You are ignoring the author of this comment. Click to temporarily show the comment.
Sounds like a fair comment:

That's the real novice mistake: The circular assumption that goes as follow - by ducking he shows that he doesn't have a stopper in diamonds or spades. Hence he should not have the J either.
I can even force him to hesitate when he discards, I can force him discard a diamond and be able to say “See! See! He didn't have the J”

I know the correct way to play the diamond suit in isolation is K then A then low to dummy. But I just want to finesse first or second round (if an entry to dummy exists).

I guess now I really know why! I have seen the hands and I am really trying to convince myself it didn't matter.

Thanks
Nov. 23, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
“Declarer could have Kx, J10xx, Axx, AKxx - or other hands”

I am a bit of a novice here, so please forgive my miscalculations as I don't see to deeply at this stage of my bridge level. I really appreciate the critique of my analysis, though. I am trying to see as much as I can (improve my speed and accuracy). So my thoughts are as follows:

I did think about it, but I kind of dismissed this hand, perhaps superficially. The play of hearts suggests heart length with East having 5 or unlikely 6 without many more points. The heart duck also suggests that.
Entries are unlikely.

So 8 top tricks and a two way finesse.
Perhaps you could force East to avoid discarding a diamond,
resulting in endplaying East. (Some sort of informational strip squeeze).

The fact is that the J is much more likely to be with West.

While there some ways Declarer could go wrong
a) hoping for a defender having QJ doubleton
b) Playing West for A9 doubleton
c) combining some of these types of chances.

By in large (superficially) I think East should conclude that
with that type of hand holding, declarer will guess that West has the J so West really does need to hold A.
Nov. 23, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
This analysis assumes a hcp 15-17 NT.

After partner plays the Ace, you know the JT984 are out. When partner returns the 9 you know partner is very likely to have the 4, as declarer played either the J or the T. So now you should think that the suit being blocked might be a possibility for the imaginative play of the 9, as partner could have the 8 and the 4 left.

Since opener has at least 15 hcp, what can partner possibly have that will prevent 3N from making, assuming opener started with JT82. Partner has to have the Qxx(x) and A, in order for it to be correct to duck the heart return. This violates the rule advocating the economy of expectation in defense. A more reasonable expectation is that partner has Axxx and is unblocking.

Note: If partner has those two key stoppers then declarer has to have Kx JT8x Jxx AKQJ for starters which seems slightly unlikely as well - to open 1NT and confidently bid 3N on a misfit.
Nov. 22, 2019
Imtiaz Husain edited this comment Nov. 22, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
You are ignoring the author of this comment. Click to temporarily show the comment.
Thanks Dale. I have been mainly looking at some old Mike Lawrence hands I guess.
Nov. 15, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
The default is non-forcing unless you have an agreement otherwise.

See http://bridgewinners.com/article/view/third-suit-forcing/
for Third Suit Forcing options
Nov. 15, 2019
Imtiaz Husain edited this comment Nov. 16, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
But more importantly, Society is always to blame, the real question is why is this choice only second place. Darn I forgot to vote for it!
Nov. 15, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
direct splinters support opener's suit; relay splinters support own suit (autosplinters)

The splinter using the 1x 1M 1N 2N route is slammish. While the splinter using the 1x 1M 1N 3oM route is warning against 3N. 3N may still be the best contract, but suggests responder prefers to play in openers minor instead.

Alex, while I prefer having a 5th diamond for the 3S bid, the 2D bid followed by 3D suggests your suit and your support are not that great. This will cause opener to bid 3N as you may have values in spades. While the intention of the 3S bid clearly warns against bidding 3N. Defenders will lead Spades, guaranteed when opener parks in 3N.

I think your bid is “theoretically” correct. Of course partnership discussions are required.
Nov. 14, 2019
Imtiaz Husain edited this comment Nov. 14, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
I play it splinter in spades setting Diamonds as trump; so 1453 or better
Not interested in slam and warning off playing 3N
Nov. 14, 2019
Imtiaz Husain edited this comment Nov. 14, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
Hi David are you still looking for a partner for testing out relays systems?
Nov. 14, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
Nov. 8, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
In that case I would look up and down too before crossing!
Sept. 13, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
Thanks for all the suggestions. I guess I need to try out some of these programs and get a better appreciation of their capabilities.

Michael - (wikipedia define 1-hot as only 1 1 and the rest (in your case 53 - so I guess you must mean 13-hot) this is kinda similar to chess and since 53 < 64 the same mechanism can apply.

I like the idea of AI vs AI to develop better bidding systems.
But until I can see some results (yields) from such an approach,
I am likely to be skeptical.

As far as explain-ability is concerned my understanding is that
this is not really an issue. Its automated explain-ability (i.e)
can the computer explain itself that is the ongoing area of research. And yes in machine learning this can be tricky.

Jeannie - I can understand Monte Carlo tree search being useful
in programs like AlphaGo.

How big is the tree space in bridge? I can't imagine it as big
I guess it is complicated by the fact that you are trying to figure out what cards players hold. (That's where Monte Carlo comes in?)
June 16, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
Oh man - I would love to get a copy of the data to do my own analysis. Chess databases have features like find innovations etc.

It would be amazing to do similar types of analysis on BBO data.
Who can I talk to see if I can get a copy of that data?
June 14, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
I don't consider double dummy solvers real AI. They just use a simple mathematical algorithm. If they do use alpha-beta it is really play all 13 cards evaluate the result - compare.
That is not really alpha-beta min max, as there is no tree-pruning - which is a requirement for alpha-beta min-max.

OK thought about it some more - there must be tree pruning for complete fails - not sure if this still constitutes alpha-beta

Thanks for responding though. I am, however, more interested in how AI handles/solves the bridge problem. I appreciate your spending the effort to answer this question.

I am assuming the correct solution should not involve alpha-beta min max. But if I were to write a program - I cannot imagine a different type of solution.
June 14, 2019
Imtiaz Husain edited this comment June 14, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
Yes if you hard code bid game at 26 combined hcp then there are a number of hands your evaluation functions will do poorly on. This is the same problem in games like chess. However, just like in chess better evaluation functions produce better results. (faster + more accurate)

I would think that in bridge you can have more complicated evaluation function, as you are not dealing with as large a tree as in chess or go.

You should in theory be able to create better evaluation functions
than experts - so I disagree with your example.
June 14, 2019
Imtiaz Husain edited this comment June 14, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
What algorithms apply to bridge playing programs? I assume some sort of alpha-beta min max is a non-starter?
June 13, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
June 13, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
Yes - I enjoyed this hand very much - Thank you Kevin for sharing!
June 12, 2019
You are ignoring the author of this comment. Click to temporarily show the comment.
Dave,

Just for my records, is this
i) a hypothetical
ii) an statement of intention
or
iii) a statement of what you are doing at the moment.

BTW are you playing online while doing so?