[han]

I've been reading the latest of these endless multi discussions. There are a lot of claims out there, but none of these claims is backed by any evidence.

Is anybody interested in collecting data about 2-openings? I'm thinking of getting a group of players together who enter their own results with their 2-openings in a spreadsheet. If we get a fair number of people to join, perhaps we can get enough data to draw some conclusions. What I suggest is this:

- We get a group of people who play "serious" with a regular partner. Only results from "serious competition" count, and only IMP results. What constitutes "serious" we can discuss, but in the end people should decide for themselves.

- If you agree to join, you enter in advance which 2-openings you play. I'm hoping for several partnerships for each of the dominant opening methods (weak 2's, multi&muiderberg, precision, ...)

- You enter some data into a spreadsheet after each serious game you play. For each 2-opening that came up you enter how many IMPs you won or lost. Also, you enter how many hands you played in total and how many IMPs you won or lost in total.

- If the game is a team game, you count IMPS won or lost in comparison with your teammates. If it is IMP pairs, you enter that.

We need some clever people to set up a joint spreadsheet that does some smart things with this data. We have plenty of know-how here on the forums.

[Fluffy]

I´d like to join, But I play actual serious IMPs twice a year. On the other hand I play with my partner Team matches each tuesday night. Sometimes the competitors are serious (had the luck of catching Phil and sathya 2 weeks ago), sometimes are just BBO experts.


We open 2♣ precision, 2♦ weak 2 in a major, 2♥ hearts+ minor, 2♠ spades+ minor, 3♣ for both minors (2NT 1 minor)

[BunnyGo]

I'm with Fluffy. I think it's a great idea, but I'm not sure any of the events I play qualify as "serious". And those that are close, I play at most once, maybe twice, a year.

I do play online a bit, but I don't think that's "serious". In f2f right now I'm experimenting with intermediate 2 bids, but so far haven't found a good game to try them in (they work great in a weak field).

[han]

I'd say, use your judgement for these team matches.

What are your 2-openings Gonzalo?

Here are mine (fairly standard here):

2C = weak with diamonds or strong.

If strong, either balanced 22-23 or 26-27, or hearts, or spades, or clubs (could be a 2-suiter involving clubs).

2D = weak with a major or strong.

If strong, either balanced 24-25 or 28-29, or diamonds (could be a 2-suiter involving diamonds).

2H, 2S = weak, 5+ major, 4+ minor. Minor must be 5+ when vulnerable (except in third seat).

[karlson]

I like the idea but it feels a little weird to me -- quite often in a local tournament (even a serious one) many people play the same methods whether those methods are weak 2s or multi. It doesn't seem to make much sense to enter imps won/lost if the opening was the same at the other table. For more offbeat methods of course it could work well. Perhaps using archives of international competitions makes more sense.

[gwnn]

sorry to interrupt, but what do you do with a two-suited hand with majors?

[mich-b]

I am willing to help and provide data from the boards I played.

I am not sure though, how exactly this data should be analyzed.
For example , in my last serious event last weekend , I played 56 boards (teams), in which we opened a 2 level opening once - a 2♦ Multi which was Passed out, making 90. (Opener had spades , dummy had 6 diamonds and 13hcp). Our teammates scored 170 in 3♥ so we won 6 IMPs. Well, was does that mean? Did we win IMPs because of the Multi? Probably not , probably the reason we won IMPs was that our player judged to overcall 2♥ , while our opponent didn't.

[Fluffy]

mich-b, on 2011-November-14, 07:41, said:

I am willing to help and provide data from the boards I played.

I am not sure though, how exactly this data should be analyzed.
For example , in my last serious event last weekend , I played 56 boards (teams), in which we opened a 2 level opening once - a 2♦ Multi which was Passed out, making 90. (Opener had spades , dummy had 6 diamonds and 13hcp). Our teammates scored 170 in 3♥ so we won 6 IMPs. Well, was does that mean? Did we win IMPs because of the Multi? Probably not , probably the reason we won IMPs was that our player judged to overcall 2♥ , while our opponent didn't.

If on a large data, you win av .3 IMPs when you open multi versus win .1 IMPs when others open 2M then multi will win over 2M for example.

[Free]

It's a good idea, but why do you restrict this to imps only? It might be interesting to see how well these openings perform with different scoring methods as well imo. For all we know, multi may be a loser in imps and a winner at MP.

However, I agree with some posters that it's difficult to interpret the data. Results from preempts depend a lot on style and on our opponents' actions over them. To have some meaning, you also need the preemptive structure from the other table, so you can compare one system over another. Moreover, when they play the same openings at the other table, it's quite impossible for the system to win/lose any imps...
And what about card play? When you have a 2-way finesse, you can win or lose. Does the system even matter in this case? Should the system be rewarded/punished because neither we nor opponents had any indication but we choose a different side to finesse?

I also have an example of multi which explains the difficulty of analyzing data: it was a MP event, NV vs V I open 2♦ minimulti in 1st seat. LHO thinks for a while and passes. Obviously my partner passed his 7 count, because we know the laws and RHO can't bid unless he has a clear reason to do so (thanks to his partner's long pauze). So we played a simple 2♦-2 NV in a 3-3 fit, while opps had 3NT available with 25HCP between them. So we won on this board because LHO hesitated! Does that count?

We can easily calculate our results after the preemptive openings, but it's very hard to know where that result comes from. Is that score a result of the opening itself or are other factors responsible?

Defining what events are serious enough will also be important. I would think that the better the opposition, the more reliable the result should be. However, if you look at BB finals for example, you see that even the best players make mistakes on defense, declarer play, bidding, lead,... So basically the data can only give some indication, but sadly it can't be considered proof.

[hrothgar]

Fluffy, on 2011-November-14, 07:50, said:

If on a large data, you win av .3 IMPs when you open multi versus win .1 IMPs when others open 2M then multi will win over 2M for example.

This sounds intriguing, however, I'm worried about the practicality of generating reliable results.

I spent some time look at board results from the Bermuda Bowl.
There is enormous amount of variance, even at the top level's of play.
One day, USA2 is beating up the Dutch, the next day the converse is true...
I suspect that you're going to need very large numbers of samples to reach any kind of reliable estimates.

Associated with this, there are going to be all sorts of weird cross effects that you need to worry about.

For example, what if Multi 2D + Muiderberg works brilliantly within the context of Polish Club but miserably within a Precision framework?

Alternatively assume that my defense is much much better than my declarer play. In that case, I might score much better with a sound opening system than a light opening system.

From my perspective, if folks are seriously interested in this topic, the best course of action would be to try to get Jack's developer's interested. (or any other serious software group)

There are a lot of advantages to being able to run this all using computer players. Once this has been coded up,

1. You can run large numbers of simulations at a relatively small cost
2. You have identically skilled players competing
3. You can play the same hands multiple times without worrying that folks with remember them

[phil_20686]

There is also the problem of self selection bias:

If you preempt more often you expect to gain less per hand on your preempts, but, presumeably, more overall. However, different strategies from "perfect hands only" to "pretty much any grotty weak two" will be vastly different in terms of the number of hands a pair opens and the chance of success in a given opening.

In order to get around this one needs to look at total imps won or lost by a given bid over X boards. This could of course be done equally of course, but my point is that if you design a spread sheet that contains only information on the hands where you bid, without knowing how many you passed, you will have a problem.

[Zelandakh]

There are further complications which this project would also not address. If you play a mini-NT with assumed fit 2 bids these interact with each other in a way some players believe is beneficial, and especially this has an impact on third seat which this does not address. Similarly for Flannery, whose main benefits come when opening 1H and not from the bid itself. I daresay that the same could be said for Ben's Roman 2M openings (although his data might contradict this).

An alternative way of running this project would be to take a (hopefully) large number of volunteers who are willing to vary the destructive part of their systems in a major way. Then play 3 weak 2s for 6 months, followed by 6 months of Dutch 2s, followed by 6 months of 2-level weak 2s, and finally 6 months of Flannery with weak 2Ms. Then you can analyse the data not only across the overall spectrum by 2 level openings but also in relation to the performance of the same pair using different methods both by 2 bid openings and for the overall system.

[hrothgar]

Zelandakh, on 2011-November-14, 12:04, said:

There are further complications which this project would also not address. If you play a mini-NT with assumed fit 2 bids these interact with each other in a way some players believe is beneficial, and especially this has an impact on third seat which this does not address. Similarly for Flannery, whose main benefits come when opening 1H and not from the bid itself. I daresay that the same could be said for Ben's Roman 2M openings (although his data might contradict this).

An alternative way of running this project would be to take a (hopefully) large number of volunteers who are willing to vary the destructive part of their systems in a major way. Then play 3 weak 2s for 6 months, followed by 6 months of Dutch 2s, followed by 6 months of 2-level weak 2s, and finally 6 months of Flannery with weak 2Ms. Then you can analyse the data not only across the overall spectrum by 2 level openings but also in relation to the performance of the same pair using different methods both by 2 bid openings and for the overall system.

[hrothgar]

Zelandakh, on 2011-November-14, 12:04, said:

There are further complications which this project would also not address. If you play a mini-NT with assumed fit 2 bids these interact with each other in a way some players believe is beneficial, and especially this has an impact on third seat which this does not address. Similarly for Flannery, whose main benefits come when opening 1H and not from the bid itself. I daresay that the same could be said for Ben's Roman 2M openings (although his data might contradict this).

An alternative way of running this project would be to take a (hopefully) large number of volunteers who are willing to vary the destructive part of their systems in a major way. Then play 3 weak 2s for 6 months, followed by 6 months of Dutch 2s, followed by 6 months of 2-level weak 2s, and finally 6 months of Flannery with weak 2Ms. Then you can analyse the data not only across the overall spectrum by 2 level openings but also in relation to the performance of the same pair using different methods both by 2 bid openings and for the overall system.

In order to do this seriously, we'd want to start with a full blown Design of Experiments.

There's a reason why I suggested that getting the Jack developers involved would be a good thing...
Doing this right involves a LOT of work...

[Phil]

Happy to participate.

We play Flannery and weak 2's. I would think you would also be interested in our results when we pass up a weak 2 in diamonds or when we choose to distort and open a 3 bid.

[Hanoi5]

I'm with hrothgar. This could end up in a contest of who plays better instead of what system is better.

Maybe an equitable solution would be to have both 'studies': all parties interested tell their 2's opening results and the complete hands are posted (saved, whatever) and then used in the experiment.

[32519]

There is another way that you can gather reliable data:

Many top-flight matches are broadcast live on BBO. No doubt many of the regular contributors to the BBO Discussion Forums kibitz some of these matches. As soon as anyone opens the bidding with two of any suit, the nature of the bid will be apparent (weak/strong/artificial/Precision etc). The result obtained can be compared with the result from the other room if a different opening bid was used there.

Anyone with true passion and dedication to put and end to this dicussion can visit the Vugraph Project http://www.bridgetoe.../index.php/home Hand records going all the way back to 1955 can be recovered here. There you can sift through endless hands and results for the data required. I believe that Pietro Campanile did something similar when he released his findings on this issue. http://www.migry.com...dfs/multi2D.pdf His data however was limited to cetain tournaments and years. Here is an extract from his findings
"The data includes all the deals played in the European Championships from 1997 to 2004 and the knock-out stages of Olympics and World Championships from 1987 to 2003, whenever there are comparisons from all tables in play."

The information is all available right here on BBO.

[Free]

32519, on 2011-November-15, 03:05, said:

As soon as anyone opens the bidding with two of any suit, the nature of the bid will be apparent (weak/strong/artificial/Precision etc).

When I open 2♦ with a long ♠ suit, do you know if 2♦ includes any strong possibilities as well?

[hotShot]

Wouldn't it be easier to set up a series of BBO tourneys with predealt hands that cover the range of all suggested weak 2 openings from 2 suited to one suited (without ♣) from 3 to 11 HCP.
This would generate a lot of results, you would have the full hands available, you could run DD or SD play analysis ....
If you like restrict access to established partnerships.

The agreements should be posted here, so everybody can prepare a defense.

[32519]

Free, on 2011-November-15, 03:38, said:

When I open 2♦ with a long ♠ suit, do you know if 2♦ includes any strong possibilities as well?

If there is any uncertainty as to the possible meaning of any bid (your example here containing weak only or strong options as well), try finding the Convention Card of the pair in question. ECatsBridge has a lot of them. I haven't looked myself, but maybe the CC's are also available on the World Bridge Federation website http://www.worldbridge.org/ Other contributors to this forum will no doubt know of other websites where the CC's can be obtained. Often the Vugraph commentators have a copy of the CC in their posession during commentary. I am convinced that Roland Wald can be contacted to obtain a copy of the CC if needed.