[helene_t]

At the Dutch site StepBridge, you get compensation for playing with a weak partner and/or against strong opponents. In theory, it should solve the problem with strong players refusing to play with weak players, or strong players earning cheap IMPs against weak players. But how to calculate that compensation? You can probably immagine that the threads on that issue have attracted 1000+ postings.

I think that if some tournament director wants to implement his own idea of a rating scheme for his regular participants, he can have his way. If it becomes popular, other TDs will adopt the same system. But please don't impose some standard rating scheme on all tournaments. We will never be able to agree on how such a rating scheme should work, and most will probably be better off without ratings.

[Free]

When stepbridge was free, I also played there, and I remember that rating system! Unless it changed, it sucks (sry to say). This is how it works if I remember correctly:

- there are 9 level: 1 is absolute beginner, 9 is the best
- after every 30 hands or so, you ranking is recalculated, based on these hands. You can go up or down 1 level, according to the scores SB gives you.
- when you play against other pairs, the levels of both pairs are summed up, and their relation gives a value to some parameter to calculate your scores. This adjusts the %-score for the experts and for the beginners, since the beginners have a very big change to get 0% on hands this way. Now they'll get at least some %. Experts make tops a lot more against beginners, so they won't get the full 100%.

As you can see, this system gets quite discriminating, since good players won't play against beginners anymore, because they'll lose their great level-9 status. If they happen to get to play new members which play very good, they get negative %'s sometimes!! The sollution a friend of mine found was to use several logins: he had 1 with level 9, 1 with level 6-7, 1 with level 4-5 and 1 with level 1-2, and keep these logins at that level (playing bad was obligated sometimes ). This way he could almost always play somewhere he wanted. When I started I couldn't play a decent game, since playing against a level-1 loses too much points! Most tables required level 4-6.

Also, beginners will stay beginners, since they won't learn much playing against other beginners. Giving players a ranking is pure psychological, and gives a negative factor! Some people will do everything to get the highest ranking, others will do everything to keep their ranking,... Imo, a non-hidden ranking system would give a very negative factor on this community. And I don't know what you could accomplish with a hidden ranking (players don't know they have a rank).

[guggie]

So when you want to improve yr rating you choose a p who can play but is a little low in rating by accident , and choose opponents who cannot play but have a high rating.
It is boring bridge, like taking candy from a baby.
I DETEST RATING

Please please lets continue to play for love of the game

(I also hate writing in capitals colours bold and with exclamation marks, but i DO feel strongly about this. I felt so happy discovering a site without rating and now it is starting on BBO too

[Chamaco]

My personal opininion is that the benefits of ading a rating system would be minimal.

1) right now, after a while you play on BBO, you get to know which players you like and which not.

2) yes, sometimes there is no known partner available, so you have to "guess" or their self-statement. Guessing on a BBO point rating would be more reliable, but by how much ?
I don't know how well the rating systems of the foreign federations work, but speaking of the ratings in Italy, I can tell you they are absolutely meaningless.
You will find MANY MANY players with a low ranking which are actualy MUCH better than other players with SIGNIFICANTLY higher rank.
And this is not an exception, but an extremely common occurrence.

3) on the other hand, my experience in any field that adopts rating systems is that people become les nice. One does not need to be a genius to recognize this in bridge clubs and bridge competitions, and my long experience as a ches player confirms that it holds true also for chess (and other posters will probably have experienced the same in many other fields).

4) one of the effects of rating introduction is also that it becomes harder and harder for bginners to play with and against good players. More and more tables become composed of beginners/low level intermediates only.
This is a damage even for the selfish people who want to play only between experts: remember, the bridge world is alive because there are a large number of players, the more players we have, the better it is for everybody.
And the best way to do it, is to let these beginers improve by keeping them in contact with stronger players.


So why should we introduce a rating system for a minimal benefit (knowing "better" the strength of your pard, based anyway on unreliable criteria), for a probable cost much worse (losing much of the frienly atmosphere and segregating low level players to a level where they can much less frequently interact and improve) ?

My bottomline is: let's leave the hierarchies to the bridge federations, and keep BBO as it is.
We do not need yet another means to pump up our ego (which the rating essentially is).

[helene_t]

guggie: I don't think you need to fear that rating is being introduced here, most people seem to be against it. Btw, you probably remember last christmas, when we played for nuts and pieces of chocolate, and nobody knew what the exchange rate between nuts and chocolate was, and besides most of the wine and cheese was given to the winners of the lottery rather than the winners of the tournament, and it was robber bridge which few people can take serious anyway. It was one of the most enjoyable bridge tournaments I've ever participated in. So the least evil rating system might be one which can't be taken serious by anyone.

Free: 90% of the members think that the Dutch system sucks. Of those, 90% only think so because they can't figure out what consequences it has and doesn't have, which is, indeed, a complicated issue. I could tell you that since I'm a mathematician, I belong to those 10% who actually know why it sucks. However, the probability that I'm telling you the truth is no more than 10%, in that case. Actually, I have no idea if it sucks more than any other rating system would.

[Free]

It's quite obvious what the consequences are imo: good player's rankings are too low, bad player's rankings are too high. You get a curve which is very high in the middle, and low on both extremes. Also beginners can become experts by bashing even bigger beginners...

But the big issue is what happens with the players, how they play, against and with who,... and this had a lot of negative influence! That negative influence is just a proof the rating system sucks imo

[paulhar]

OK, looks like I was just plain WRONG! Apparently my experiences with a rated site were most unusual - in the poll, only one other vote for equal manners between rated and unrated sites/tables and that was 'everybody is rude!'. About 90% of you stated that your experience showed that rated tables increased boorishness and/or unethical behavior. (The undisclosed agreements argument was something I hadn't thought of.) And if that's really true, I don't want ratings either.

By the way, I feel like I understand the Step rating system also, and agree with Helene's assessment of its merits I did a simulation once and wondered what would happen using the Step system without a top level of 9. After many hands, everybody's ratings went up, and eventually even the most pathetic players would levatate from their '1' rating, being pulled up by the massive ratings earned by others. The skew came from the fact that 3% above average raised your rating but 5% below lowered it, so there's an automatic upward bias for attendance. The same was true for IMPs but I don't remember what the exact IMP totals needed for advancement/demotion were.

[DrTodd13]

Just a couple of requirements on any ratings system that would be developed. First off, somebody should not be given a rating unless the number of their partners and opponents is sufficiently connected to the larger BBO membership. For example, if Meckwell played on BBO but they only played against some other world class pair but not as good as them then if analyzed in isolation then Meckwell would be above average and the other world class pair would be below average. So, there has to be a certain amount of intermixing of partners and opponents for a rating system to have any meaning. Does BBO currently have enough such intermixing for the average player? I don't know but I do know there probably are people who play set games with few enough partner's and opponents that a rating for them wouldn't make any sense.

If a ratings system were to be any good then it has to be predictive. One of the big failings of the lehman system I think is that it tried to compute how well a 54/61 pair would do against a 47/52 pair. Moreover, the lehman system treated a 55/55 pair the same as a 50/60 pair which is obviously not try. A pair (except for declarer play) is limited in general by its weakest member. Rather than trying to create another formula (which is pure guess work), I think the right approach is to have a 4-dimensional matrix where each dimension represents the ability of one of the people at the table. This matrix is initialized at week 1 with some guess as to what the result should be but each entry in the matrix is itself a weighted moving average so after each board (or possibly once a week/day) that entry itself is updated to reflect the actual score seen. So, if a 50/60-55/55 matrix entry is initially 0.0 and the 55/55 pair win 5 IMPs on a board then the 55/55 pair will move up slightly to 55.2/55.2 the 50/60 pair move down slightly to 49.8/59.8 and the matrix entry moves up slightly to 0.2. Then we'd have to deal with issues like fractional ratings and how to look those up in the matrix. Moreover, they may be specific ratings combinations that are rarely seen in actual play and so those matrix entries would be updated very slowly. Holding 3 dimensions constant and varying the 4th dimension, the result should be at least be smooth and not-chaotic so to some degree you could change adjacent matrix entries to maintain this property.

[hrothgar]

couple quick comments here
(my Internet access is very limited these days)

Back when I was trying to develop a ratings structure at OKB, I reached a couple significant conclusions:

#1. Start with the easy stuff.
Work on developing a rating system that can accurately rate PAIRS.
If/When you can evaluate the performance of a pair, you can try to
decompose this into rankings for individual players.

#2. Use signal processing models.
Most of the work on rating systems was based on statistical methods such
as regression analysis. from what I can tell, signal processing is a much
more fruitful avenue for exploration. Discrete Time Kalman Filters look
particularly appropriate.

[luis]

Since this topic is active I will re-insert my idea about rankings here:
Bridge is a pairs game, so instead of having a ranking for each player keep a ranking for each pair.
An individual will then have different rankings with every different partner.
In this way you can see how good your opps are as a pair or how good or bad you are doing with some partner.

Doesn't it make more sense?

Luis

[Free]

Yes it makes sense, but it still has the same consequences: some people won't play with others, and some will cheat to get a better ranking. Always the same story, so no ratings in any form plz!!!

[hotShot]

Well if we start a discussion about ranking systems lets start with the problems:

First lets define some rules that must be applyed to a rating system, to make it work!
1) Winning against a weaker rated Pair/Team/Player may not influence the rating , that way playing weaker opps will not help you to improve your ranking, nor will it ruin theirs.
2) Winning against a better rated Pair/Team/Player should increase your rating
3) Loosing agains a weaker Pair/Team/Player should decrease your rating

Now lets think about what d data do we take into account, and what the rating system should tell us.
At a team event the rating system should tell us which team will me more likely to win.
This is the most simple case, as to teams meat and one ist winner and one is looser.

At a pairs event the system should predict the final ranking of the tournament.
Here we have a problem. Let us assume we have a perfect “7 Table Howell”, we play 13 rounds. We have 2 top pairs, that make 100% against each opponent and get 50% if playing against each other. The other pairs are of exactly the same strength and play 50% against each other loosing only 2 the top pairs. What would we expect ?
2 Pairs at the top off the list with almost 96% and all other pairs should be equil at about 46%.
But we won’t get that scoring in Matchpoints. We will have the weaker pairs distributed from 46% down to 41%.
This range is created by the movement we used. At some boards both top pairs score NS or EW so the avarage score for the “normal” pairs is different depending on the fact that they are on the same side as the top pairs or on the other side.
In any other movement than the “7 Table Howell” it is even worse.
This is why typical BBO 8 Boards/ 4 Round/ lots of pair tournaments should be compared to lotteries and not to tournaments.

This is why there should be special ranking pair events, with a small number of players and a big number of boards. Subbing should not be allowed ( because who’s strength do you take for the final result that of the sub or that of the player he subbed in for, and who’s ranking is adjusted?) All results of pairs that are eliminated should be erased from the tourney.

Considered all that there are still 2 Problems to solve:

1) How do we find a rating for a new player if everybody else has a valid rating (this is easy if everybody else has one)
2) How do we get the initial rating for all, if nobody has one


Have a nice day
hotShot

[Chamaco]

hotShot, on Jun 23 2004, 09:48 PM, said:

1) How do we find a rating for a new player if everybody else has a valid rating (this is easy if everybody else has one)

In chess servers with rating, one enters with a fixed rating, but the fluctuations in rating at the beginning are multiplied by a coefficient which is larger at the beginning and decreases in time.

So basuically: at the begiinning you experiene large loss and gains in rating very quickly, then after a while you start to stabilize your level while your coefficient decreases.

After a while you find yourself at the level you deserve.

[paulhar]

I saw some posts suggesting that pairs get ratings; that you should have a different rating with each partner. OK, so I have 600 ratings, almost all based on 1 to 3 boards played in an individual! (An aside: individuals are better slapstick comedy than any you would go pay to see. Keep your eyes on the chat for maximum comic effect.) Unfortunately, these ratings are pretty meaningless, as are any ratings of others that are derived from them. I would think that most people's playing profile looks a lot like that - BBO encouranges people to click into rooms and play with whomever for a few hands, and under this rating system, that would be just one more meaningless rating that will play a factor in determining other pairs' ratings. A pair probably shouldn't be rated until they have played at least 20 deals with each other, and even then problems arise because their opponents on some of those deals won't have played the requisite boards. You could 'fake' their opposing pair's rating by combining their totally hidden from all individual ratings, which would make it a lot easier for pairs to play twenty meaningful boards. Also, there comes the problem mentioned earlier that a pair would have had to play some cross-section of BBO; i.e. if pair A and pair B are all new to BBO and play 20 boards against each other, they could both be awesome pairs or both be beginners so any numerical rating assinged to them would be meaningless. (Individual ratings would suffer the same defect.)

I would not be at all surprised if over half the pairs I play against have not played 20 boards with each other - usually they just clicked into my table. I know for a fact that there are not that many people on BBO that I have played 20 boards with.
Pairs ratings work if you play a lot with the same people and can find other pairs that play together a lot, and those pairs also play against other pairs besides you. In the BBO environment, at least as I see it, there would be too few pairs with meaningful ratings to matter. But the whole point is moot - because it has been pointed out on many occasions that 'Ratings are evil!' or 'Ratings make people evil!' or perhaps it's the love of a high rating that is the root of all evil!

[Cascade]

Chamaco, on Jun 23 2004, 07:24 PM, said:

In chess servers with rating, one enters with a fixed rating, but the fluctuations in rating at the beginning are multiplied by a coefficient which is larger at the beginning and decreases in time.

So basuically: at the begiinning you experiene large loss and gains in rating very quickly, then after a while you start to stabilize your level while your coefficient decreases.

After a while you find yourself at the level you deserve.

I have experimented with this sort of thing at the local club.

I estimated the score of a player as a function of rating, partner's rating, the ratings of player's in the field and then adjusted this based on the score obtained. When I had very little data about a player I made big adjustments and when I had lots of data about a player I only made small adjustments.

This seemed to work well for me and was much more useful than some infamous online rating systems I have seen.

[helene_t]

What is the purpose of a rating system anyway?

I really don't care if my partner is good at hand evaluation and declarer play. It may influence the results, but it will not influence joy or frustrations. Of course, if I were to give partner instructive comments, or receive instructive comments from him, it would be nice if we knew each other's skills. But most people don't want comments, or don't have time for it. Besides, instructive comments are useful when given from an expert to an beginner, but rating schemes would tend to prevent experts from playing with beginners. I do have a slight preference for players who don't play advanced conventions unless they really understand those conventions. A rating would give me little clue to that except that a real expert would have a somewhat higher probability of being ok than an intermediate or advanced player would, but a real expert wouldn't waste his time playing with me anyway.

I would like the profiles to be more elaborate, though. The self-rating is not very informative.

[paulhar]

helene_t, on Jun 25 2004, 02:27 AM, said:

(portions deleted) I do have a slight preference for players who don't play advanced conventions unless they really understand those conventions. A rating would give me little clue to that except that a real expert would have a somewhat higher probability of being ok than an intermediate or advanced player would (portions deleted)

Ah, the solution is simple. People can rate themselves as 'I really don't fully understand these convetions, and I don't want anybody to comment on my use of them, but I want to play them anyway'. How many people would partner this person? Nobody would ever admit that anyway - since everybody assumes that they know the convention. And they're telling the truth - they know everything they have heard about the convention. Are they playing it right? Not likely. If I agree to play 'multi', I'm probably not playing it right either, but I'll happily admit that. Most people are just not aware that there's a lot more to a convention than they know about it. Take transfers. We don't like to teach transfers to newer players because they get the taste of it and they think they know how to play them, whereas they are quite in the dark in many special situations (like what if the opponents overcall?)

We were discussing this philosophy with a couple of students who totally agreed, but of course, they had totally discussed transfers and they knew the ins and outs of transfers and had been playing it for some time so there was no problem. I decided to test this theory, giving one of them a 1NT opening and the other one

---

and let them bid it. I threw in a 2D overcall expecting the responder to bid hearts and the other to take it as a transfer, or for the responder to double (stolen bid - arrgh!) and the other to correctly (in theory) pass, or some other disaster to befall them, but the heart hand stated 'Transfers are off over interference' negating all chance for a misunderstanding. Then she said "so I have to pass." If she weren't playing transfers, she would have competed to 2 hearts! But merely playing the convention has clouded her normal thinking.

A couple more examples pointed out that it was obvious that they needed a lot more discussion about transfers (one of the two was ready to give them up but that was not my intention.)

So, it is quite practical to assume that your partner is not going to understand some infrequent nuance of a convention and to avoid these if at all possible, unless trying to form a very regular partnership with this partner.

A real expert would have a somewhat higher probability of being OK? How on earth would you know? Judging from the response I got when proposing ratings that NOBODY ELSE CAN SEE - the probability of knowing somebody is an expert by seeing their rating on this site and having the players happy about it is approxmently the same as the probability of seeing complete peace and harmony in the Middle East in our lifetime.

[paulhar]

You've seen in my past posts what I was trying to avoid by having ratings that nobody can see. Now I have encountered a new unpleasent occurrence twice in two days that non-visual ratings would have prevented.

Because the open slots in a game go so fast, one doesn't have time to examine the profiles of all three players at the table - if you are lucky to find an empty seat opposite a partner whose profile you like, you must jump at it. So, I did. Turns out that my opponents were marked 'intermediate' which is supposed to mean 'average player on BBO' but in reality, means 'average player in an under 300 game'. We play about five hands, and one of my opponents who was down 31 IMPs said something like "I've had enough of this! I didn't come here to lose 31 IMPs!" and promptly left. Today I found a friend available when I logged on and he asked that I find a couple of opponents. I clicked on a table which had two advanced players as opponents. Again, almost like clockwork, 5 boards later, they were grumbling about 'enough of this' (like it was my fault!) and left. Now, as unpleasant as this was for us (my 'friend' didn't seem too happy about the whole affair), I'm sure it was more unpleasant for them - and they showed their displeasure, which is what made it more unpleasant for us.

We would have avoided both these pairs (and they both would have avoided us and been a lot happier) had we been able to put a minumum rating on the table. No doubt, there was probably another pair that would have loved a tough game - I just didn't happen to find them.

Lest the sceptic call me arrogant again, let me just say that I am not intending to brag. (I know it sounds that way, but how else could I have made that point?) I make many too many errors to be a top player and being rated isn't going to boost my ego - I'm all too painfully aware that I'll never be one of the top 10,000 players in the world. But I would like to avoid the players that are so outclassed that they're going to leave in a huff after a few boards. Don't tell me to only play against World Class players, most of them won't allow me at their table, presumably because they don't know me, because as far as I know, I haven't ever played with or against any of them. I actually don't mind being outclassed that miserably - I'm bound to learn a lot when I am. However, I also feel that I have no right to inflict myself on a table where they want world class play, because I'm not going to be able to deliver the goods.

[dogsbreath]

Hi Paulhar..
first of all I'd like to say you're a hell-of-a typist Not only the quantity, but the lack of spelling mistakes and generally accurate grammar will really fox 'Slothy'. I'm generally uneasy about rankings on BBO but not enough to get into a row about it. However i do suspect it would discourage stronger players from playing with beginners/intermediates. As to finding a (decent) table to play at.. you are welcome at mine anytime .. and I will be happy to introduce you to many fine players on BBO.
I have regular games with 'beginner/intermediate' players which I enjoy and regard as a greater challenge than playing with experts since I must take more care to try to ensure I get the best out of them and that they enjoy their game. I know plenty of others who do the same and wouldnt like to see a situation where a ranking system discouraged this. When a 'serious' game is planned we simply set player permissions ON - we never ask for 'expert' as most 'advanced' players here are solid players. If someone is disrupting a table - ask them to leave or boot them if they wont go quietly .
If you want to join a table or team match where they are asking for expert/WorldClass ..just edit your profile
Rgds Dog

Quote

furnulum pani nolo

[guggie]

I fully agree with Paulhar. Playing in mainbridge is very hard as p and opps come and go very very fast. I almost completely quitted it, being not able to find a table with players who play for love of the game and wanting to improve.
The self chosen skill level is 80% inadequate (also in my own case, I should be intermediate followong BBO standards but viewing other people's skills I proceeded to advanced;-. I cannot bring me to call myself an expert, every real good player who knows me should have a fit of laughter. )

Nevertheless rating is worse.

Cannot we create a corner in BBO were abuse and roaring monkey behaviour is strictly forbidden, and tables are open to those who love the game?