[inquiry]

Well, i never meant to hold this example up as proof flannery is good. There are several problems with the shown data that none have mentioned, so i will

The way the search was performed was I looked at all first seat opening that meet the necessary criteria. First there was 118,145 times 1♥ was opened and only 825 times when 2♦ was opened. The obvious implication of this is on the great majority of DEALS 2♦ was never opened. By this, I mean if each deal was played 16 times, and 2♦ only opened once on any deal, There would have been 7435 deals, and only 11% of them include a 2♦ opening.

Second, anytime you look at a large number of standard type opening, like 1♥ it averages out to zero imps or 50% matchpoints. By applying pressure to the bids (quality of suit, number of hcp) you can see whether a weak 2♥ or a atronger 2♥ does better or worse, or if opening on a five card suit, etc. But as an example, if you look at all weak 2♥ opening bids, they average out to zero.

So the statement that the players that opened 2♦ (flannery) were better than the ones that opened 1♥ somewhat missed the point. 1♥ averaged to 0 because that is what normal bids do. But the 2♦ bidders did better, but perhaps (and i am ready to believe it) 825 hands (not deals, hands) is too small. On the ohter hand, there is an argument that the flannery guys had good agreements. Of the 825 deals, 56 of them played in things like a 3-1 diamond fit and did very poorly, so not all of them were on the same wavelength.

To do studies like this adequately, what needs to be done is to grab all the DEALS (not hands) where someone opened 2♦ with flannery hand. So only the 1♥ openings are compared to 2♦ opening on the same HANDS. I just found it interesting given the bad name some people try to force on flannery that it does just fine.

It is possible to group players making the 2♦ and 1♥ bid by ranges of lehman's scores.. any range i want, using bridgebrowser. To see how they do on the hand before and next hand, however would be technically challenging

[kenberg]

This is all to the good. If it were easy to settle which conventions are best then we would all play the same ones and have nothing to argue about.

[pclayton]

Couldn't you do this:

1. Flag all of the pairs that play Flannery.
2. Set up a criteria for a weak 2♦ - 5-10 points, 6 bagger and 2 of the top 5 honors or better.
3. On hands where the Flannery pairs ("1") were dealt the criteria in ("2"), calculate the IMP odds of what their counterparts who could have opened 2♦ (and did) received.

Compare the IMP average (as well as the relative frequency) of ("3") against the IMP average of the pairs that opened a Flannery 2♦.

Wouldn't this be more meaningful?

[helene_t]

Yes, but your suggestion still misses any (dis)advantage the Flannery players may have when they open 1♥. So the 1♥ openings need to be included in the analysis. The problem is of course that you need to identify the Flannery pairs and then find some 1♥ openings they made as well as some failures to open with a weak hand with 6 diamonds. Given that the database contained only 825 Flannery openings it's a fair bet that the results will be inconclusive.

Oh btw some Flannery players put the weak 2♦ in the 2♣ opening, intending to pass the 2♦ waiting bid. But I suppose you would have to ignore that since if you also need to include the 2♣ opening in the analysis (and some Flannerist may play Precision or such, arghhhh) it becomes really hopeless.

[awm]

Here's a bridgebrowser question I think might be interesting:

Take hands where the person in first chair is 5-5 (exactly) in any two suits and has 8-10 hcp. Compare the results from passing, opening one of a suit, and opening at the two-level.

The hope is that this gives some input as to the right action with these hands playing relatively "standard" agreements. Obviously this will be biased somewhat by folks playing a light-opening system (i.e. opening light with special agreements/ranges to handle it) but I suspect that these people are pretty rare on BBO and that most people opening 1M on 8-10 hcp are "upgrading for shape" in a system like standard american, acol, bbo advanced, 2/1, etc rather than playing moscito or magic diamond.

[han]

inquiry, on Feb 18 2008, 11:15 AM, said:

Well, i never meant to hold this example up as proof flannery is good. There are several problems with the shown data that none have mentioned, so i will

True, you never explicitly stated that this data had any relevance, but you started with "The question is, what bid worked best" and you ended with "the flannery bid did "ok"". That is very suggestive. Maybe I am a purist but I think that if one posts research that is as flawed as this then it would be good to start with a disclaimer stating that no conclusions can be drawn. Not everybody has the statistical background that you have.

I don't understand the problems with the data that you mention at all:

Quote

The way the search was performed was I looked at all first seat opening that meet the necessary criteria. First there was 118,145 times 1♥ was opened and only 825 times when 2♦ was opened. The obvious implication of this is on the great majority of DEALS 2♦ was never opened. By this, I mean if each deal was played 16 times, and 2♦ only opened once on any deal, There would have been 7435 deals, and only 11% of them include a 2♦ opening.

OK, so why is this a problem when looking at the results for a 2♦ opening? 825 is a reasonable number, the fact that this is a small part of the total number of hands is not a problem. The deals where nobody opened 2D are of course completely irrelevant when you compute the average number of IMPs gained when opening 2D.

Quote

Second, anytime you look at a large number of standard type opening, like 1♥ it averages out to zero imps or 50% matchpoints. By applying pressure to the bids (quality of suit, number of hcp)  you can see whether a weak 2♥ or a atronger 2♥ does better or worse, or if opening on a five card suit, etc. But as an example, if you look at all weak 2♥ opening bids, they average out to zero.

Yes of course. Again, I do not see why this is a problem here.

Quote

So the statement that the players that opened 2♦ (flannery) were better than the ones that opened 1♥ somewhat missed the point. 1♥ averaged to 0 because that is what normal bids do. But the 2♦ bidders did better, but perhaps (and i am ready to believe it) 825 hands (not deals, hands) is too small. On the other hand, there is an argument that the flannery guys had good agreements. Of the 825 deals, 56 of them played in things like a 3-1 diamond fit and did very poorly, so not all of them were on the same wavelength.

The argument was that the Flannery players and partnerships were better than the average. This didn't miss the point at all!

Quote

To do studies like this adequately, what needs to be done is to grab all the DEALS (not hands) where someone opened 2♦ with flannery hand. So only the 1♥ openings are compared to 2♦ opening on the same HANDS. I just found it interesting given the bad name some people try to force on flannery that it does just fine.

I don't understand why this is an issue. If the field was completely homogenuous you would do fine by looking at all hands. And again, you are not showing that flannery does fine, you are showing that the people who play Flannery do fine when Flannery comes up compared to the others playing these hands.

Quote

It is possible to group players making the 2♦ and 1♥ bid by ranges of lehman's scores.. any range i want, using bridgebrowser. To see how they do on the hand before and next hand, however would be technically challenging.

To get reliable results you would have to group by partnership, not by player.

[helene_t]

inquiry, on Feb 18 2008, 05:15 PM, said:

It is possible to group players making the 2♦ and 1♥ bid by ranges of lehman's scores.. any range i want, using bridgebrowser.

It is not good to reduce the lehman scoring to range groups. This throws information away.

What you have to do is to include the lehman score as a (continuous) covariate in your analysis. The baseline model (before we become concerned with Flannery) is something like

resultNS ~ lehmanNS - lehmanEW

Now for the subset of hands that are affected by NS playing Flannery we have

resultNS ~ lehmanNS - lehmanEW + Flanneryeffect

and for the subset of hands affected by NS not playing Flannery we have

resultNS ~ lehmanNS - lehmanEW - Flanneryeffect

[pclayton]

helene_t, on Feb 18 2008, 09:29 AM, said:

Yes, but your suggestion still misses any (dis)advantage the Flannery players may have when they open 1♥. So the 1♥ openings need to be included in the analysis. The problem is of course that you need to identify the Flannery pairs and then find some 1♥ openings they made as well as some failures to open with a weak hand with 6 diamonds. Given that the database contained only 825 Flannery openings it's a fair bet that the results will be inconclusive.

Oh btw some Flannery players put the weak 2♦ in the 2♣ opening, intending to pass the 2♦ waiting bid. But I suppose you would have to ignore that since if you also need to include the 2♣ opening in the analysis (and some Flannerist may play Precision or such, arghhhh) it becomes really hopeless.

I understand.

I'm less concerned with the 'advantages' or 'disadvantages' of a 1♥ opening when we play Flannery. Don't they have the same issues when pard bids a forcing NT? Presumably they are skipping spades with 4 to bid a forcing NT.

The vast majority of the gains / losses would come from the 2♦ opening (or lack thereof).

[mikestar]

True, this doesn't prove anything. But it is suggestive that further research could be worthwhile. If (as I have tended to believe) that Flannery totally sucks, then this data is a surprise.

It may well be that Flannery gains when used but the losses from the system changes outweigh it.

[han]

Imagine the following search:

Opener has 15-17 balanced and responder has a 9-count with a 6-card major. Search for hands where the final contract is 4M, opener bid 1NT and responder either bid 4M or 2M-1.

I'm pretty sure that 2M-1 will score about average and 4M-1 will show considerable positive IMPs.

Naive conclusion: Texas transfers is a great convention!

While the conclusion may be correct, the method is of course ridiculous.

[gwnn]

OK the search makes little sense, but surely that's not the point. The point is to reach a conclusion that teases Roland So the conclusion has been reached, what difference do the means make?