[JavaBean]

Scoring: IMP

---

Unopposed, my partner and I bid these hands up to 6♠ South (seems reasonable, though 6NT may be better). I received a heart lead, and clearly the main issue is how to tackle trumps for one loser. There is some remote possibility of a minor-suit loser, but there are several ways to take care of that and I'm not terribly interested in them.

At the table I ran the ♠10, imagining that the only 4-1 split I can pick up is West's ♠9. This was a riotous success when he in fact did have a singleton nine, but when I thought about it some more it struck me that running the ten is a terrible play, gaining against one 4-1, but losing against two (East's stiff honor). I concluded that the best line is to just cash the ace, succeeding on any 3-2, and also when East holds a singleton honor.

But then I gave the hand to Suitplay, which has an improvement: run the eight, which retains the ability to pick up East's stiff honor, and as an added chance gets West's singleton seven. So now I am confident that is the best line for four tricks, and I clearly understand why.

But when I give the whole hand to Jack (the bridge program I play against when no humans are available), his play surprises me. He starts by leading the 8 but then, when East follows with the 2, he rises with the ace and cashes the king. Since this is demonstrably a poor line in spades, and it should be very easy for a computer to derive the best line in isolation (the same way Suitplay does), I wonder if there is some facet of the whole deal that means spades should be tackled differently. Does anyone see it?

[fred]

Nice suit combination!

I write up a suit combination problem each month for The Bridge World magazine. Expect to see this combination appear in about 6 months.

If you want to see your (real) name in print, please post it here or send it to me by e-mail (fred@bridgebase.com) so I can include a mention like "combination suggested by...". Also let me know what city and country you live in.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[foo]

JavaBean, on Dec 24 2008, 07:36 PM, said:

... I gave the hand to Suitplay, which has an improvement: run the eight, which retains the ability to pick up East's stiff honor, and as an added chance gets West's singleton seven. So now I am confident that is the best line for four tricks, and I clearly understand why.

But when I give the whole hand to Jack (the bridge program I play against when no humans are available), his play surprises me. He starts by leading the 8 but then, when East follows with the 2, he rises with the ace and cashes the king. Since this is demonstrably a poor line in spades, and it should be very easy for a computer to derive the best line in isolation (the same way Suitplay does), I wonder if there is some facet of the whole deal that means spades should be tackled differently. Does anyone see it?

♠T8643♥A♦K982♣AQT
+
♠AK5♥KQ♦AJ5♣K9543

6S by N, ♥x led

My 1st thought is that Jack wants to get as many trumps off the table as possible before letting Them in to protect against a possible bad minor suit break and potential ruff.

When Jack plays the ♠8 and 2nd hand plays the ♠2, the possible layouts where We can take 4+ tricks on the single suit line are:
7+QJ92
97+QJ2
J7+Q92
J9+Q72
Q7+J92
Q9+J72
QJ+972
J97+Q2
Q97+J2
QJ7+92
QJ9+72
On all the 3-2 layouts, We are cold for only 1 ♠ loser. Unless They can get a ruff.

IMHO Jack decided at that point in the play that the odds of a 4-1 ♠ break that can be handled by following the original line are less than the apriori odds of a bad minor suit break that could set the contract if They got in too soon and got a ruff. Therefore Jack changed from the single suit optimum line to the line that its calculations said optimized the chances to make the contract.

EDIT: different copies of Jack seem to be doing different things on this board. Some play the single suit optimum line. Maybe this thread is more about a SW setting / bug than about bridge?

[JavaBean]

fred, on Dec 24 2008, 09:25 PM, said:

Nice suit combination!

I write up a suit combination problem each month for The Bridge World magazine. Expect to see this combination appear in about 6 months.

If you want to see your (real) name in print, please post it here or send it to me by e-mail (fred@bridgebase.com) so I can include a mention like "combination suggested by...". Also let me know what city and country you live in.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

Thanks! I'm glad you find it as interesting as I did, and in fact I read your message just after reading the January BW suit combination, so I would be delighted to see my name attached to one of them. My name is Alan Malloy, living in Santa Clara, California.

[MickyB]

Yes, very nice. Your original line of running the ten also picks up stiff seven on your left, so it isn't actually worse than cashing AK (until you take into account the possibility of a ruff).

Jack may well be trying to avoid a ruff - are you able to give it a hand where that isn't a concern - a 7NT contract, probably - and see how it plays the spades now?

[kgr]

- Playing AK is better when one opp has QJ doubleton and you have a looser in a side suit. (when one of the opps has a 5 card ♣).
Maybe that is enough reason for Jack to play high?

[kgr]

JavaBean, on Dec 25 2008, 02:36 AM, said:

But when I give the whole hand to Jack (the bridge program I play against when no humans are available), his play surprises me. He starts by leading the 8 but then, when East follows with the 2, he rises with the ace and cashes the king. Since this is demonstrably a poor line in spades, and it should be very easy for a computer to derive the best line in isolation (the same way Suitplay does), I wonder if there is some facet of the whole deal that means spades should be tackled differently. Does anyone see it?

I tried this in Jack. Heart lead and you return ♠8 in trick 2 and East plays small (♠2). Then ctrl-A for analysis:
Jack prefers now small ♠ over A or K.
 
But if you do analysis from the start then Jack seems to prefer to cash ♦K and ♣A before playing the ♠8 and taking it with the ♠A when East follows low.
...I wonder how this can be explained.

[foo]

kgr, on Dec 25 2008, 08:02 AM, said:

JavaBean, on Dec 25 2008, 02:36 AM, said:

But when I give the whole hand to Jack (the bridge program I play against when no humans are available), his play surprises me. He starts by leading the 8 but then, when East follows with the 2, he rises with the ace and cashes the king. Since this is demonstrably a poor line in spades, and it should be very easy for a computer to derive the best line in isolation (the same way Suitplay does), I wonder if there is some facet of the whole deal that means spades should be tackled differently. Does anyone see it?

I tried this in Jack. Heart lead and you return ♠8 in trick 2 and East plays small (♠2). Then ctrl-A for analysis:
Jack prefers now small ♠ over A or K.
 
But if you do analysis from the start then Jack seems to prefer to cash ♦K and ♣A before playing the ♠8 and taking it with the ♠A when East follows low.
...I wonder how this can be explained.

This not the behavior I got.

"Analyze Position" at the start if T2 rates CQ or CA as best play.
Let Jack choose what to do and it plays the S8
*huh?*

After E inserts the S2,
Analyze Position rates the S5 as best
...and if I let Jack choose what to do... ...it plays the S5

Which makes me wonder what this thread is really about.

I'm running Jack 4.01 with playing strength settings of
time 15, auction 2, cardplay 3, lead 3, declarer play 4, memory 10, and Finish calculations checked.

[hotShot]

I don't know if is published how Jack goes single dummy, but if it is done similar to GIB that uses a set of random deals using the known restrictions to search for the best play.
This would mean, that 2 user at different computers might get a different behavior from from the software, because each case has its own set of random deals.