[msheald]

Hello! Just curious. Why would the robot lead a singleton against a 3 NT game?

I don't think it mattes much is this game. However, I was taught to lead a singleton when I only have a point or two and will never get back to my hand to cash a length trick. The theory is that partner may have something in my short suit.

With 5 points and a 6-card heart suit, it does not appear to fit that category.

It does not seem to matter in this hand. But in other hands, it would just fool partner into returning a dead suit. Best regards.

Mike


https://tinyurl.com/2ovfquxq

[pilowsky]

I had the same hand.
They do it all the time.
Apparently, simulations suggest it's a good lead.
With a human partner it might be a little confusing,

[smerriman]

Humans make leads that help their partner know what to do if they get in. Robots make leads under the assumption that everyone will play double dummy after that, so "fooling partner" is not a consideration.

It therefore usually prefers leads which it knows are less likely to immediately give up a trick; so much so that it's an important part of robot play to assume the opening lead was unlikely to be from an unsafe honor holding (and if you find out it was, they probably have unsafe holdings in all other suits too, so you can place some cards in their hand).

[johnu]

If you are interested in learning more, read "Winning Notrump Leads" by David Bird and Taf Anthias. There are a number of examples where leading a singleton against NT is the best lead.

Bird and Anthias also have a companion book "Winning Suit Contract Leads".

Both books use double dummy simulations to determine their lead rankings.

[thepossum]

All you know is to not assume its 4th in longest suit - could be anything - maybe a singleton suggests partner has a lot of that suit. Who knows
Or is it one of the non descript passive leads. Nothing else seemed any good. This one will have to do
I do find it strange though - likely establishing long cards in opps, no chance of leading back to partner if they get in again. Things like that

EDIT Apparently par for that hand is 7Hx-6, and sure enough it did lead to your long suit. Did you establish your 9?
EDIT 2 Curiosity got the better of me and I hadn't run a little sim for a while. Any card seems much as good as any other +/- a little - sorry I lie. Misreading the numbers. How about a diamond - innocuous passive lead - sorry I mean a small heart

[mycroft]

One thing I remember from back in the day when the Bird/Anthias book came out is the caveats from others:

• It can work very well, *if partner has read the book also and understands how you think*.
  
• the simulations strongly value Ace leads that "can work out" the killing switch after seeing dummy (the book notes this as an issue with their simulation methods). I'm not as smart as double-dummy computers; you probably aren't either.
  
• Following the ideas gleaned from the NT book are much more effective than those from the suit leading book, even if 1. above hasn't happened

NT openers are just so much more prescribed than suit openers, especially when they then *play* in NT.

But I will note that "when I don't have entries, and the opponents have a limited HCP range, maybe try setting up partner's suit instead?" is in fact an effective strategy. And when they have a NT auction that *didn't look* for a major fit, it's even more likely that your short major is "partner's suit instead". Of course, I also play (Reverse) Smith echo, so even when partner can't count my hand to the "no entries" I can see, I can usually get the point across.

[thepossum]

One thing I am curious about more broadly, ignoring the minute and insignificant differences in this hand is that the best lead (in my sim) is a small heart to defeat the contract but the singleton is better in terms of tricks. But it is so minute as to not matter on this hand

[pilowsky]

 thepossum, on 2024-June-25, 18:05, said:

One thing I am curious about more broadly, ignoring the minute and insignificant differences in this hand is that the best lead (in my sim) is a small heart to defeat the contract but the singleton is better in terms of tricks. But it is so minute as to not matter on this hand

That's something Bird&Anthias discuss. Each simulation they provide gives the 'best' lead in terms of IMP's vs MP's.

[thepossum]

Thx. Mine aren't that sophisticated but I compare tricks versus defeating the contract

[1175]

I just played a hand that certainly demonstrates the "unsafe honor holding" principle:

---

At nine of the fifteen tables in the duplicate IMPs, South opened and played 2NT. At every table, West lest the ♣4, the only lead (double dummy) that holds declarer to eight tricks. Three of the tables (including mine) made nine tricks anyway. The unsafe honor holding principle apparently only applies on opening lead. I floated a spade early, and West won the jack, and shifted to the ♥K. Later, when West won the ♦K, it continued with the ♥J, giving me three heart tricks.

[thepossum]

I find it curious that the singleton actually works DD on that hand specifically but sims still show a small diamond (the traditional lead) as best for tricks and defeating the contract - on average by a fair bit
Its just a shame the bots don't communicate well with humans in defence, let alone each other - do I have that phrase back to front. It sounded right at the time

[1175]

 thepossum, on 2024-June-29, 01:27, said:

Its just a shame the bots don't communicate well with humans in defence...

I certainly agree with that. I often feel like I don't have all of the information I should have when defending with the Robot as a partner.

[msheald]

Another example of a traditional lead defeating the contract while the lead of the singleton does not.


https://tinyurl.com/2zwlfj3c


Mike

[1175]

 msheald, on 2024-July-01, 05:07, said:

Another example of a traditional lead defeating the contract while the lead of the singleton does not.


https://tinyurl.com/2zwlfj3c


Mike

The hand evaluation shows (double dummy) that declarer can make exactly seven tricks on any lead (assuming optimal play by both sides). If North leads a heart, West simply wins the third round, and North can never get it to cash the long hearts.

[msheald]

Thank you, 1175! When I first went over this hand, I miscounted and gave south an extra heart as entry to north. (I keep telling myself not to count when I'm distracted early in the morning with the needs of the day!) Again, thanks for the correction.

I had thought that my misunderstanding of Gib was failing to read the bidding manual closely.

Now, I'm just realizing what smerriman meant when Gib assumes all robot partners play double dummy. I had thought that the program allowed Gib to calculate card distribution frequencies for the hands and calculate the best pay based on that, updating itself after each round of cards are played. I thought that it did what David Bird does in his book, "Winning Duplicate Tactics" - find the best play from the most likely distribution of cards based on bidding and past cards played.

However, if all robots play double dummy, that means, basically, that Gib always knows the cards. By extension, this also means that the program has to assume that the human player does as well in that case. That it is why it does not need to signal! And plays the way it does that often leaves my scratching my head in puzzlement.

Having all robots play double dummy saves computing power and was important before when computing power was more expensive. I'll be curious how modern programs play bridge as computing power significantly drops in cost. Best regards.

Mike

[smerriman]

You misread my message. It does do what Bird did in his book, though with smaller samples. I did not say it plays double dummy - I said it *assumes* everyone will play double dummy *after* whatever card it plays, just like in Bird's book (that's how they calculate what the average score will be for each card). It doesn't *actually* see the hands.