Comparing a 5-3 fit with a 4-4 fit

[jogs]

When boards are normalized, the 4-4 fit generates 8 tricks. While the 5-3 fit seems to generate 8 1/3 tricks. My studies suffer from small sample size. Tallying results by visually inspecting the boards is slow and error prone.
Still it is possible that the 5-3 fit plays better. The 5-3 fit has fewer flat joint hand patterns.

[peter cheung]

Post by jogs
When boards are normalized, the 4-4 fit generates 8 tricks. While the 5-3 fit seems to generate 8 1/3 tricks. My studies suffer from small sample size. Tallying results by visually inspecting the boards is slow and error prone.
Still it is possible that the 5-3 fit plays better. The 5-3 fit has fewer flat joint hand patterns.

5/3 fit overall will be better then 4/ 4 fit. its main advantage is it will lose less trump tricks.

[jogs]

Post by peter cheung

Post by jogs
When boards are normalized, the 4-4 fit generates 8 tricks. While the 5-3 fit seems to generate 8 1/3 tricks. My studies suffer from small sample size. Tallying results by visually inspecting the boards is slow and error prone.
Still it is possible that the 5-3 fit plays better. The 5-3 fit has fewer flat joint hand patterns.

5/3 fit overall will be better then 4/ 4 fit. its main advantage is it will lose less trump tricks.

Don't know about losing less trump tricks. The 5/3 fit guarantees that 5 tricks will be won by a trump. Not necessarily won by us. The 4/4 fit can only guarantee 4 tricks will be won by a trump.
Peter, aren't you able to test this? Don't you have a massive amount of data history in digital and programmable form? Is the 6/2 fit another 1/3 of a trick better?

[t***@att.net]

4-4 does not guarantee that only four tricks can be won by trump. It could be as high as 8 (or 13 with sufficient voidage).

[p***@infi.net]

Post by t***@att.net
4-4 does not guarantee that only four tricks can be won by trump. It could be as high as 8 (or 13 with sufficient voidage).

That isn't what he said: "only guarantee" and "guarantee only" are quite different. When a player has four trumps, at least four tricks must be won by trumps. When a player has five trumps, at least five trucks must be won by trumps, etc.

[peter cheung]

I have not set up my software for many years now. If you use just common sense length from 5 on is at least equal to one hcp and that is equal to about .3 of a trick.

[smn]

Post by jogs
When boards are normalized, the 4-4 fit generates 8 tricks. While the 5-3 fit seems to generate 8 1/3 tricks. My studies suffer from small sample size. Tallying results by visually inspecting the boards is slow and error prone.
Still it is possible that the 5-3 fit plays better. The 5-3 fit has fewer flat joint hand patterns.

Hi , the conventional wisdom is that the 4-4 fit often plays 1 trick better but the 5-3 is safer . Some slams can be made by choosing 4-4 rather then 5-3 but sometimes a hand will make at 5-3 and go down at 4-4 .Of course if a hand has only one of these available,who cares. amn

[jogs]

Post by smn
Hi , the conventional wisdom is that the 4-4 fit often plays 1 trick better but the 5-3 is safer . Some slams can be made by choosing 4-4 rather then 5-3 but sometimes a hand will make at 5-3 and go down at 4-4 .Of course if a hand has only one of these available,who cares. amn

I wasn't investigating boards that contain both a 5/3 and 4/4 fit. Was studying boards with 5/3 fit against boards with a 4/4 fit. These were a separate set of boards. Also was only interested in boards where our side had between 17 to 25 HCP. Purposely excluded slam range boards. That's for another study.

[smn]

Post by jogs

Post by smn
Hi , the conventional wisdom is that the 4-4 fit often plays 1 trick better but the 5-3 is safer . Some slams can be made by choosing 4-4 rather then 5-3 but sometimes a hand will make at 5-3 and go down at 4-4 .Of course if a hand has only one of these available,who cares. amn

I wasn't investigating boards that contain both a 5/3 and 4/4 fit. Was studying boards with 5/3 fit against boards with a 4/4 fit. These were a separate set of boards. Also was only interested in boards where our side had between 17 to 25 HCP. Purposely excluded slam range boards. That's for another study.

Hi ,ok , sorry .Try a study restricted to hands for your side of 23-25 hcp .You should find 4-4 is more likely to make game . However an 8 card fit .particularly if opponents have at most an 8 card fit does not generate a lot of total tricks compared to a 9 card fit .So stretching to game with an 8 card fit is more risky in general. Buy what really do I know .Regards smn

[KWSchneider]

Hi , the conventional wisdom is that the 4-4 fit often plays 1 trick be=

tter but the 5-3 is safer . Some slams can be made by choosing 4-4 rather t=
hen 5-3 but sometimes a hand will make at 5-3 and go down at 4-4 .Of course=
if a hand has only one of these available,who cares. amn

=20
I wasn't investigating boards that contain both a 5/3 and 4/4 fit. Was s=

tudying boards with 5/3 fit against boards with a 4/4 fit. These were a sep=
arate set of boards. Also was only interested in boards where our side had=
between 17 to 25 HCP. Purposely excluded slam range boards. That's for a=
nother study.
Hi ,ok , sorry .Try a study restricted to hands for your side of 23-25 hcp =
.You should find 4-4 is more likely to make game . However an 8 card fit .p=
articularly if opponents have at most an 8 card fit does not generate a lot=
of total tricks compared to a 9 card fit .So stretching to game with an 8 =
card fit is more risky in general. Buy what really do I know .Regards smn

10,000 hand DD simulation - 44 in spades, 53 in hearts. Dealer (South) has 13HCP, dummy has 11HCP.

44 fit = 9.82 tricks, sDev = 1.07, Success = 62.4%
53 fit = 9.69 tricks, sDev = 1.03, Success = 58.4%

Kurt

[KWSchneider]

Post by KWSchneider
10,000 hand DD simulation - 44 in spades, 53 in hearts. Dealer (South) has 13HCP, dummy has 11HCP.
44 fit = 9.82 tricks, sDev = 1.07, Success = 62.4%
53 fit = 9.69 tricks, sDev = 1.03, Success = 58.4%

Another one - 44 in spades, 62 in hearts. Dealer (South) has 13HCP, dummy has 11HCP.

44 fit = 10.3 tricks, sDev = 1.04, Success = 79.2%
62 fit = 10.1 tricks, sDev = 1.03, Success = 74.5%
44fit is 0.22 tricks better than 62fit.

Interesting to note that simply making the south hand 46[xy] instead of 45[xy], and north 42[xy] instead of 43[xy] increased the result by ~0.5 tricks, a huge benefit.

Kurt

[jogs]

Post by KWSchneider

Post by KWSchneider
10,000 hand DD simulation - 44 in spades, 53 in hearts. Dealer (South) has 13HCP, dummy has 11HCP.
44 fit = 9.82 tricks, sDev = 1.07, Success = 62.4%
53 fit = 9.69 tricks, sDev = 1.03, Success = 58.4%

Another one - 44 in spades, 62 in hearts. Dealer (South) has 13HCP, dummy has 11HCP.
44 fit = 10.3 tricks, sDev = 1.04, Success = 79.2%
62 fit = 10.1 tricks, sDev = 1.03, Success = 74.5%
44fit is 0.22 tricks better than 62fit.
Interesting to note that simply making the south hand 46[xy] instead of 45[xy], and north 42[xy] instead of 43[xy] increased the result by ~0.5 tricks, a huge benefit.
Kurt

If you lower responder's hand to 7 HCP keeping dealer at 13 HCP, my experience is the 6-2 fit plays better.
Also I think DD the 4-4 fit bias favors declarer. Real live declarers often have problems handling some 4-4 holdings. Assume declarer holds only 2 of the 4 honors. There are 6 possible combinations. AK, AQ, AJ, KQ, KJ, and QJ. 4 of those 6 are difficult to play.
Kxxx // Jxxx Live declarers lose 3 or more trump tricks much more often than DD declarers.

[jogs]

On 25 Jul 2016 05:34 PM ,"

Post by KWSchneider
10,000 hand DD simulation - 44 in spades, 53 in hearts. Dealer (South) has 13HCP, dummy has 11HCP.
44 fit = 9.82 tricks, sDev = 1.07, Success = 62.4%
53 fit = 9.69 tricks, sDev = 1.03, Success = 58.4%

Another one - 44 in spades, 62 in hearts. Dealer (South) has 13HCP, dummy has 11HCP.
44 fit = 10.3 tricks, sDev = 1.04, Success = 79.2%
62 fit = 10.1 tricks, sDev = 1.03, Success = 74.5%
44fit is 0.22 tricks better than 62fit.
Interesting to note that simply making the south hand 46[xy] instead of 45[xy], and north 42[xy] instead of 43[xy] increased the result by ~0.5 tricks, a huge benefit.
Kurt

How does these simulations work? Can you set the starting seed for the random number generator? Can you run 100 trials DD and then rerun the same 100 trials SD?
In certain 4-4 trump fits, I think without proof that DD declarers navigate the suit much better than SD declarers. And SD declarers are playing closer to human declarers. Only SD declarers are AX dummy players. Human declarers from online hand histories are absolute beginners to world champs. Also assuming the programs are well written SD declarers play extremely well.

[KWSchneider]

How does these simulations work? Can you set the starting seed for the ran=
dom number generator? Can you run 100 trials DD and then rerun the same 10=
0 trials SD?
In certain 4-4 trump fits, I think without proof that DD declarers navigate=
the suit much better than SD declarers. And SD declarers are playing clos=
er to human declarers. Only SD declarers are AX dummy players. Human decl=
arers from online hand histories are absolute beginners to world champs. A=
lso assuming the programs are well written SD declarers play extremely well=

1) I can run SD and DD simultaneously on the same hands.
2) SD declarer can be made to be various levels - from beginner to better than AX. I can't vary the capability during a simulation unless I run a third concurrent simulation with a differentiated SD declarer.

Kurt

[jogs]

Post by KWSchneider
1) I can run SD and DD simultaneously on the same hands.
2) SD declarer can be made to be various levels - from beginner to better than AX. I can't vary the capability during a simulation unless I run a third concurrent simulation with a differentiated SD declarer.
Kurt

I'm convinced SD is closer to real ppl results. Use the strongest SD available.
4-4 vs 5-3
With combined 20HCP 5-3 fit should have higher expected tricks.
With combined 32HCP 4-4 fit should have higher expected tricks.

The question is where do the two line cross?

[KWSchneider]

Post by jogs

Post by KWSchneider
1) I can run SD and DD simultaneously on the same hands.
2) SD declarer can be made to be various levels - from beginner to better than AX. I can't vary the capability during a simulation unless I run a third concurrent simulation with a differentiated SD declarer.
Kurt

I'm convinced SD is closer to real ppl results. Use the strongest SD available.
4-4 vs 5-3
With combined 20HCP 5-3 fit should have higher expected tricks.
With combined 32HCP 4-4 fit should have higher expected tricks.
The question is where do the two line cross?

Interesting Results...
44spade and 53 heart fits. 1000 simulations, same hands played 4times. DD44, DD53, SD44, SD53

13vs11
DD44 = 9.81tricks, sDev = 1.08
DD53 = 9.65, 1.04
SD44 = 9.92, 1.10
SD53 = 9.61, 1.06

So DD predicts a 0.16 trick advantage to a 44fit, whereas SD predicts double, 0.31 trick. What is more interesting is that while there was the expected DD advantage on defense for the 44fit, there was none (in fact a slight benefit to declarer) for the 53fit. (I'm guessing that far too few of us lead trump against 44fits...)

16vs14
DD44 = 11.70, 0.87
DD53 = 11.59, 0.84
SD44 = 11.70, 0.90
SD53 = 11.49, 0.89

While the advantage diminished, this confirms that 44 is still better than 53, even when the overarching strength is with the declaring side. And as expected, the defensive advantage for DD is decreasing as declarer's assets increase.

This seems to indicate that defending a 44fit is more difficult in real life than a 53 fit - and this is exacerbated by weakening declarer.

I'm currently running an 11vs9 test and will report tonight.

Kurt

[KWSchneider]

Post by KWSchneider
Interesting Results...
44spade and 53 heart fits. 1000 simulations, same hands played 4times. DD44, DD53, SD44, SD53
13vs11
DD44 = 9.81tricks, sDev = 1.08
DD53 = 9.65, 1.04
SD44 = 9.92, 1.10
SD53 = 9.61, 1.06
So DD predicts a 0.16 trick advantage to a 44fit, whereas SD predicts double, 0.31 trick. What is more interesting is that while there was the expected DD advantage on defense for the 44fit, there was none (in fact a slight benefit to declarer) for the 53fit. (I'm guessing that far too few of us lead trump against 44fits...)
16vs14
DD44 = 11.70, 0.87
DD53 = 11.59, 0.84
SD44 = 11.70, 0.90
SD53 = 11.49, 0.89
While the advantage diminished, this confirms that 44 is still better than 53, even when the overarching strength is with the declaring side. And as expected, the defensive advantage for DD is decreasing as declarer's assets increase.
This seems to indicate that defending a 44fit is more difficult in real life than a 53 fit - and this is exacerbated by weakening declarer.
I'm currently running an 11vs9 test and will report tonight.

11vs9 results
DD44 = 8.33, 1.10
DD53 = 8.19, 1.08
SD44 = 8.60, 1.19
SD53 = 8.22, 1.08

Conclusions - 44vs53 over 20-30hcp range
1) 44 is always better than 53, independent of combined strength of declaring hands. This varies from 0.2tricks at slam level hands to 0.4tricks at part score levels.
2) 53 fit is much less impacted by SD analysis than 44 fit - ergo defense plays much more of a part in 44 fit results. Stated differently - as declarer it is much easier to match DD results playing a 53 fit and you expect to do so until you reach the slam level. Conversely, declarer should be exceeding DD results when playing a 44 fit until reaching the slam level, where you expect to match them.
3) 53 fit slightly favors defenders - ie DD results are equal or better than SD results, independent of combined strength of declaring hands. The "benefit" varies from none to 0.1tricks at slam levels.
4) 44 fit significantly favors declarer - the SD "benefit" varies from none to 0.27tricks at part score levels.

Kurt

[jogs]

Post by KWSchneider

Post by KWSchneider
I'm currently running an 11vs9 test and will report tonight.

11vs9 results
SD44 = 8.60, 1.19
SD53 = 8.22, 1.08
Conclusions - 44vs53 over 20-30hcp range
1) 44 is always better than 53, independent of combined strength of declaring hands. This varies from 0.2tricks at slam level hands to 0.4tricks at part score levels.
2) 53 fit is much less impacted by SD analysis than 44 fit - ergo defense plays much more of a part in 44 fit results. Stated differently - as declarer it is much easier to match DD results playing a 53 fit and you expect to do so until you reach the slam level. Conversely, declarer should be exceeding DD results when playing a 44 fit until reaching the slam level, where you expect to match them.
3) 53 fit slightly favors defenders - ie DD results are equal or better than SD results, independent of combined strength of declaring hands. The "benefit" varies from none to 0.1tricks at slam levels.
4) 44 fit significantly favors declarer - the SD "benefit" varies from none to 0.27tricks at part score levels.
Kurt
11vs9 results
SD44 = 8.60, 1.19
SD53 = 8.22, 1.08

I'm surprised by your results. But you didn't preform the same test as I.

My tests were on 4-4 fit and a separate test on the 5-3 fit.
The sample on the 4-4 fit contained many boards where neither partner held a 5 card suit, hence no additional source of tricks.
4432 // 4333 The mean for this was about 7.5 trks/bd.

[KWSchneider]

Post by jogs
I'm surprised by your results. But you didn't preform the same test as I.
My tests were on 4-4 fit and a separate test on the 5-3 fit.
The sample on the 4-4 fit contained many boards where neither partner held =
a 5 card suit, hence no additional source of tricks.
4432 // 4333 The mean for this was about 7.5 trks/bd.

I'm not sure as to the relevance of your test. The whole concept was to determine, when faced with the 53 vs 44 decision, which one was the "best" over the long haul.

Why do we want to know whether a 44 fit is better than a 53 fit, when we have one or the other and we have no decision to make? Or are you saying that you might "push" harder for a tight contract, more so with one vs the other?

Kurt

[jogs]

On Friday, August 5, 2016 at 1:07:37 PM UTC-7,

Post by KWSchneider
I'm not sure as to the relevance of your test. The whole concept was to determine, when faced with the 53 vs 44 decision, which one was the "best" over the long haul.
Why do we want to know whether a 44 fit is better than a 53 fit, when we have one or the other and we have no decision to make? Or are you saying that you might "push" harder for a tight contract, more so with one vs the other?
Kurt

No, I started this thread, not you. I want to know the expected tricks for a 4-4 fit. Also the expected tricks for a 5-3 fit. Boards may or may not contain both a 4-4 and 5-3 fit. They usually don't.
I concluded the 5-3 fit had a higher expected value. But my sample size was small.

[KWSchneider]

Post by jogs
No, I started this thread, not you. I want to know the expected tricks for a 4-4 fit. Also the expected tricks for a 5-3 fit. Boards may or may not contain both a 4-4 and 5-3 fit. They usually don't.
I concluded the 5-3 fit had a higher expected value. But my sample size was small.

Fair enough...
No restrictions on point count - 1000 sims, SD and DD, 53 and 44 separately.

DD44 -> 8.07 tricks, sDev = 2.01
SD44 -> 8.35, 1.83

DD53 -> 8.19, 2.11
SD53 -> 8.45, 1.91

So your conclusion over the broad spectrum of all point counts is correct. Interesting that SD had better sDev than DD for both cases - and that sDev for 53 is higher than for 44.

Kurt

[jogs]

Post by KWSchneider

Post by jogs
No, I started this thread, not you. I want to know the expected tricks for a 4-4 fit. Also the expected tricks for a 5-3 fit. Boards may or may not contain both a 4-4 and 5-3 fit. They usually don't.
I concluded the 5-3 fit had a higher expected value. But my sample size was small.

Fair enough...
No restrictions on point count - 1000 sims, SD and DD, 53 and 44 separately.
DD44 -> 8.07 tricks, sDev = 2.01
SD44 -> 8.35, 1.83
DD53 -> 8.19, 2.11
SD53 -> 8.45, 1.91
So your conclusion over the broad spectrum of all point counts is correct. Interesting that SD had better sDev than DD for both cases - and that sDev for 53 is higher than for 44.
Kurt

My studies are from hand histories. Therefore often or sometimes with a 4-4 spades fit the final contract is in some other strain. 9 cards in hearts means the final contract would usually be in hearts. It may require 10 cards in a minor for the final contract to be in that minor.
My sDev was lower than yours. Probably because many of those boards was played in the strain with more trumps. Also I normalized the results. Meaning if our side had 23 HCP, I subtracted 1 trick from the outcome.
1/3 of a trick for each HCP over 20.

jogs

[KWSchneider]

Post by KWSchneider
=20

No, I started this thread, not you. I want to know the expected tricks=

for a 4-4 fit. Also the expected tricks for a 5-3 fit. Boards may or may=
not contain both a 4-4 and 5-3 fit. They usually don't.

Post by KWSchneider

I concluded the 5-3 fit had a higher expected value. But my sample siz=

e was small.

Post by KWSchneider
=20
Fair enough...
No restrictions on point count - 1000 sims, SD and DD, 53 and 44 separate=

ly.

Post by KWSchneider
=20
DD44 -> 8.07 tricks, sDev =3D 2.01
SD44 -> 8.35, 1.83
=20
DD53 -> 8.19, 2.11
SD53 -> 8.45, 1.91
=20
So your conclusion over the broad spectrum of all point counts is correct=

. Interesting that SD had better sDev than DD for both cases - and that sDe=
v for 53 is higher than for 44.

Post by KWSchneider
=20
Kurt

My studies are from hand histories. Therefore often or sometimes with a 4-=
4 spades fit the final contract is in some other strain. 9 cards in hearts=
means the final contract would usually be in hearts. It may require 10 ca=
rds in a minor for the final contract to be in that minor.
My sDev was lower than yours. Probably because many of those boards was pl=
ayed in the strain with more trumps. Also I normalized the results. Meani=
ng if our side had 23 HCP, I subtracted 1 trick from the outcome.
1/3 of a trick for each HCP over 20.

More narrow simulation - 1000 sims, SD and DD, 53 and 44 separately, 13HCP declarer, 11HCP dummy.

DD44 -> 9.63 tricks, sDev = 1.05
SD44 -> 9.66, 1.10
DD53 -> 9.72, 1.05
SD53 -> 9.67, 1.12

Based on SD, 53 and 44 are identical for game going strengths.

Conclusion -> On a standalone basis, 53 better than 44 for part score hands, the same for game hands, and 44 better than 53 for slam hands.

Kurt

[jogs]

Post by KWSchneider
Based on SD, 53 and 44 are identical for game going strengths.
Conclusion -> On a standalone basis, 53 better than 44 for part score hands, the same for game hands, and 44 better than 53 for slam hands.
Kurt

That was what I thought before making any studies.

[p***@infi.net]

Post by KWSchneider

Post by KWSchneider
=20

No, I started this thread, not you. I want to know the expected tricks=

for a 4-4 fit. Also the expected tricks for a 5-3 fit. Boards may or may=
not contain both a 4-4 and 5-3 fit. They usually don't.

Post by KWSchneider

I concluded the 5-3 fit had a higher expected value. But my sample siz=

e was small.

Post by KWSchneider
=20
Fair enough...
No restrictions on point count - 1000 sims, SD and DD, 53 and 44 separate=

ly.

Post by KWSchneider
=20
DD44 -> 8.07 tricks, sDev =3D 2.01
SD44 -> 8.35, 1.83
=20
DD53 -> 8.19, 2.11
SD53 -> 8.45, 1.91
=20
So your conclusion over the broad spectrum of all point counts is correct=

. Interesting that SD had better sDev than DD for both cases - and that sDe=
v for 53 is higher than for 44.

Post by KWSchneider
=20
Kurt

My studies are from hand histories. Therefore often or sometimes with a 4-=
4 spades fit the final contract is in some other strain. 9 cards in hearts=
means the final contract would usually be in hearts. It may require 10 ca=
rds in a minor for the final contract to be in that minor.
My sDev was lower than yours. Probably because many of those boards was pl=
ayed in the strain with more trumps. Also I normalized the results. Meani=
ng if our side had 23 HCP, I subtracted 1 trick from the outcome.
1/3 of a trick for each HCP over 20.

More narrow simulation - 1000 sims, SD and DD, 53 and 44 separately, 13HCP declarer, 11HCP dummy.
DD44 -> 9.63 tricks, sDev = 1.05
SD44 -> 9.66, 1.10
DD53 -> 9.72, 1.05
SD53 -> 9.67, 1.12
Based on SD, 53 and 44 are identical for game going strengths.
Conclusion -> On a standalone basis, 53 better than 44 for part score hands, the same for game hands, and 44 better than 53 for slam hands.
Kurt

That's why I don't chase a 4-4 fit in the other major when we have a known 5-3 fit and no slam interest. But conversely, I despise the popular "limit notrump" style of bidding 3NT over 1M with 2344 and 13-15 points -- if opener has a good hand, we could have slam in either of those 4 card suits.

[[e-mail]]

Post by p***@infi.net

Post by KWSchneider

Post by KWSchneider
=20

No, I started this thread, not you. I want to know the expected tricks=

for a 4-4 fit. Also the expected tricks for a 5-3 fit. Boards may or may=
not contain both a 4-4 and 5-3 fit. They usually don't.

Post by KWSchneider

I concluded the 5-3 fit had a higher expected value. But my sample siz=

e was small.

Post by KWSchneider
=20
Fair enough...
No restrictions on point count - 1000 sims, SD and DD, 53 and 44 separate=

ly.

Post by KWSchneider
=20
DD44 -> 8.07 tricks, sDev =3D 2.01
SD44 -> 8.35, 1.83
=20
DD53 -> 8.19, 2.11
SD53 -> 8.45, 1.91
=20
So your conclusion over the broad spectrum of all point counts is correct=

. Interesting that SD had better sDev than DD for both cases - and that sDe=
v for 53 is higher than for 44.

Post by KWSchneider
=20
Kurt

My studies are from hand histories. Therefore often or sometimes with a 4-=
4 spades fit the final contract is in some other strain. 9 cards in hearts=
means the final contract would usually be in hearts. It may require 10 ca=
rds in a minor for the final contract to be in that minor.
My sDev was lower than yours. Probably because many of those boards was pl=
ayed in the strain with more trumps. Also I normalized the results. Meani=
ng if our side had 23 HCP, I subtracted 1 trick from the outcome.
1/3 of a trick for each HCP over 20.

More narrow simulation - 1000 sims, SD and DD, 53 and 44 separately, 13HCP declarer, 11HCP dummy.
DD44 -> 9.63 tricks, sDev = 1.05
SD44 -> 9.66, 1.10
DD53 -> 9.72, 1.05
SD53 -> 9.67, 1.12
Based on SD, 53 and 44 are identical for game going strengths.
Conclusion -> On a standalone basis, 53 better than 44 for part score hands, the same for game hands, and 44 better than 53 for slam hands.
Kurt

That's why I don't chase a 4-4 fit in the other major when we have a known 5-3 fit and no slam interest.

Except it may be a 5-4 fit in the other major.

Carl

[jogs]

Post by ***@verizon.net

Post by p***@infi.net

Post by KWSchneider
Based on SD, 53 and 44 are identical for game going strengths.
Conclusion -> On a standalone basis, 53 better than 44 for part score hands, the same for game hands, and 44 better than 53 for slam hands.
Kurt

That's why I don't chase a 4-4 fit in the other major when we have a known 5-3 fit and no slam interest.

Except it may be a 5-4 fit in the other major.
Carl

That did happen in a regional once. 5=5=1=2
1S - (3D) - 3S - (4D)
4S all pass
The correct bid was 4H. Partner had 3=4=2=4.
4H made while 4S failed. Nearly the entire field was in 4S.

[p***@infi.net]

Post by jogs

Post by ***@verizon.net

Post by p***@infi.net

Post by KWSchneider
Based on SD, 53 and 44 are identical for game going strengths.
Conclusion -> On a standalone basis, 53 better than 44 for part score hands, the same for game hands, and 44 better than 53 for slam hands.
Kurt

That's why I don't chase a 4-4 fit in the other major when we have a known 5-3 fit and no slam interest.

Except it may be a 5-4 fit in the other major.
Carl

That did happen in a regional once. 5=5=1=2
1S - (3D) - 3S - (4D)
4S all pass
The correct bid was 4H. Partner had 3=4=2=4.
4H made while 4S failed. Nearly the entire field was in 4S.

You mean the successful bid, given these specific hands. I would have to see opener's hand before deciding whether 4H might gain more than it loses -- and will it be clear to partner you are offering a choice of games, not simply control bidding? Responder had a choice as well -- double and then support spades if partner doesn't bid hearts. Again, I would have to see the actual hand before deciding between double and 4S. Double, of course, risks partner passing for an inadequate penalty.

[KWSchneider]

Post by KWSchneider

Conclusion -> On a standalone basis, 53 better than 44 for part score han=

ds, the same for game hands, and 44 better than 53 for slam hands.

=20
Kurt
=20

That's why I don't chase a 4-4 fit in the other major when we have a known =
5-3 fit and no slam interest. But conversely, I despise the popular "limit =
notrump" style of bidding 3NT over 1M with 2344 and 13-15 points -- if open=
er has a good hand, we could have slam in either of those 4 card suits.

Paul - but this is not what the results show. The conclusions I posted above are for STANDALONE (ie separate instances of 53 and 44 fits). If you reread my earlier posts in this thread, where I compare 53 and 44 fits IN THE SAME HANDS, the 44 fit is always better. As jogs points out, the necessary shape benefits of having the 53 fit (for discards and shortness in the 45xy hand) with the 44 fit makes the 44 fit superior, across ALL ranges of strength - being significantly better (1/3 trick) in part score hands.

Kurt

[p***@infi.net]

Post by KWSchneider

Post by KWSchneider

Conclusion -> On a standalone basis, 53 better than 44 for part score han=

ds, the same for game hands, and 44 better than 53 for slam hands.

=20
Kurt
=20

That's why I don't chase a 4-4 fit in the other major when we have a known =
5-3 fit and no slam interest. But conversely, I despise the popular "limit =
notrump" style of bidding 3NT over 1M with 2344 and 13-15 points -- if open=
er has a good hand, we could have slam in either of those 4 card suits.

Paul - but this is not what the results show. The conclusions I posted above are for STANDALONE (ie separate instances of 53 and 44 fits). If you reread my earlier posts in this thread, where I compare 53 and 44 fits IN THE SAME HANDS, the 44 fit is always better. As jogs points out, the necessary shape benefits of having the 53 fit (for discards and shortness in the 45xy hand) with the 44 fit makes the 44 fit superior, across ALL ranges of strength - being significantly better (1/3 trick) in part score hands.
Kurt

Fair enough. 1/3rd trick is approximately 1 point, so if partner opens 1H and you have 4 spades and 3 hearts, you can gain a point if you bid spades and partner raises (and has 4 trumps, not 3.) Odds of partner having 4 spades are about 1 in 3 a priori, surely less given his 5+ hearts. So, with a hand worth a minimum raise, searching for the 4-4 spade fit doesn't seem worthwhile. In some cases, playing constructive raises, you would find a 4-4 heart fit after a 1S opening and 1NT response on a hand too weak for an immediate 2S.

[Steve Willner]

This is a really nice study! I've rearranged some text.

Post by KWSchneider

Post by KWSchneider
44spade and 53 heart fits. 1000 simulations, same hands played
4times. DD44, DD53, SD44, SD53

11vs9 results
DD44 = 8.33, 1.10
SD44 = 8.60, 1.19
DD53 = 8.19, 1.08
SD53 = 8.22, 1.08

Post by KWSchneider
13vs11
DD44 = 9.81tricks, sDev = 1.08
SD44 = 9.92, 1.10
DD53 = 9.65, 1.04
SD53 = 9.61, 1.06
16vs14
DD44 = 11.70, 0.87
SD44 = 11.70, 0.90
DD53 = 11.59, 0.84
SD53 = 11.49, 0.89
This seems to indicate that defending a 44fit is more difficult in
real life than a 53 fit - and this is exacerbated by weakening
declarer.

This makes a lot of sense. A 4-4 fit, with a 5-3 on the side, might
play well either by cross-ruffing or by establishing the 5c suit. The
opening leader is often on a guess which one to defend against. Leading
trumps is vital on the first type but just loses a tempo on the second type.

Post by KWSchneider
Conclusions - 44vs53 over 20-30hcp range
1) 44 is always better than 53, independent of combined strength of
declaring hands. This varies from 0.2tricks at slam level hands to
0.4tricks at part score levels.
2) 53 fit is much less impacted by SD analysis than 44 fit - ergo
defense plays much more of a part in 44 fit results.
3) 53 fit slightly favors defenders

In a 5-3, side losers can't be discarded, so it's mostly a matter of
whether declarer guesses well or not. Defenders should pretty much
always lead trumps if they know the situation.

Post by KWSchneider
4) 44 fit significantly favors declarer - the SD "benefit" varies
from none to 0.27tricks at part score levels.

[Jeff Miller]

Post by KWSchneider
1) I can run SD and DD simultaneously on the same hands.
2) SD declarer can be made to be various levels - from beginner to better than AX.

Kurt,

Just out of curiosity, what program are you using to do the SD simulations, and can you vary the strength level of the declarer independently of the strength level of the defenders (i.e., SD simulate weak declarer against strong defenders and vice versa)?

Thanks,

[KWSchneider]

Post by Jeff Miller

Post by KWSchneider
1) I can run SD and DD simultaneously on the same hands.
2) SD declarer can be made to be various levels - from beginner to better than AX.

Kurt,
Just out of curiosity, what program are you using to do the SD simulations, and can you vary the strength level of the declarer independently of the strength level of the defenders (i.e., SD simulate weak declarer against strong defenders and vice versa)?
Thanks,

I use a commercial, purchased version of the GIB engine - which handles both SD and DD - and custom TCL subroutines (to interface with Thomas Andrew's Deal) that I've written.

GIB allows a significant number of switches - one of which allows for separate defensive and offensive capability. It allows the defense and offense to separately initiate more (or less) simulations to analyze and pick a play (250 is max). So 10 sims might be considered weak and 200 sims strong (20x longer to make a play).

I did a number of tests years ago when I wrote the interfaces to determine the lowest number of simulations that resulted in what I thought were "acceptable" AX defense and play decisions, recognizing that each play (all four members of a deal) required a measurable amount of time to pick a play. I settled on 25 sims for offense and 50 sims for defense for matchplay (IMPS has different decisions due to safety plays, etc). HOWEVER, until I was into very low sim counts, was I able to determine any difference in results. This indicates that even basic logic will get the job done most of the time.

Kurt

[Lorne Anderson]

Post by jogs
When boards are normalized, the 4-4 fit generates 8 tricks. While the 5-3 fit seems to generate 8 1/3 tricks. My studies suffer from small sample size. Tallying results by visually inspecting the boards is slow and error prone.
Still it is possible that the 5-3 fit plays better. The 5-3 fit has fewer flat joint hand patterns.

Not certain what exactly you want to do but if it helps the results from
1000 random deals with N+S biased to have 0-4 cards in each of 3 suits,
S with 9-13 points and N with 8-12 and in the first case exactly 4 cards
each in the other suit, or in the second case 5 in S and 3 in N for the
other suit show the following tricks:

Tricks 3 4 5 6 7 8 9 10 11 12
4-4 1 2 13 78 226 313 254 94 18 1 ave 8.066
5-3 0 0 13 76 193 323 260 116 19 0 ave 8.165

NB: do not use this to decide 5-3 fits play better than 4-4 when both
are available for the same deal. When both deals are available for the
same deal the results reverse and 4-4 plays better. The above are for 2
separate runs as per the info in the OP.

[t***@att.net]

"Morehead on Bidding" has a good discussion of 4-4 vs 5-3. The 5-3 fit is superior in cases where the short hand can ruff or if there is another 5-3 fit for discards or if the 4-4 alternative is too weak to be a good trump suit (among others.)

The 4-4 fit can often generate one more trump trick (and sometime two) because having 4 trumps generally indicates more shortness in side suits than having 3 trumps. A cross-ruff may play better in a 4-4 fit than in a 5-3 (though I'd guess not always.)

Sometimes one finds that a 5-3 fit may be safer for game but a 4-4 fit may be the only way to make a slam.

Every hand is an adventure.