Okay. This thread is starting to go in circles. So let me try to refocus and wrap this up.
pescetom, on 2019-September-03, 09:50, said:
If you want a precise answer I suggest you post the same question to Barry Rigal and other historians on BridgeWinners. But Goren and others were playing "Standard American" in the late 1940s. Whenever that became 5-card majors based, it's hardly surprising that they retained a "normal" non-forcing 1NT response and non game forcing 2/1, just as Jais and Lebel did later in Europe. Kaplan Sheinwold was just too far ahead of its time to gain widespread acceptance.
Okay. I didn't know that was the better place to ask that question. Thank you for the suggestion.
Should I just post a question to their intermediate forum? Or is there a better way to do it? I don't normally read their forums like I do here.
hrothgar, on 2019-September-03, 03:05, said:
If you want people to critique your use of information theory then you need to present your calculations, not just make reference to the fact that you have done so...
I did a non-mathematical version above. I can do the math precisely if you guys really need it. But it seems to me that the problem is more that people keep lumping a bunch of different things together. So let's try to clear that up and then I'll do a sketch.
Goren (and Culbertson and ACOL) are all 4-card major systems. They have different rules for continuing the bidding (limiting vs forcing bids), but it is very hard to know the precise distribution and they do in fact have the problems that people keep attributing to canape.
There are "canape tendency" systems where normal bidding is (effectively) 5-card majors and reverses are canape. I don't understand why you'd want to do this. But multiple top-level systems have. And the Blue Team's version is super-complicated because it has a bunch of exceptions and special sequences. Internet folklore says that the point of "tendency" is to let you show strength better at the cost of distribution. I don't know if that's true or where it came from; tendency just shows up in the Italian systems. But whatever the reason, Blue Team is *not* Alberan's method.
Alberan's system is a pure canape system. The bidding solves exactly the same problem that 5-card majors does and for exactly the same reason. The bids you say are different but you are communicating the same information give or take a symmetry transformation.
So if you want to say that there's some flaw with Alberan's approach that 5-card majors doesn't have, then you'd have to show why he and everyone else missed it. Simply referencing what is popular doesn't work -- people once bid 4-card majors instead of the older strong club. And they swapped from limit raises (Culbertson) or forcing ones (Goren). I could go on. But the history of the game is rife with examples of this stuff.
So it's a facially bad argument to say, "it isn't popular and therefore doesn't work despite all of the theorists considering it to be just a different way to show 5-card majors."
OTOH, saying, "All of the theorists say that Alberan's canape is just a different way to show 5-card majors. Since the bids are more or less equivalent, there's no point in playing his version instead of the 5-card majors that we use. So no one bothers since it has no advantages and would leave you without the benefit of everyone else's bidding developments."
This is what people said above. And I think that's a completely sound reason for why it would have fallen out of use.
===
As for the information theory stuff, rhm is the one who cited two books that made very primitive (i.e. flawed) use of the idea.
His claim that canape gives less information is false. In most sequences it gives more. This isn't necessarily good or bad. (Bidding a careful slam that only goes down on a particular, very unusual lead is not going to work if you spell out which card that needs to be.)
Those books more or less say that each higher bid in the ladder should eliminate 50% of the remaining possible hands. If you had no competition to worry about, this *might* be kinda true. But you don't want to lump together raw hands; "Open 1 club with an odd number of black cards," and "Open 7NT on exactly 4321S 321H 321D 321C," both satisfy that criterion.
Similarly, it's not just about grouping hands by your score or even the differential; you care about the optimal contract: a 4S sacrifice and a 4S game don't actually need to be distinguished in theory. If your bidding system always gets you to 4S when it is the best outcome, the rest doesn't matter.
If it was just you and your partner, then the ideal bidding system would always maximize the variance reduction per bit of information the bidding conveyed; modulo making sure that you don't overshoot the correct bid. The books he's suggesting acknowledge this more or less, but they don't really account for it in a proper way. (You'd want to account for it with the same math that lets you use different cost/penalty functions for Bayesian estimators to trade-off type I and type II error; except you be trading off between expected points vs penalties.)
And since this is a competitive game, you actually have a more complex task: you want to minimize the variance that you and your partner have when estimating the correct contract. But you want to maximize the variance of the opponents. It's generally impossible to do both with crypto methods disallowed, so the trade-off depends on seating position, vulnerability, whose hand this really is, and a host of other factors.
So this is why "on paper" a 15-17 1NT opening is "right", but people get positive IMPs with a weak one. The same is true of a lot of other bids as well.
=====
Now that brings us to Alberan's canape system vs. 5-card majors.
Essentially, the rule is that if you bid the major second it has exactly five cards. (Rare things like 6-6 hands excepted.)
If you have a 4-4 fit, you discover it immediately and find a 5-3 fit on a rebid. So all you did was just swap these two sequences.
You end up with more information in canape on both auctions. With the 1M opening, you know that if opener has a minimum, it includes a 5-card minor and is unbalanced. If you have a super-fit because opener had 6+ in the major, you'll discover it on opener's rebid and he'll be able to easily deal with competition.
If he doesn't open a major, then you know that he doesn't have one except in one special circumstance. And in that circumstance, you'll end up in the same place in the bidding after opener's rebid on an uncontested auction and he'll be in a better position to deal with interference knowing that you have *at best* a 5-3 fit (as will you when he bids the major since you'll know if you have a fit at all and exactly what it is.)
With the other sequences, things are mostly a wash: With canape, you can find one 5-4 fit on responder's first bid and the other on opener's rebid. Both of these are slightly sooner, however you find out about 5-5 and other extreme fits a bit later.
If you are a rigorous adherent to the LoTT, then canape is slightly better. Otherwise, then I'm not seeing how the differences matter. You end up at the same bid. You give away slightly more information, for good and bad. And you can compete about as well. It changes who the captain is on some bids, and needs somewhat different conventions for a few situations. But on the all, it's just a matter of which fit you find first: 4-4s or 5-3s.
5-3 is less desirable. And given your cards, it's more likely to find a 4-4 fit than a 5-3 one. Canape finds the 4-4 faster and leaves the minor nebulous, so it makes the opening lead more challenging. It makes over-calling 1M slightly riskier because it means that the canape bidder will immediately know that you don't have a fit or that you have a really bad trump break. On the flip side, while canape lets you compete better in lots of auctions, in the close part-scores with dueling 5-3 fits, you are going to probably come out behind.
So this stuff cuts both ways. We can run the exact numbers if you really need me to, but I don't really see any basis for believing that this slight change of bidding emphasis is going to make any noticeable difference. (And per the discussion above, that's probably why no one is using it.)
P.S. Having written all of this out, it occurs the me that there is actually a way in which canape would be better or worse that we haven't considered.
If canape communicates the same information with fewer principles and conventions, then it would be superior in some sense. I.e. we should be looking at the information necessary to use the system itself. Alberan's canape was made to be an ACOL variant with the distributional precision of 5-card majors. He justified having to change major principles of the system on the grounds that doing that would result in fewer special sequences and artificial conventions.
AFAIK, no one has seriously tried to evaluate this claim. But now I want to figure it out.
If he's right, then there's an argument that ACOL players should be using his distribution showing rules instead of their current ones. If he's wrong, then that's the answer -- his system uses more natural bids without covering the corner cases that standard American covers with special understandings in the unusual bidding sequences.
Help
