarer's Q, instead of just exiting a heart to partner for the set.

Other NT tutorial test hands were also of highly dubious quality, perhaps too complicated for teaching hands (no clear guaranteed lines), and recommending lines that to me were clearly inferior, attacking suits in wrong order, not preserving entries, not maximizing chances, also relying on defensive errors to make.

In Bridge Master, if you take suboptimal line, opps defeat you, even if they have to swap cards so that finesses fail both ways. In this one, program recommends inferior line, then defense helps you out? It's like Bridge Master written by a low intermediate bridge player who can't see much better lines that are available.

I randomly chose the slam tutorial, it recommended 2nt-3d-3h-4c = Gerber with a 0-5-4-4 hand, LOL. Sorry Ross, I think you need to get a LOT better at bridge before you attempt to teach other people how to play. Try some of the bridgebase.com products yourself, you'll see how far you need to improve, and how below standard your current effort is. Also many excellent books from www.baronbarclay.com are available, and you can post questions here as well, for stuff not easily looked up in books and opinions on particular hands.

I am just an interrmediate player and as I said it's just a hobby. The program has sold thousands of copies however. I never spent much time at all on the tutorials, which are indended for absolute novices.

[RossSCann]

lalldonn, on 2012-July-07, 10:20, said:

I tell you what Ross, I'm no Frank Stewart but run it for 50 hands and post the results here, then send them to me and I'll look at them. jdonn@hotmail.com

I appreciate your interest and I will send you the file with all 500. You can look at any you wish.
Best,
Ross

[RossSCann]

Stephen Tu, on 2012-July-06, 18:40, said:

What exactly do you mean by "tops", do you just mean NT outscores the major on a MP basis? Did you take into account that North is taking preference to 2 spades fairly often, not passing 2 hearts? Are you comparing mp score of the resulting major partial vs. NT? How are you scoring the hand, do you have a double-dummy engine you wrote or got from somewhere else?

What were your parameters for opener rebidding 1nt? (does north raise to 2s with wk nt with three spades and outside small doubleton rather than bid 1nt? Does north ever bid 1nt with a singleton spade?) These kind of things can affect the outcome of your sim.

Lots of us use dealing programs and double dummy analysis to figure out whether borderline decisions are correct or not, in situations which are unclear. Less common for columnists. Generally bridge columnists tend to make their recommendations based on their personal experience of what works or not, or their knowledge of what other experts tend to do, which also tends to be based on what works or not.

It's pretty automatic for people to bid 2 hearts on this sort of hand. If you feel people are wrong, that passing 1nt is better, then present your results. But you have to give more complete data. Is this MP or IMPs? What were parameters for the opposite hand? Are you sure you are scoring each hand vs. the correct major partial (i.e. if opp hand is 3-3 in majors, you want to score 2s vs. 1nt, not 2h vs. 1nt)?

I would be happy to email you my results. When you say people do you mean interrmediate players? N would raise to 2 S's with 3 to A or K. I am using the book "Standard Bidding with SAYC" for much of the bidding in the program, which has that agreement. We are talking about intermediate bidding and play here.

[Stephen Tu]

RossSCann, on 2012-July-07, 17:59, said:

I would be happy to email you my results. When you say people do you mean interrmediate players? N would raise to 2 S's with 3 to A or K. I am using the book "Standard Bidding with SAYC" for much of the bidding in the program, which has that agreement. We are talking about intermediate bidding and play here.

When I say people I mean anyone who has been taught correct basic bidding, including beginners who have gotten up to what responder's rebids mean. Bidding 2h is completely standard.

Now that you have given more data, your problem is evident. You are under the mistaken impression that 2H is forcing in standard bidding. It is not. New suits by responder are forcing in standard in most situations, but NOT AFTER 1NT REBID. After 2h rebid, the only things opener is allowed to do is pass 2h (with 4h, or 1-3 in the majors if he is allowed to rebid 1nt on a stiff in responder's suit; this is subject to partnership agreement), or return to 2s (all hands with 2-3 spades and not 4 hearts) which responder will pass. Most int+ players do play conventional checkback mechanisms over 1nt, playing 1 or both of 2c/2d as artificial and forcing over 1nt, but 2H is universally non-forcing, allowing one to escape in a cheap 2M partial when responder is 5-5/5-4/6-4 in the majors. There is a slight flaw in that with 2-3 in the majors, opener does normally bid 2s which may be worse vs. pass if responder is 5-5, but tends to be better opposite 5-4/6-4. Can't cater perfectly to everything in std schemes.

So all your results with contracts reached other than 2h and 2s are totally bogus. Also it seems you have some weird hands for North, with 15-16 hcp (shouldn't happen, would have opened 1nt), 4 card spades (towards bottom of your edited first post), and a couple with 5-5 minors (would have rebid 2c, bottom of your post). Also, it's not clear if you are analyzing results on double-dummy basis, or based on simulated playout of the hand by non-double-dummy computer engines. Did you use someone else's DD engine, did you write your own, or are you having computers play out the hands only seeing single dummy, and who wrote the play engine in that case?

Your program, IMO, is too hard for novices (who should use "Learn to Play Bridge", parts I and II, from the ACBL, instead). They need a lot more step-by-step guidance, and hands as complicated as you present should not be in a "tutorial" mode. And the analysis/recommended bids & plays contain far too many errors to be useful for aspiring intermediate players, who should be reading books (Root, Kantar, Lawrence, others), using bridgebase software (Bridge Master, other titles), and/or looking to be mentored or taking lessons from better players. Hope the thousands who bought your program didn't assume the advice was correct, or at least eventually figured that out! Your program should come with disclaimer IMO "bridge advice from intermediate- player".

[RossSCann]

Stephen Tu, on 2012-July-07, 23:09, said:

When I say people I mean anyone who has been taught correct basic bidding, including beginners who have gotten up to what responder's rebids mean. Bidding 2h is completely standard.

Now that you have given more data, your problem is evident. You are under the mistaken impression that 2H is forcing in standard bidding. It is not. New suits by responder are forcing in standard in most situations, but NOT AFTER 1NT REBID. After 2h rebid, the only things opener is allowed to do is pass 2h (with 4h, or 1-3 in the majors if he is allowed to rebid 1nt on a stiff in responder's suit; this is subject to partnership agreement), or return to 2s (all hands with 2-3 spades and not 4 hearts) which responder will pass. Most int+ players do play conventional checkback mechanisms over 1nt, playing 1 or both of 2c/2d as artificial and forcing over 1nt, but 2H is universally non-forcing, allowing one to escape in a cheap 2M partial when responder is 5-5/5-4/6-4 in the majors. There is a slight flaw in that with 2-3 in the majors, opener does normally bid 2s which may be worse vs. pass if responder is 5-5, but tends to be better opposite 5-4/6-4. Can't cater perfectly to everything in std schemes.

So all your results with contracts reached other than 2h and 2s are totally bogus. Also it seems you have some weird hands for North, with 15-16 hcp (shouldn't happen, would have opened 1nt), 4 card spades (towards bottom of your edited first post), and a couple with 5-5 minors (would have rebid 2c, bottom of your post). Also, it's not clear if you are analyzing results on double-dummy basis, or based on simulated playout of the hand by non-double-dummy computer engines. Did you use someone else's DD engine, did you write your own, or are you having computers play out the hands only seeing single dummy, and who wrote the play engine in that case?

Your program, IMO, is too hard for novices (who should use "Learn to Play Bridge", parts I and II, from the ACBL, instead). They need a lot more step-by-step guidance, and hands as complicated as you present should not be in a "tutorial" mode. And the analysis/recommended bids & plays contain far too many errors to be useful for aspiring intermediate players, who should be reading books (Root, Kantar, Lawrence, others), using bridgebase software (Bridge Master, other titles), and/or looking to be mentored or taking lessons from better players. Hope the thousands who bought your program didn't assume the advice was correct, or at least eventually figured that out! Your program should come with disclaimer IMO "bridge advice from intermediate- player".

My program knows it is not forcing and N is free to pass as a preference with much better H's than S's and a dead minimum open.

[antonylee]

RossSCann, on 2012-July-08, 12:59, said:

My program knows it is not forcing and N is free to pass as a preference with much better H's than S's and a dead minimum open.

Then why is a 2♥ contract never reached according to the data you posted?

[Stephen Tu]

antonylee, on 2012-July-08, 13:16, said:

Then why is a 2♥ contract never reached according to the data you posted?

+1. Also, what's up with all the 2nt, 3 level, and 4h results you posted?

In standard bidding, after 2h is bid, the only possible contracts are 2h and 2s. Either North passes 2h, or he takes a 2s preference which South passes. Anything else, your program is bidding in a very non-std fashion, leading to minus scores and the faultiness of your initial conclusion.

Also it reads as if you are just having the program play out the hand, I assume not double-dummy accurate? If the program's play and defense is as bad as its bidding, then conclusions about the # of tricks taken in the final contract are also suspect. Double-dummy results are at least something concrete, and comparison vs. studies of okbridge & BBO hands have shown that double-dummy results aren't too far off from what humans achieve in real-life. Posting single-dummy results is fine, but you should also post data on how much / how frequently your program deviates from the DD result, and how much on average it favors declarer or defense.

[nigel_k]

Exactly what Stephen said. 2♥ is not forcing and will lead to the hand being played in either 2♥ or 2♠. If you do the simulation again comparing a 1NT contract with either a 2♥ or 2♠ contract, I'll bet you find that bidding 2♥ is better.

[lalldonn]

Ross sent me the hands but the formatting was bad so it was hard to look. I looked at the first 10 and on half of them one opponent or the other had a clear overcall. It was too hard for me to analyze play the way the formatting was or to look for a pattern that would make me think the hands were configured wrong.