Kurt: questions on DD and SD analysis

Post by jogs
Can these programs be told the game is imps? Can one specify the final contract?
Run a sample of 100 deals. Contract is 3NT. Run the same 100 deals again. Contract is 4NT. Will declarer and defenders use a different line of play? Will the table of tricks made be different for the 3NT sample and the 4NT sample?
TIA,
jogs

What is single dummy analysis?

GIB robots playing the hand?
What is the objective?
More precise results reflecting at the table experience?
You must be joking.
Anyone who has played with GIB robots knows what I am talking about.
Not that GIB robots are entirely playing bad Bridge.
The problem is rather they are unbalanced.
They excel at some aspects of the game.
For example they never miscount a hand, something most human are known to have done form time to time, not to mention the humans who never do.
They also are good at finding the only chance making a contract.
But at other aspects they are worse than any human past the novice stage.

I have made more than once contracts against GIB robots no one would let me make at the table.

No thanks, I prefer double dummy analysis.
I know the deficiencies, but at least it is a reliable, fast and proven method.
Saying single dummy is superior, usually means overlooking the practical problems arising from analyzing large number of deals and the deficiencies which crop when doing so.

The true answer is robots are no match for simulating human mind so far.
Note: Even if they will play better than humans at some stage it does not follow that they will be useful to predict what would happen at the bridge table with humans sitting there. .

p***@infi.net

2016-08-13 14:57:39 UTC

Post by rhm

What is single dummy analysis?
GIB robots playing the hand?
What is the objective?
More precise results reflecting at the table experience?
You must be joking.
Anyone who has played with GIB robots knows what I am talking about.
Not that GIB robots are entirely playing bad Bridge.
The problem is rather they are unbalanced.
They excel at some aspects of the game.
For example they never miscount a hand, something most human are known to have done form time to time, not to mention the humans who never do.
They also are good at finding the only chance making a contract.
But at other aspects they are worse than any human past the novice stage.
I have made more than once contracts against GIB robots no one would let me make at the table.
No thanks, I prefer double dummy analysis.
I know the deficiencies, but at least it is a reliable, fast and proven method.
Saying single dummy is superior, usually means overlooking the practical problems arising from analyzing large number of deals and the deficiencies which crop when doing so.
The true answer is robots are no match for simulating human mind so far.
Note: Even if they will play better than humans at some stage it does not follow that they will be useful to predict what would happen at the bridge table with humans sitting there. .

GIB's skill level depends on how much time you give them. I do agree that it seems unlikely to completely replicate human behavior. But it does avoid some of the known problems with DD analysis, such as always finding the killing lead, never making a poor lead, always finding a Queen, dropping a stiff honor, etc.

KWSchneider

2016-08-16 02:51:20 UTC

GIB's skill level depends on how much time you give them. I do agree that i=
t seems unlikely to completely replicate human behavior. But it does avoid =
some of the known problems with DD analysis, such as always finding the kil=
ling lead, never making a poor lead, always finding a Queen, dropping a sti=
ff honor, etc.

Other than attempts at slipping a trick by an unwary defender, I will put GIB up against any human on the play of the hand when properly configured.

DD overvalues honors, undervalues trump, overvalues defense, undervalues the declarer, and always finds the killing lead. Other than that, it is a perfect method of simulation. Oh - and it drops ALL offside honors, and uses deuces to drop singleton Aces...

Kurt

--
Posted by Mimo Usenet Browser v0.2.5
http://www.mimousenet.com/mimo/post

KWSchneider

2016-08-16 02:46:13 UTC

Post by rhm

What is single dummy analysis?
GIB robots playing the hand?
What is the objective?
More precise results reflecting at the table experience?
You must be joking.
Anyone who has played with GIB robots knows what I am talking about.
Not that GIB robots are entirely playing bad Bridge.
The problem is rather they are unbalanced.
They excel at some aspects of the game.
For example they never miscount a hand, something most human are known to have done form time to time, not to mention the humans who never do.
They also are good at finding the only chance making a contract.
But at other aspects they are worse than any human past the novice stage.
I have made more than once contracts against GIB robots no one would let me make at the table.
No thanks, I prefer double dummy analysis.
I know the deficiencies, but at least it is a reliable, fast and proven method.
Saying single dummy is superior, usually means overlooking the practical problems arising from analyzing large number of deals and the deficiencies which crop when doing so.
The true answer is robots are no match for simulating human mind so far.
Note: Even if they will play better than humans at some stage it does not follow that they will be useful to predict what would happen at the bridge table with humans sitting there. .

Rainer - you are better than this. Send me 4 hands, give me the contract[s],the auction, and I'll provide you with the play by play by GIB. Then you can make your sweeping judgmental comments.

And you don't know the deficiencies of DD, otherwise you wouldn't have posted such nonsense.

Kurt

--
Posted by Mimo Usenet Browser v0.2.5
http://www.mimousenet.com/mimo/post

jogs

2016-08-16 13:53:00 UTC

Post by KWSchneider
Rainer - you are better than this. Send me 4 hands, give me the contract[s],the auction, and I'll provide you with the play by play by GIB. Then you can make your sweeping judgmental comments.
And you don't know the deficiencies of DD, otherwise you wouldn't have posted such nonsense.
Kurt

Can I send you a hand starting from trick 2?
Declarer has mistakenly ducked the opening lead.
Can SD East find the switch to set the contract?

3NT South

S T75
H AT65
D 4
C AK763
----- S Q64
----- H J84
----- D AK75
----- C 854

T1: S3, 5, Q, 8
What does East lead on trick 2?

TIA

f***@googlemail.com

2016-08-16 15:22:37 UTC

Can I send you a hand starting from trick 2?
Declarer has mistakenly ducked the opening lead.
Can SD East find the switch to set the contract?
3NT South
S T75
H AT65
D 4
C AK763
----- S Q64
----- H J84
----- D AK75
----- C 854
T1: S3, 5, Q, 8
What does East lead on trick 2?
TIA

This is a truly absurd question.
You aren't prepared to give the auction so we know nothing about declarer's hand.
But you are prepared to say that 'South has mistakenly ducked at trick 1'.
At the table, you (i) don't actually know if South has ducked or if your side are cashing the entire spade suit, and (ii) if South has ducked, you don't know if it is a mistake or not.

If the auction was, say, 1NT-2C-2D-3NT (not my choice on the North cards) then the "SD" correct continuation after a spade to the queen holds, is to play partner to have started with AK932 and cash five rounds of spades.

jogs

2016-08-16 16:13:27 UTC

Can I send you a hand starting from trick 2?
Declarer has mistakenly ducked the opening lead.
Can SD East find the switch to set the contract?
3NT South
S T75
H AT65
D 4
C AK763
----- S Q64
----- H J84
----- D AK75
----- C 854
T1: S3, 5, Q, 8
What does East lead on trick 2?
TIA

This is a truly absurd question.
You aren't prepared to give the auction so we know nothing about declarer's hand.
But you are prepared to say that 'South has mistakenly ducked at trick 1'.
At the table, you (i) don't actually know if South has ducked or if your side are cashing the entire spade suit, and (ii) if South has ducked, you don't know if it is a mistake or not.
If the auction was, say, 1NT-2C-2D-3NT (not my choice on the North cards) then the "SD" correct continuation after a spade to the queen holds, is to play partner to have started with AK932 and cash five rounds of spades.

IMP pairs, dealer North

EW silent

The two most common auctions were

1C - 3NT all pass

1C - 2NT, 3NT all pass

South had SA, HAKQ, CAKQJT. No communication problems.

Are you willing to concede that South erred by ducking?

f***@googlemail.com

2016-08-16 17:31:02 UTC

Can I send you a hand starting from trick 2?
Declarer has mistakenly ducked the opening lead.
Can SD East find the switch to set the contract?
3NT South
S T75
H AT65
D 4
C AK763
----- S Q64
----- H J84
----- D AK75
----- C 854
T1: S3, 5, Q, 8
What does East lead on trick 2?
TIA

This is a truly absurd question.
You aren't prepared to give the auction so we know nothing about declarer's hand.
But you are prepared to say that 'South has mistakenly ducked at trick 1'.
At the table, you (i) don't actually know if South has ducked or if your side are cashing the entire spade suit, and (ii) if South has ducked, you don't know if it is a mistake or not.
If the auction was, say, 1NT-2C-2D-3NT (not my choice on the North cards) then the "SD" correct continuation after a spade to the queen holds, is to play partner to have started with AK932 and cash five rounds of spades.

IMP pairs, dealer North
EW silent
The two most common auctions were
1C - 3NT all pass
1C - 2NT, 3NT all pass
South had SA, HAKQ, CAKQJT. No communication problems.
Are you willing to concede that South erred by ducking?

Obviously that wasn't South's hand because it is a 1305 23-count. I assume you that that those were the combined honours between the NS hands.

The point is that when you have a single dummy defensive problem, you aren't given the additional information '....and declarer erred by ducking'. You have to decide, as the player whose SQ has held the trick, what the best continuation is. This may include an analysis of probability of beating the contract if declarer has played correctly; the probability if declarer has played incorrectly and the probability of that happening and the relative importance of overtricks at 1 imp per time.

As a defensive problem, it is usually not right to assume that South has 10 top tricks and no diamond stop because it's an insane play to duck the spade - even at matchpoints, there's no gain, and this isn't matchpoints. Perhaps your opponents play like that but mine don't (and if they do, I shall win).

Give South Ax Kxx Qxxx QJxx and a spade continuation beats the contract while a diamond switch lets it through. It is only because you give the unauthorised information that South has erred that it becomes evident we need to play on diamonds.

(I have seen players duck here as declarer from Kxx in spades - which is right if they have a card to knock out and the opening lead is from Jxxxx. But it's wrong when the opening leader has the Ace.)

f***@googlemail.com

2016-08-16 17:36:25 UTC

Can I send you a hand starting from trick 2?
Declarer has mistakenly ducked the opening lead.
Can SD East find the switch to set the contract?
3NT South
S T75
H AT65
D 4
C AK763
----- S Q64
----- H J84
----- D AK75
----- C 854
T1: S3, 5, Q, 8
What does East lead on trick 2?
TIA

This is a truly absurd question.
You aren't prepared to give the auction so we know nothing about declarer's hand.
But you are prepared to say that 'South has mistakenly ducked at trick 1'.
At the table, you (i) don't actually know if South has ducked or if your side are cashing the entire spade suit, and (ii) if South has ducked, you don't know if it is a mistake or not.
If the auction was, say, 1NT-2C-2D-3NT (not my choice on the North cards) then the "SD" correct continuation after a spade to the queen holds, is to play partner to have started with AK932 and cash five rounds of spades.

IMP pairs, dealer North
EW silent
The two most common auctions were
1C - 3NT all pass
1C - 2NT, 3NT all pass
South had SA, HAKQ, CAKQJT. No communication problems.
Are you willing to concede that South erred by ducking?

Obviously that wasn't South's hand because it is a 1305 23-count. I assume you that that those were the combined honours between the NS hands.
The point is that when you have a single dummy defensive problem, you aren't given the additional information '....and declarer erred by ducking'. You have to decide, as the player whose SQ has held the trick, what the best continuation is. This may include an analysis of probability of beating the contract if declarer has played correctly; the probability if declarer has played incorrectly and the probability of that happening and the relative importance of overtricks at 1 imp per time.
As a defensive problem, it is usually not right to assume that South has 10 top tricks and no diamond stop because it's an insane play to duck the spade - even at matchpoints, there's no gain, and this isn't matchpoints. Perhaps your opponents play like that but mine don't (and if they do, I shall win).
Give South Ax Kxx Qxxx QJxx and a spade continuation beats the contract while a diamond switch lets it through. It is only because you give the unauthorised information that South has erred that it becomes evident we need to play on diamonds.
(I have seen players duck here as declarer from Kxx in spades - which is right if they have a card to knock out and the opening lead is from Jxxxx. But it's wrong when the opening leader has the Ace.)

I suspect that the right single dummy card at trick 2 is a top diamond honour This lets through the contract if declarer has Ax KQx Jxxxx QJx or (in practice) Axx KQx Q10x QJxx although the latter gives partner a very odd choice of lead.

But I am really struggling to construct a hand where it is right to switch to diamonds at trick 2 unless declarer has misplayed; and it's easy to construct hands where continuation spades is correct.

KWSchneider

2016-08-16 17:40:47 UTC

Can I send you a hand starting from trick 2?
Declarer has mistakenly ducked the opening lead.
Can SD East find the switch to set the contract?
3NT South
S T75
H AT65
D 4
C AK763
----- S Q64
----- H J84
----- D AK75
----- C 854
T1: S3, 5, Q, 8
What does East lead on trick 2?
TIA

This is a truly absurd question.
You aren't prepared to give the auction so we know nothing about declarer's hand.
But you are prepared to say that 'South has mistakenly ducked at trick 1'.
At the table, you (i) don't actually know if South has ducked or if your side are cashing the entire spade suit, and (ii) if South has ducked, you don't know if it is a mistake or not.
If the auction was, say, 1NT-2C-2D-3NT (not my choice on the North cards) then the "SD" correct continuation after a spade to the queen holds, is to play partner to have started with AK932 and cash five rounds of spades.

IMP pairs, dealer North
EW silent
The two most common auctions were
1C - 3NT all pass
1C - 2NT, 3NT all pass
South had SA, HAKQ, CAKQJT. No communication problems.
Are you willing to concede that South erred by ducking?

Obviously, GIB or a human has to assume that all of the players are trying to play their cards to the best of their ability - and not making errors. Consequently, GIB would assume that the "duck" of the spade was appropriate.

Nonetheless, on the basis of a 1C (P) 2N (P) 3N (AP) auction, GIB looked at all of the potential returns (DA, Sx, D5, D7, C5, C8, H4, H8, HJ) and ranked them. The DA ranked significantly above the spade return (which came 2nd). The only "positive" return rankings were the DA, Sx, D5, D7. All others were immediately disregarded.

So the answer is that GIB lead the DA at trick 2, and would have defeated the contract. I can provide you with a screenshot if you would like documentation.

Kurt

--
Posted by Mimo Usenet Browser v0.2.5
http://www.mimousenet.com/mimo/post

KWSchneider

2016-08-16 02:40:54 UTC