[Wayne_LV]

For a study I am doing I need names of players involved in a Regular Partnership on BBO that meet the following criteria:

• Play mostly IMP in Club games or tournaments where boards are played 16 times or more.
• Play at least 400-500 Boards per month, more is better
• Average IMP/Board of .5 or better is preferred
• Skill level of Advanced or better. Expert or World Class is preferred
• Rarely play against GiBs as opponents. Any board with a GiB in any seat will be excluded from the study.
• Normally play against opponents of equal or higher skill levels

If and when the results of this study are ever published online or in any other public media, express permission will be obtained before any player names are published.

If you or anyone you know is in such a partnership, please submit BBO name(s) to me in a private forum message.
To send such a message, click on my forum name. From the profile page, under my picture, click on "Send Me a Message" and submit the names.

As Bartel and Jaymes says: "We thank you for your support"

[FM75]

Are you looking for partners who play against better opponents than they are, and average a plus against them?

Or did I misread this?

[Antrax]

I think he means self-rating, so it's not that difficult. I suspect a real advanced player can easily trounce random BBO experts.

[Wayne_LV]

FM75, on 2013-May-07, 20:47, said:

Are you looking for partners who play against better opponents than they are, and average a plus against them?

Or did I misread this?

Based on limited data so far, it looks like the pairs that consistently do well against all comers at all levels are those pairs that don't make a lot of huge errors. It would be way too time consuming to review every bad board by every pair we study to determine why boards were lost by a large number of IMPs. Some guidelines are beginning to emerge. I suspect that the numbers I am about to quote are valid ONLY for regular partnerships that play their system well and play a lot of boards. The larger the sample, the more likely the conclusions will be valid.

The most important thing for this study is a lot of regular partnerships that play a lot of IMP boards that are played in a relatively large field ... i.e. Club rooms games 16 plays, larger tournaments - boards my be played 20-30 or more times. The larger the samples, the more likely statistical aberrations will be either averaged out or excluded from the data base.

The following are preliminary conclusions that well may change with more sample pairs and more overall data:
A bad board is one lost by 7 IMP or more. Below that level a lot of lost boards are not preventable for various reasons. A big swing result at another table, Opps made a great bid or play. Opps made a bad bid and got really lucky on distribution. Cheating? It does happen
I can post all IMP boards played by partners into an Excel data base and quickly get a count of the total boards and those lost by 7 or more IMPs
The ratio of preventable bad boards vs fixes seems to be about 40% fix, 60% preventable.
The better pairs will have a lower ratio of preventable losses but it looks like 40/60 is fairly typical range
However the actual ratio of preventable to fix losses is not important. If the pairs all play in a similar environment then the percentage of fixed boards (mostly stuff at other tables you cannot prevent -- the extreme being a pair goes 7NTxx-13 in your direction, you are doomed to a big loss and nothing you can do about it. Over time every pair in that universe will get their fair share of gifts and fixes and those will average out. So it stands to reason that a raw number of bad boards should provide a fair comparison over time. The better pairs will get their fair share of bad boards that cannot be prevented and by making fewer errors will reduce their number of preventable bad boards thereby lowering their composite ratio of bad boards to boards played.
So far that ratio has ranged between 1:25 to 1:12 with most being close to 1:12. IMP/Board is another indicator of performance. So far the best has been .97 and the worst -.01. Oh I have seen much much worse than -.01 but that is not players I want in this study. The average is about .3 to .5. But ......... the sample size is too small to draw ANY hard and fast conclusions yet. This is why I need more regular partners to expand the sample size. It is looking more like the group of players that interests me is a pretty small percent of the BBO players. Rather than partnership bridge, it seems BBO is made up of mostly maverick individual players that roam from partner to partner or play with GiBs in ever increasing numbers. I have found what we all suspect is true: The typical "Expert" on BBO is not even a good novice. There are, of course exceptions but few and far between. I have looked at boards played by "Experts" that play 2000 IMP boards a month, rarely with the same partner on more than one set, average losing .3 IMPs a board and having a bad board (-7 or worse) 1:10. At the same time I have observed Intermediates that have a ratio of 1 bad board in 25 and IMP/Board over .5 .

This is a long answer to a short question. It well may be the kinds of partnerships I am looking for are a practically extinct specie. So far I have yet to get ONE pair to study from this post and shotgunning random players on BBO has yielded few finds. It looks to me like online bridge is the domain of unattached players in constant search for a good partner for the moment. Kinda like cruising bars for a one night stand

[Wayne_LV]

Antrax, on 2013-May-07, 21:35, said:

I think he means self-rating, so it's not that difficult. I suspect a real advanced player can easily trounce random BBO experts.

Precisely, I am looking for winning pairs and could care less what their self rating is.

[barmar]

Wayne_LV, on 2013-May-08, 09:07, said:

Oh I have seen much much worse than -.01 but that is not players I want in this study.

Aren't you biasing your data to reach a particular conclusion? If you're trying to determine the IMPs/board for a particular class of players, it's circular to exclude players who don't achieve a particular IMPs/board.

[JLOGIC]

Good pairs don't play against randoms on BBO for very long, they end up playing against their friends and in set games of people who are generally their level.

This is going to lead to few pairs being eligible for your study. I mean, it's not fun to play against bad people who leave all the time and you keep having to replace them and you just crush them. Back in the day we used to play vs randoms with the goal of seeing the least amount of boards it took to get +100 imps. That was the only goal.

So what you're goingto have is good pairs not playing against randoms in the main bridge club, they will probably play vs people their own level and thus it will be hard to be +.5 a board. I'm sure jec/miamiwiz are +.5 a board (just looked, in the last month miamiwiz is +1.2 playing only with jec) etc, they are just so good that even though they only play vs good opps in jec matches they can average that.

[Wayne_LV]

barmar, on 2013-May-08, 09:16, said:

Aren't you biasing your data to reach a particular conclusion? If you're trying to determine the IMPs/board for a particular class of players, it's circular to exclude players who don't achieve a particular IMPs/board.

I have a sufficient number of pairs that are not winners. What I really need is pairs that are winners to balance the sample.

There are really only 2 things I expect to get from this study:

• What is an acceptable ratio of bad boards to total boards? At what point does it become obvious the partners must review all of their bad boards and look for reasons.
• Identify some really good pairs (consistently high IMP winners) and review say the top 10% of their boards by IMPs to learn what they are doing to perform so well.

And yes, I could simply pick a few of the known top players on BBO and study their best boards .......... but those players normally play systems only understood by themselves and their partners and they perform mental gymnastics of which only a rare few are capable. What I seek is how do good partnerships consistently win in the environment most of us play. I suspect the answer is they don't make a lot of dumb mistakes. But how many dumb mistakes can you make and still be consistent winners? The answer to that question is NOT zero.

Trying to measure bridge performance is in many ways like measuring apples with a micrometer. I am simply looking for some yardsticks.

[JLOGIC]

Quote

suspect the answer is they don't make a lot of dumb mistakes. But how many dumb mistakes can you make and still be consistent winners? The answer to that question is NOT zero.

Even if you answered this, how would the answer help you? You're right, the people who do well make less mistakes than their opponents. Most mistakes are dumb mistakes. If you want to improve, strive to make less mistakes. No matter how good you get it will always be a game of making less mistakes than your opps combined with some luck (many mistakes don't end up costing, some end up gaining,etc). Even the best make dumb mistakes way more frequently than people would think.

THe exact number of mistakes you can make and still win against randoms on BBO does not seem important. For starters, the most important number for that answer is what is the average number of dumb mistakes that randoms make on BBO. That is basically what you are measuring. I would look at it probabilistically, the less mistakes I make the higher % chance I have to win. The better my opponents, the less mistakes they make, so the lower my chance to win is.

The most important thing is that traditional wisdom is right on this one, bridge is a game of mistakes not brilliancies.

[Wayne_LV]

JLOGIC, on 2013-May-08, 10:22, said:

Good pairs don't play against randoms on BBO for very long, they end up playing against their friends and in set games of people who are generally their level.

This is going to lead to few pairs being eligible for your study. I mean, it's not fun to play against bad people who leave all the time and you keep having to replace them and you just crush them. Back in the day we used to play vs randoms with the goal of seeing the least amount of boards it took to get +100 imps. That was the only goal.

So what you're goingto have is good pairs not playing against randoms in the main bridge club, they will probably play vs people their own level and thus it will be hard to be +.5 a board. I'm sure jec/miamiwiz are +.5 a board (just looked, in the last month miamiwiz is +1.2 playing only with jec) etc, they are just so good that even though they only play vs good opps in jec matches they can average that.

Thanks Justin. A very good observation. I agree that IMP/Board is not a real good indicator of performance since that number heavily depends on the level of the competition. But dumb mistakes are dumb mistakes at any level. Could you give me, based on your vast experience, a ball park ratio (1 of x) of how many boards can you make dumb mistakes on and still be competitive?

Thanks

[JLOGIC]

If you make 1 dumb mistake per session (26 boards) you would have a great chance to win a national pair game and would almost always be in the top 10 (these are the elite players). If you made 2 per session you would always have a shot at winning but you'd need your mistakes to not be costing much.

I mean it's hard to say, it depends how much your mistakes cost. If my partner and I combined drop 1.5 boards a session (so 6 percent) then I feel like with above average luck we are gonna win a huge amount of the time. If we drop 2 boards a session then we might win. That is why top/bottom swing decisions are brutal, most bad mistakes will cost about half a board. So if you and your partner make 3 dumb mistakes per session total that cost half a board you are likely to win. Of course there is a lot of luck involved in terms of how many gifts/fixes you get, and how you do on the close hands (those are not gonna be dumb mistakes but you need your close decisions to be working).

This is talking about national pair games.

[Free]

For this kind of study you need to be very careful with the input data, because there's a huge trap called "trash goes in, trash comes out". Not only the studied pairs are important, but the opposition is equally important.

[Zelandakh]

Free, on 2013-May-09, 01:56, said:

For this kind of study you need to be very careful with the input data, because there's a huge trap called "trash goes in, trash comes out". Not only the studied pairs are important, but the opposition is equally important.

Quite so. Here is an example or trash in:-

Wayne_LV, on 2013-May-08, 09:07, said:

A bad board is one lost by 7 IMP or more. Below that level a lot of lost boards are not preventable for various reasons. A big swing result at another table, Opps made a great bid or play. Opps made a bad bid and got really lucky on distribution. Cheating? It does happen
I can post all IMP boards played by partners into an Excel data base and quickly get a count of the total boards and those lost by 7 or more IMPs
The ratio of preventable bad boards vs fixes seems to be about 40% fix, 60% preventable.

A 7 IMP loss is preventable? So if I bid to a 75% slam and go down while every other table is in game making, this is a preventable mistake? To be honest, I think you would be better off measuring against Par than this, despite the problems that this method has. For one thing, you might start by defining a mistake. For example, if I play a line that is 1% inferior to the best line, is this a mistake? 2%? 5%? 20%? What if the inferior line works because of the actual layout. Is that getting lucky or table feel? You cannot measure such mistakes by results. I would say that the average intermediate makes something like 2 mistakes per board or so, some serious, some minor. Justin's 2 mistakes per session seems wildly optimistic to me, unless only including large mistakes, at least judging from the Vugragh matches I have seen. But the expert mistakes are often of a different character, a slightly inferior bid or line that will not usually cost. Perhaps inaccuracy is a better term than mistake here, borrowing from chess.

[Fluffy]

fluffy-jsabate play against jrge-rval each tuesday, we make a lot of mistakes, but I'd bet way less that most pairs you will find. You will see a lot of team matches rather than main bridge club hands that won't fit your study, and once a week won't get 500 deals per month but stil....

[Wayne_LV]

This post deleted by author

[Wayne_LV]

Zelandakh, on 2013-May-10, 02:04, said:

Quite so. Here is an example or trash in:-

A 7 IMP loss is preventable? So if I bid to a 75% slam and go down while every other table is in game making, this is a preventable mistake? To be honest, I think you would be better off measuring against Par than this, despite the problems that this method has. For one thing, you might start by defining a mistake. For example, if I play a line that is 1% inferior to the best line, is this a mistake? 2%? 5%? 20%? What if the inferior line works because of the actual layout. Is that getting lucky or table feel? You cannot measure such mistakes by results. I would say that the average intermediate makes something like 2 mistakes per board or so, some serious, some minor. Justin's 2 mistakes per session seems wildly optimistic to me, unless only including large mistakes, at least judging from the Vugragh matches I have seen. But the expert mistakes are often of a different character, a slightly inferior bid or line that will not usually cost. Perhaps inaccuracy is a better term than mistake here, borrowing from chess.

Boards lost by any number of IMP's may or may not have been preventable. What I said (or tried to say) was a loss of 7 or more IMP is a preventable loss a majority of the time. To determine if a loss is preventable or not requires a board by board evaluation. If the threshold is set at -7 IMP, then MOST of the time the loss could have been preventable. Preventing a bad loss does not mean that you would win the board. Preventing a bad loss may mean taking an action that reduces the loss below the threshold. Such as doubling a sacrifice bid.

Unpreventable losses should even out in the long run, offset by the unpreventable losses of your opponents. Therefore a % or ratio of bad boards (7 or more IMP loss) to total boards played should be a number that can be used for comparison. Such a comparison might only be valid for pairs that normally play in the same environment i.e. club games and tournaments where boards are played numerous times. Team match players could only be compared to other team match players and none are included in the study. In team matches there is virtually no such thing as an unpreventable loss. If the loss was due action at another table, the perpetrators are your team mates. Results of team matches depend on the performance of 4 players, not 2 as in pairs games.


I am not looking for a microscope; a yardstick should suffice.


I am not able to scientifically determine a benchmark percentage of bad boards to total boards. But, I think anything under 5% or 1 board in 20 is probably good.

My definition of a unpreventable loss: A loss over which you have no control. A board where no bid or play by your partnership could have prevented the loss.

Some of the reasons for such losses are:

• A skewed board resulting from an unusual and totally silly action by a pair sitting in your direction. Such as 7NTxx-13.
• Your opponents bid a long shot game and make it solely due to a favorable lie of the cards and the field is in a part score.
• You bid an odds on game or slam and a low % distribution makes it impossible to make. Like a 5-0 trump split. The field magically ignores the odds during bidding and avoids the failed contract.
• The opponents make an outstanding bid or play that allows them to make their contract while the field is failing. It is to their credit, but no action by your partnership was available to prevent the loss.
• A dozen other reasons for losing a board where no possible action is available to you to change the result.

As I already stated, preventable losses will be offset by those of your opponents in the long run. Some days you eat the bear, some days the bear eats you.

[Zelandakh]

Here is another example. Say we play a more aggressive preempting style than our opponents. On a particular hand, this style gains us 5 IMPs 10% of the time; 2 IMPs 10% of the time; breaks even 60% of the time; loses 1 IMP 15% of the time; and loses 7 IMPs 5% of the time, unlikely as that is in reality. That makes bidding a winner, yet it appears in the statistics only as a preventable error. My point is that playing a higher variance style, or against higher variance pairs (aggressive preempting, bidding borderline games and slams, etc) will increase the incidence of preventable errors without actually having changed anything negative; and might even be positive. The errors are surely preventable if you can look at all 4 hands, but actually EV neutral, perhaps even EV+. Yes, in the long run such things will offset; but that takes a lot of hands. In the meantime, yes we could have bid that 25% game/slam to avoid the loss; but it was surely not an error not to do so. Similarly, I might have chosen not to open a weak 1NT and avoided that -800; but I would be missing out on all the benefits too. If I can play methods that involve us making 10% more errors but induce the opponents to make 30% moer errors, then I will probably find this to be a good tradeoff. In this respect I disagree with the basic premise that making more errors necssarily means we are worse. Rather, we need to make fewer errors than the opponents - "Winning Ugly" as Brad Gilbert would say.

[barmar]

Zel's example can be viewed two ways:

1. Zel's point: It's a mistake to look only at preventable errors if you don't also take note of "preventable successes". If the latter outweigh the former, you're playing winning bridge.

2. Suppose you play the same aggressive style, but somehow add the judgement to know when to rein it in. You'll have the same winning boards, but you'll reduce the big losses. That will make you even better than you were before. Finding the players who are able to do this seems to be the point of the study.