[zasanya]

I see a significant number of people trying to 'improve' the self-rating system of BBO.The idea appears to be to expose those who wrongly call themselves experts .Why are so many people interested in exposing phoney experts?
Personally I play with some unknown people and a lot of known people.If I meet somebody whose bridge skills impress me I put him as a friend.I try to include him im my team matches without bothering about his skill level.I have people ranging from world class with stars to novices in my Friends list.
My question is why worrry about the self rated skill levels of others?How does it matter if your purpose in being a member of this wonderful site is to enjoy bridge?

Aniruddha

[barmar]

If you're playing in an Individual tournament, and your partner rates himself as an expert, it may set some expectations. You would make some bids or plays that you think an expert is more likely to understand than a novice. You might bid more aggressively, expecting him to be able to play the cards well.

Online bridge is frequently anonymous like this, so the ratings are the only thing you have to go by. If they're totally meaningless, then you're left guessing. As well, why waste space in the profile for a meaningless designation?

[Free]

I'm not really obsessed with ratings, but I find it a challenge to find a fair system to rate bridge players if they only have 1 account. Fact is that you play with several people together, and a skill level is private, which makes this a difficult problem...

[hotShot]

If you try to get better, it is nice to have some means of measurement to evaluate your performance.
If you meet someone and you want to compare your skills, a measure would be handy.
If your sports is marathon, you can use the time as a measure.
So there is a need to have some mesurement of bridge skill.
But what could be a measure in bridge?
Masterpoints won't do, because they include the size of you bridge club and the number of events you played in. Someone who made 1MP's in his first tourney may be much better than someone who collected 20MP's in 40 years.
MP-percentages won't do, because the strength of the field is an important factor. Even in an all expert field some player have to have less than 50%.

So it is hard to think of a ranking system that works.

When people log in at BBO they do that for different reasons.
Some just want to play a few boards in a relaxed atmosphere, others are looking for a tough fight.
Depending on your own level, novices, beginners or intermediates are no competition especially if you play with a regular partner against some first time partnership.
If you partner plays a contract where you think xx tricks are fool prove and he/she makes less than that, you might be a little upset.
If you are no match for your opps, they might find it booring to play with you.
This is where the trouble starts, if you host a tourney for expert+ players, every player who would like to play, needs to set his level to expert.
......

[ehhh]

Ratings are an honesty issue. Generally you feel more comfortable playing 'with your own kind' though playing up as in golf will help you improve your game so long as 'up' isn't so high as to embarrass yourself. I note everyone I play with as their rating readily becomes apparent. As such, ratings as nebulous as they are, are just a benchmark no more no less but nevertheless in the cyberspace of bridge necessary as they are in ftf.

[Chamaco]

ehhh, on Oct 9 2006, 03:03 PM, said:

As such, ratings as nebulous as they are, are just a benchmark no more no less but nevertheless in the cyberspace of bridge necessary as they are in ftf.

How necessary are ratings in ftf bridge ?

[bid_em_up]

ehhh, on Oct 9 2006, 10:03 AM, said:

As such, ratings as nebulous as they are, are just a benchmark no more no less but nevertheless in the cyberspace of bridge necessary as they are in ftf.

Ratings should be reasonably unecessary in f2f bridge.

In f2f bridge, you already know (usually) the skill level of your partner (unless you paired at the partnership desk, and even then you have the opportunity to talk to them before deciding to play). In many cases, you know most of the opponents as well, unless you are playing in an out of town tournament.

Online, you usually dont have this information available to you prior to making a decision regarding playing with a person or deciding to allow a person to sit at a table as an opponent.

My problem with inaccurate self-ratings is.....I really dont enjoy beating up on people who claim to be expert, but are no better than intermediate. Its not fun or challenging for me, and its usually not much fun for them. Nor do I enjoy playing with them as my partner.

If you tell me you're an intermediate player and we agree to play, and you make a mistake in the bidding/play/defense, then I expect it to occur and will be happy to explain what you could have done differently (if you wish to hear it), so that you can improve. If you dont wish to learn but just want to play, thats fine also, but at least the mistakes are expected to occur.

If you tell me you are expert, and consistently play like a beginner/intermediate, I will mark you as "not expert" in my player notes for you and move on. If I consider you a "friend", I will politely suggest that you change your skill level to a representation that is more indicative of your skills.

But I would much prefer to be able to sit down at a table, ask for advanced+ opponents and get someone who is actually at those levels than have to constantly continue to sort out/weed through bad players who are claiming to be expert.

I get really tickled (or annoyed) with players whose profile will claim Expert status, and then it says "16-18 NT, no transfers, strong twos, blackwood".

[inquiry]

I don't think a rating system would improve our favorite site. But I have to say that I have a way of calculating such, and using it it to comapare a lehman like score to average imps/mp per hand. Doing this it quickly becomes clear that it is not a good idea to use average imps or average mp, unless the player you are checking plays against a lot of different opponents. Some people play team games against weak opposition and their average imps/mp will be fairly high.

I can tell you, if you are intersted, the scores I have calculated has fulvio2002 at 70.12, fred at 67.83, ritong at 63.0, Roland at 61.06, sillafu at 58.32, me at 56.18, two member (or past members) of the BIL at 42, and 46. Here 75 is basically a theoretical and unreachable maximum, and 25 is a theoretical and unreachable minimum, with 50.0 the proposed average. If you have a good day, your score moves up, a bad day down, and it can move down lot in a short time, or up a fair amount too... Thus, the VAST majority of players are between 48 and 52. Complete novices fall quickly into the low 30's, I have only seen one player in 20's. Beginners are in the mid to upper 30's. Intermediates would probably be in the low 40's. Advanced in the mid 40's to low 50's, experts in the mid 50's to low 60's, and world class in the upper 60's and low 70;s. There ratings can not be perfect, for one thing sillafu score seems low.For what it is w

The reason why I say it is not a good idea to use high imps/board is that some people with high imps per board have a fairly low rating (a kind of bunny bashing where beating up on weak players does not help your rating, and even though they can't see their rating, these people just can't help themseovles... or maybe they are teachers).

[hrothgar]

One quick comment on the topic of ratings:

I've always found it very curious that people have such an enormous interest in rating individuals, but (typically) show very little interest in rating partnerships. Personally, I believe that it is significantly easier to rank partnerships than individuals. Case in point:

Lets consider 4 well bridge players

Fred Gitelman
Bred Moss
Jeff Meckstroth
Eric Rodwell

Assume that you were forced to play this team for money. Which lineup would you rather face?

Gitelman - Moss / Meckstroth - Rodwell or
Gitelman - Meckstroth / Moss - Rodwell

I'd be willing to bet that most people would rank their odds are significantly better facing the second lineup. (I doubt that many folks would claim that their chances are good, mind you, just better) Silly example, I know, but I think that the point is valid. If you focus on partnerships rather than individuals, you've removed a lot of complexity from the problem space.

Personally, if I were doing any serious work on this problem, I'd focus on getting a good solution for partnerships. After which, I might consider trying to decompose individual ratings from the different partnership ratings.

[DrTodd13]

All this depends on the purpose of the rating. Some people might want a rating to track their own progress as an individual or their progress as part of one or more regular partnerships. Some people might want others rated individually for the purpose of finding a reasonable partner or opponents or teammates. A rating that includes some measure of friendliness or stick-to-it-ness would be useful for finding a partner or opp. You want someone who is nice who isn't going to bolt the first time somebody does something a bit odd. Conversely, a rating to track your own progress should not include such features. For rating your own progress you might even like the ability to selectively exclude boards where partner took some ridiculous action. I suspect that people are most interested in finding potential partners, then tracking their own progress, then tracking partnership progress. So, while partnership ratings may be easier than individual ratings, there are so few regular partnerships playing against each other right now that I fear the graph would be so disconnected that even getting a partnership rating would be difficult. I still contend that peer-rating is the way to go to solve the first problem of finding matching partners or opponents and it doesn't suffer from the problems of the Lehman system.

[pclayton]

I suppose if I was motivated, or cared, I could start a website where people could enter the player's handle and rate them on a 1 to 10 scale. Or 1 to 100.

Is this system error-prone? Not neccesarily. Different weights to how someone rates another could be based on the RATER's own ability. I'd rather trust Fred's opinion of my game, then some random that sits at my table and leaves. But Fred's opinion would carry a higher weight than the random.

The wings could discarded, and a sensible # could be obtained. After awhile, the system would work if it was used enough.

Abuses would be easy to spot too.

[DrTodd13]

pclayton, on Oct 11 2006, 01:46 PM, said:

I suppose if I was motivated, or cared, I could start a website where people could enter the player's handle and rate them on a 1 to 10 scale. Or 1 to 100.

Is this system error-prone? Not neccesarily. Different weights to how someone rates another could be based on the RATER's own ability. I'd rather trust Fred's opinion of my game, then some random that sits at my table and leaves. But Fred's opinion would carry a higher weight than the random.

The wings could discarded, and a sensible # could be obtained. After awhile, the system would work if it was used enough.

Abuses would be easy to spot too.

Also need to weight the review based on number of boards played together. You could get that from myhands I suppose. Big problem is that any external website wouldn't be used enough to matter. It really has to be integrated into BBO or it won't be used enough.

[helene_t]

DrTodd13, on Oct 11 2006, 11:08 PM, said:

For rating your own progress you might even like the ability to selectively exclude boards where partner took some ridiculous action.

I know that strategy from clinical research - if you selectively exclude patients or lab animals that behaved in an "odd" way, you will end up with data that are biased towards confirmation of your hypothesis because you (unconsciously) tend to rate behavior as "odd" if it contradicts the hypothesis you want support for.

The same will happen here: a "ridiculous" action may be included if it turns out well and excluded if it turns out bad. Also, if my partner's take mor ridicolous actions than other people's partners, it may something about me. It could also say something about my partners, of course, but the Lehman algortihm handles that automatically.

Ridicolous actions are parts of bridge and must be included in the statistical analysis.

[mike777]

"Ridicolous actions are parts of bridge and must be included in the statistical analysis."


Why? Who says what must be included and why? This seems to show a bias between Math and real life. I assume there is Math for nonlogical decision making?

Call it Congnitive errors or whatever.

If you want to declare some Math/logic bias ok but if you want to include all nonlogical real life bias so be it .

[helene_t]

hrothgar, on Oct 11 2006, 10:18 PM, said:

Assume that you were forced to play this team for money. Which lineup would you rather face?

Gitelman - Moss / Meckstroth - Rodwell or
Gitelman - Meckstroth / Moss - Rodwell

I'd be willing to bet that most people would rank their odds are significantly better facing the second lineup. (I doubt that many folks would claim that their chances are good, mind you, just better)

There is some truth in this, of course, but I think the importance of well-oiled partnerships is overrated.

Jansma-Muller just won the Dutch Pairs Championship. They are not a regular partnership, allthough they know each other well because both are part of the national selection. Muller used to play 4-card majors but switched to Relay Precision one and a half year ago. Jansma used to play Dutch Doubleton but switched to an advanced version of Polish Club a couple of years ago. But expertscan play well with non-regular partners as they have a good understanding about what is "standard".

An example from another part of the spectrum: in Riccione, I played with a pick-up partner who spoke very little English and was used to playing Blue Club with Italian carding, something I have no experience with at all. All other pairs at least joined the festival as pairs and many had very detailed CCs. We did better as a pair than our combined results in the indy. Probably, we were fortunate both to have a good understanding of the bridge-cultural diversity so that we had less misunderstaings than most intermediate pick-up partnerships from different cultures. But intermediate players from different cultures generally try to avoid each other as pick-up partners.

Two Gozillas may have slightly more absurd bidding misunderstandings if they are a pick-up pair than if they are an established partnership. But as a pick-up partnership, they have some advantages:
- they can say "no agreement" whenever opps ask them for explanations, so that their confusing actions will not lead to adjustments.
- they know that partner is unlikely to understand anything so they do not base decisions on false assumptions of understanding
- they don't get angry when partner misunderstand their bidding since they didn't expect otherwise.

[helene_t]

mike777, on Oct 12 2006, 11:26 AM, said:

"Ridicolous actions are parts of bridge and must be included in the statistical analysis."

Why? Who says what must be included and why? This seems to show a bias between Math and real life. I assume there is Math for nonlogical decision making?

If you want to compute your average score restricted to non-ridicolous actions, fine. I just don't see the usefullness of such a figure.

Consider this:
Player A has a rather conservative textbook-like style, plays as few conventions as possible and never psyches.
B likes to experiment with all kind of crazzy gadges whether agreed or improvised, hyper-aggresive bidding, speculative leads, potentially costly falsecarding and psyches.

B provokes her partners to make much more ridicoulous actions than does A. If you filter the ridicolous actions out, the comparison between A and B will be biased. This is especially true if the definition of "ridicolous" is confounded by the result an action happens to lead to.

[hotShot]

If you want to rate your declarer play, you eliminate all contracts where you are not declarer.
If you want to evaluate your leads, you focus on the deals you had to lead to.
Now what if your lead was succesful or unsuccesfull, because you followed partners dead directing bid?
Your partner bid/lead out of turn, so you are "forced" to lead or not to lead the suit shown. Should this board have impact of your lead quality statistics?

Your partner preempted 5m and all pass. In what way does this deal reflect your bridge skills?

If you want to rate your own performance, it is valid to erase all scores that are beyond your control. But you should be very carefull only to erase results that are really -beyond any doubt- unrelated to your performance. Otherwise your rating gets worthless.

[helene_t]

hotShot, on Oct 12 2006, 01:27 PM, said:

Your partner preempted 5m and all pass. In what way does this deal reflect your bridge skills?

Maybe he preempted 5m because
-you did not give him reason to expect a good result from a slower auction, or from defending. -
-you failed to open in first seat.

Or maybe his 5m preempt turned out well because you correctly judged not to raise to 6m, or it turned out badly because you failed to make a lead-directing bid before raising to 6m.

An analogy: when evaluating the efficiency of surgery in terms of reducing death rates, it seems obvious only to count deaths related to the surgery itself or to the condition for which the surgery was done. After all, how could the fact that a patient was murdered, or killed in a trafic accident, be related to the surgery?

But this would introduce all kinds of bias. For example, it could be that the fysician who establishes the cause of death is influenced by his knowledge about the kind of surgery the patient had.

Therefore, it is good clinical practice to evaluate surgery in terms of its correlation with the total death rate, regardless of the cause of death.

It's a common mistake to try to "improve" the data by post-hoc removal of data that would seem "irelevant" according to "common sense".

[bid_em_up]

hotShot, on Oct 12 2006, 06:27 AM, said:

If you want to rate your declarer play, you eliminate all contracts where you are not declarer.
If you want to evaluate your leads, you focus on the deals you had to lead to.
Now what if your lead was succesful or unsuccesfull, because you followed partners dead directing bid?
Your partner bid/lead out of turn, so you are "forced" to lead or not to lead the suit shown. Should this board have impact of your lead quality statistics?

Your partner preempted 5m and all pass. In what way does this deal reflect your bridge skills?

If you want to rate your own performance, it is valid to erase all scores that are beyond your control. But you should be very carefull only to erase results that are really -beyond any doubt- unrelated to your performance. Otherwise your rating gets worthless.

Then should you also throw out any boards that your opponents made idiotic bids and went for big numbers? After all, you had nothing to do with it, right?

How about if you bid a crazy slam that happens to make (lets say you were off two cashing tricks, but the opponents failed to cash)? Should you be rewarded for their bad defense? After all, it wasnt your "skill" that had anything to do with it making. (Although, you could be aware that opps are poor defenders and decide to bid the slam anyway, on the chance that their poor defense will let it make. Now is it skill?).

Continue down this path long enough, and the next thing you know, you have no more hands left to evaluate.

Partner taking a flyer on one board and opening 5m really shouldnt hurt an overall rating over the long haul. If partner continues to take flyers, and you continue to play with that partner, well then, why shouldn't your rating go down as well? After all, he's found someone that will tolerate his nonsense, thereby encouraging him to continue to do it. If you only put up with it for a couple of boards, then your score would not be seriously affected.

The bottom line is that any "rating" system, should include all results. Who would decide which results were due to your "brilliance", or the opponents or partners? And in the end, over the course of hundreds (even thousands) of boards, the crazy results and the brilliant results will offset each other. The objective is to have fewer crazy results, and more of the brilliant kind, which would increase your rating. The higher caliber a persons level of play is, then you would expect this to have more brilliant/good results, and fewer bad ones (unless of course, its a bunch of the juniors fooling around @ nite!!).

[hotShot]

bid_em_up, on Oct 12 2006, 03:36 PM, said:

Then should you also throw out any boards that your opponents made idiotic bids and went for big numbers? After all, you had nothing to do with it, right?

Right! If you are talking about unforced errors. If you want to analyse your own individual skill level, you should look at the best defence possible. If opps play worse than that, good for your result, but opps gifts are not part of your skills.

Mistakes and errors are the only things that matter in duplicate bridge. You win because of the imperfection of your opps bidding, play and defence. You loose because of your sides mistakes and idiotic actions. If no side makes mistakes both sides (should) get the average score.
Expert+ players have a very low rate of unforced errors and can keep their "forced error level" low too. This is why they reach the top spots.
The only way to improve is reducing the number of errors you make.

As bridge is a complex team effort it's sometimes hard to pinpoint who has to take the blame. Partners bidding, signals or play may lead us the wrong way.

Of cause if you blame your partner for every bad score and take credit for every good score, there is nothing left for you to improve.