[inquiry]

I did a short little BridgeBrowser study. Here was the conditions.

1. Dealer held 11-15 hcp, and precisely 5 hearts, precisely 4 spades (any other)
2. Dealer opened something between 1♥ and 2♥

The question is, what bid worked best?

In the one database I checked, I found 121,963 hands that meet the above requirements (dealer pass would not meet the requirements, dealer had to bid to be included)

1♥ was opened 96.8% of the time (118,145 times)
2♦ was opened 2% of the time (825 times)

Other opening bid include 1♠, 1NT, 2♣, and 2♥.

Here are the imp averages for these bids...

1♥ 0.01 imp
1♠ -0.66 imp
1N 0.29 imp
2♣ -0.38 imp
2♦ 0.29 imp
2♥ 0.56

Note, 2♥ in these times was "flannery 2♥" and 2♦ was flannery. Each of the hands were 2♦/2♥ was opened, obviously the majority of people opened 1♥, and yet, the flannery bid did "ok".

BTW, I do not play flannery, because i have a better use for the bid.

[pclayton]

Ben, I don't think this means much, since we don't know what the IMP averages for a 2♦ led to when it wasn't flannery.

[hrothgar]

Why would Roland care? It's not like the results of this type of study mean anything...

People have noted many times that drawing inference from this type of data dredging is wildly inappropriate.

Case in point: Let's look at what appears to be a Flannery type 2♥ opening which is scoring a ridiculous +.56 IMPs per board. What does this number actually tell us?

1. Does the +.56 IMPs per board measure the effectiveness of the opening bid or the skill level of the players who use this opening? (I'll be dollars to donuts that there is a statistically significant relationship between the two)

2. Even if we went and compensated for the skill level, its still impossible to judge these sorts of bids in isolation from the rest of the system... Let's assume that I made the decision to devote one of my opening bids to show some specific hand type (perhaps a 2♥ opening that shows 5 Hearts and 4 Spades and 11 - 15 HCP). Having assign a nice precise meaning to this bid, I damn well better score well if/when it occurs. However, we have no way of knowing whether the gains that we enjoy from this opening offset the strain that this very specific opening places on other parts of the system...

[rogerclee]

Omitted Variable Bias.

If you want to perform a statistical study like this, you're going to need to factor "skill level" into this. This could be measured by considering the average IMPs/bd for each of the flannery bidders, then considering correlation again.

I think you'd also need to consider the IMP losses from not opening 2♦ on weak hands.

(FWIW, I think Flannery is okay .)

[han]

We would always expect regular partnerships on BBO to do significantly better than two random players who sit down right? The people who play Flannery and especially 2H as Flannery are much more likely to are regular partners, and at least they have had some discussion. Therefore, you would expect them to score significantly above 0.

BTW I think I know what the word "yet" means but I don't understand its function in your next to last sentence.

[jtfanclub]

I have to agree with Richard. The problem with Fl*****y isn't that it does poorly. It does great. If I had a two diamond opening that showed 6-6 in the majors, it would probably do even better. It just doesn't come up often enough to be worth spending a bid on.

I play a modified Flannery in which a 2♦ opening is 4+/4+ in the majors, usually not 4432. Other people playing Precision may open 2♦ with 4513 or 4504 shape (along with lots of hands without 5 hearts), but not with two or more diamonds.

Is it Magic Diamond where a 2♥ opening shows 11-15 hcp, exactly 5 hearts? I know I've played against it, but never learned it.

[han]

Also agree with hrothgar's second point of course, but I think the skill level part is most important. If it wasn't for that, we would at least be able to conclude that if one decides to play Flannery, opening 2D is better than opening 1H. But even that isn't the case.

[helene_t]

I don't think I know any bad players who play Flannery. Bad players may play multi or wilcocz or mini-roman, but not Flannery.

Anyway, it should be possible to compensate for that by looking at the IMPs for 2♦ openings minus the average IMPs for the same partnerships. This still doesn't tell us if Flannery is a good convention since the most important thing is how the 1♥ opening is affected by playing Flannery, how the pass/1♦/3♦ openings are affected by playing a natural weak 2♦ etc.

[Free]

Second all the comments that you can't draw the right conclusions from this set of data.

To add one more thing: you don't know WHY these openings were made. Some probably opened 2♣, 2♦ or 2♥ as a weak hand with both Majors, so obvious these score better since responder will stay low when it's right. Some opened 1♥ as 14+ (Fantunes) which might get you higher when it's wrong (invite with 8-9), or 2♥ as 10-13 (Fantunes) which puts up the pressure and lets opps make mistakes. A 1♥ opening in 2/1 also has a different range. There are probably a lot more meanings for various opening bids...

All this aside, it's obvious that a bid that describes your hand so accurate scores better (if I could open 2♦ showing a 4=5=3=1 with 13-14HCP, 4-5 controls, and no lost values in ♣, partner will know what to do very well!). But the question is: is it worth the sacrifice of another opening bid? Will the gains compensate the losses?

So, with this set of data, you can't conclude if the opening is really the cause of the better results or if it's the player or another meaning of the opening bid. And even if you could conclude something, you still need another set of data to determine the loss of several kinds of 2-level openings before you can make an accurate analyse about the efficiency of the opening.

[gwnn]

helene_t, on Feb 16 2008, 03:31 AM, said:

I don't think I know any bad players who play Flannery. Bad players may play multi or wilcocz or mini-roman, but not Flannery.

So you don't know me any more?

[rbforster]

I agree with all the caveats put forward so far - different skill levels, costs of alternative bids, effects on overall system effectiveness, etc, etc.

However, that doesn't mean that this sort of study is without merit. I would be happy to do a straight up comparison of the IMP gains of

1. a "natural weak 2♦"
2. the above mentioned 2♦ Flannery results

and if 2♦ scored a bigger average IMP gain than 2♦ weak, I'd think about adopting it (shudder). I feel like all the other factors are likely to be secondary or less to the cost of the alternative use of the bid.

Of course since 2♥ is better as Flannery than 2♦ (on general grounds as well as in the above standings), I suppose you should want to compare 2♥ weak with this as the alternative instead (or 2♦ multi instead of 2♦ natural, and 2♠ something else... gets more complicated).

[hotShot]

The big problem with a study as given above, is that it misses several points.
They have already been given and I would like to focus on this one.

Lets take a Precision 1♣ opening, it's biggest impact on your bidding is not in those cases where you open 1♣. The biggest impact on your system is that every other bid is limited to 15- HCP.
So even if the 1♣ bid would produce a negative result, your overall benefit might be be positive, because you benefit every time you don't open 1♣ which is much more often.

To make Ben's study interesting, we would need to know how well those who open Flannery did compared to the rest of the field, when they opened 1♥ (or 1♠ if the play canape style). It might be possible that Flannery openings just move some hands that would also work well with a 1♥ opening to another bid, making the 1♥ bid less effective and losing a better use of the 2♦/♥ bid.

The "skill factor" would not be much of an issue, if there are enough scores for the boards used. Stupidity on both sides should average out.

[rbforster]

hotShot, on Feb 17 2008, 05:47 AM, said:

To make Ben's study interesting, we would need to know how well those who open Flannery did compared to the rest of the field, when they opened 1♥ (or 1♠ if the play canape style). It might be possible that Flannery openings just move some hands that would also work well with a 1♥ opening to another bid, making the 1♥ bid less effective and losing a better use of the 2♦/♥ bid.

I don't think the issue you're focusing on is that significant. 45xx is a pretty rare specific shape out of all possible shapes that normally open 1♥ playing 5 card majors. I get about 1.5% of all shapes being 45xx, while normal 1♥ openers (including these hands) are almost 10%. This makes the Flannery hand only about 15% of normal 1♥ openers, and then there's still the question of what fraction of the time you would reach a normal contract (4H/4S/3N etc) regardless of whether or not the Flannery shape was an option. In short I wouldn't expect its presence or absence to have a big effect on the rest of the system. That's why I think in this case a straight comparison with 2♦ natural and 2♦ Flannery is likely to be reasonably indicative. The precision example you give is much more complicated.

hotShot, on Feb 17 2008, 05:47 AM, said:

Stupidity on both sides should average out.

Amen to that - plenty of stupidity out there to go around. More to the point though, arguments like "better players play/don't play convention X and therefore declare or defend better" often misses that better players typically play against better opponents if given the choice, so even if there is an expert preference for a particular convention it's unclear if there would be a relative scoring advantage from this.

[awm]

I'd suggest a few amendments to this if you want to more accurately evaluate flannery. In particular:

(1) You need to hold the skill levels fixed. Ideally you'd want to hold "partnership ratings" fixed to account for the advantage of regular partnerships, but this may be harder to manage.

(2) You also need the comparison of 2♦ vs. pass on "weak 2♦" hands. It seems clear that flannery should benefit when you open flannery, the question is whether it's worth what you lose when you have to pass with a weak 2♦.

(3) The relative frequencies of weak 2♦ vs. flannery also need to be taken into account -- if on average you win 0.4 imps every time you open flannery and lose 0.2 imps every time you have to pass with a weak 2♦ bid, but weak 2♦ is 3 times more frequent than flannery, then you're still losing on average...

[jtfanclub]

Rob F, on Feb 17 2008, 06:55 AM, said:

In short I wouldn't expect its presence or absence to have a big effect on the rest of the system.

Sure it can. For example, I play that a 1♠ response to 1♥ promises 5, not 4. That also means that the 1NT response to 1♥ (forcing) can have 4 spades, making it tougher to interfere. I don't know if it's better or worse, but it's certainly very different.

[benlessard]

Ive wanted to buy to buy bridge browser but ive chicken out after checking the average level of the players on BBO. I truly believe that database are the future for bidding judgement except that there is so many clueless players that are corrupting results.

[kenberg]

The data, no doubt, should not be taken at face value. Most things shouldn't be taken at face value. Trying to look forward, I offer a suggestion for this and other studies. Hopefully it would be easy to implement on BridgeBrowser.

Every time you find an auction where the criteria are met for inclusion, look not only at the score on this board but the score on the next board or the previous board that the foursome played. If the foursome did not play either the next hand together or the previous hand together, then dump the current hand from your data. This would not, by itself, eliminate the objection that the pair using, say, a 2H bid to show 5-4 major distribution is getting the good result because they are a regular partnership, but it could be evidence in that direction. Say that on the Flan hand they get their .56 average imp but on the companion hands they break even. That would be at least somewhat unexpected and therefore interesting. There is still the issue of "when you are doing X you are not doing Y". That is, if 2H shows 5-4 then you cannot open a weak 2H. Maybe they have another bid for it, maybe not (but if 2D shows a weak 2H then 2D doesn't show diamonds, etc). That's harder to deal with via BB I think. But at least you could, partly but not wholly, deal with the "well, they are an experienced partnership" objection, both here and in other studies. Assuming BB allows this sort of maneuver.

This whole business of drawing conclusions from data is highly misused. My own pet peeve regards education. Someone noticed that kids who take algebra in the eighth grade are more likely to go on to college than those who don't (duh). So they started stuffing kids, ready or not, into algebra in the eighth grade. There are correlations and there are causes, and they aren't the same thing.

[han]

Rob F, on Feb 17 2008, 04:23 AM, said:

I would be happy to do a straight up comparison of the IMP gains of

1. a "natural weak 2♦"
2. the above mentioned 2♦ Flannery results

and if 2♦ scored a bigger average IMP gain than 2♦ weak, I'd think about adopting it (shudder). I feel like all the other factors are likely to be secondary or less to the cost of the alternative use of the bid.

I think you are really underestimating the significance of these other factocs. The level of the players and partnerships involved will have a much greater effect on the results than what system is being played.

[helene_t]

I think the fair thing to do would be to look at the results from Flannerist obtained when they
1) Open 1♥ or 2♦, or pass with a hand which non-flannerists are likely to open 2♦
2) Do something else (e.g. open 2N with a balanced 21 count, fail to open before opps open in 1st seat, etc).

If Flannery is any good, their results from 1) would be better than their results from 2).

There are some minor flaws in that analysis. For example it could be that the specificity of the 3♦ opening is affected by the lack of a natural 2♦ opening, but I wouldn't worry too much about this.

In order to identify the "Flannerists" one would have to filter on tables where at least one Flannery-opening was recorded. So the amount of data would be limited.

I agree with Han that the effects of player strength and level partnership understanding are more important. If done properly, I would expect the analysis to reveal a nett effect of Flannery (or any other non-essential convention) very close to zero.