EBU National Grading Scheme

[RMB1]

I'm not even on there as I rarely play pairs.

 

Ditto: I rarely play at all.

 

You can still find your own grade by logging in to the member area.

 

If anyone has looked, I am not from WOR (Worcestershire).

[Zelandakh]

I think we should all congratulate Frances and Jeffrey for being the third best pair in the whole of England! One wonders if these ratings might one day be used as a basis for international selections...

 

One thing I did notice flicking through this is that some top players are completely absent. Is there some reason for this? Are you allowed to opt out? To mgoetze: do you have a link to the German version, please?

[WellSpyder]

One thing I did notice flicking through this is that some top players are completely absent. Is there some reason for this? Are you allowed to opt out?

Yes, you can opt out if you want to - a grade will still be calculated for you in order to provide data for other calculations, but the grade will not be published.

 

I suspect that a more likely reason for omissions, however, is that some top players simply aren't playing enough pairs games to earn a grade.

[daveharty]

One thing that I found very interesting was that on the "Top Partnership" list, five of the top ten partnerships (and all of the top three) are mixed pairs. If this sort of list were generated in ACBL-land, I suspect this would not be the case. I know nothing about the EBU, could anyone explain why this is? Are there just a lot more mixed pairs events?

[gnasher]

One thing that I found very interesting was that on the "Top Partnership" list, five of the top ten partnerships (and all of the top three) are mixed pairs. If this sort of list were generated in ACBL-land, I suspect this would not be the case. I know nothing about the EBU, could anyone explain why this is? Are there just a lot more mixed pairs events?

Of that list of the top ten pairs by grade, only two pairs play in the EBU Premier League, which contains almost all of the best players in the country.

 

I think it's just that they don't have enough data for most of the top players playing in serious partnerships. Many top players play in their serious partnerships only in the major teams events; in pairs events they will often play with sponsors or friends, or stay at home. And most of them don't play any club bridge together.

[mgoetze]

To mgoetze: do you have a link to the German version, please?

 

http://vu2109-rails.delta.railshoster.de/frame

[Finch]

One thing that I found very interesting was that on the "Top Partnership" list, five of the top ten partnerships (and all of the top three) are mixed pairs. If this sort of list were generated in ACBL-land, I suspect this would not be the case. I know nothing about the EBU, could anyone explain why this is? Are there just a lot more mixed pairs events?

 

 

No, there's just not enough data. For example, Jeffrey and I (one of your top three pairs) only have about 380 boards together counted (which is basically 2x Brighton Swiss Pairs + 2x National Pairs), which is a long way short of the 1000 that gives you a 'proper' grade. If you 'hide' evolving grades, so you only include pairs with 1000 boards of matchpoints, there is exactly one player who is a regular in the premier league (and he's playing with wife, not his usual partner).

 

Because you are looking mainly at club matchpoint results (+ some national pairs events) you will get a lot of couples rather than the country's top partnerships.

[gordontd]

It was a Swiss Pairs with IMP scoring, so I suspect that's what confused the scoring system. I did check the original XML file, and it does show your 103 VPs, so I've forwarded it to the EBU and deleted the session from the club's record. Hopefully it'll all get corrected soon.

I was correct that the unusual form of scoring wasn't recognised, but they are correcting it and we can expect to see Andy in his rightful place once they have.

[awm]

This seems to suffer from three of the major problems that the "power rating" system has as well:

 

(1) Presumption of linearity in expected result. It's assumed that if you play in a field that is different from the norm, your expected percentage will change by the same amount regardless of what your starting percentage was. I don't think this is true. For example, suppose Frances has a rating of 68% in a national "average" field and shows up to a weak club game with a partner of comparable skill. The club game might be 10% below national average, and Frances and partner are very likely to win... but will they score 78%? Even in a weak field, it is very hard to consistently score at that level. Similarly, if you take two very weak pairs and put them in a national championship event, their final scores will be quite bad. But there will be much more luck as to who has the better score, than there would be for them in an event where the standard is weaker. My point is that when a pair is much better (or much worse) than the field their expected MP score should tail off. I don't think expected scores in excess of 75% are really reasonable or accurate regardless of the caliber of field (same could be said of scores below 25%).

 

(2) Presumption of linearity in partnership caliber. Carrying a very weak partner to a good result in a mediocre event is really a very different skill from getting a good result with a comparable partner in a top event. I know a lot of people who are much better at one of these skills than the other. It doesn't seem reasonable to presume that two strong players who obtain comparable results when playing with their regular partners will necessarily do comparably well when partnering a beginner in the pro-am, yet the rating system seems to presume precisely this.

 

(3) Partnership. Frances is likely to do a lot better playing with a regular partner of comparable caliber, than playing with a random pickup of comparable caliber (this is the same for everyone, I am just using her as an example). So her rating benefits from playing mostly with regular partners. There are ways to adjust for this (for example, weighting the effect on rating based on number of prior boards with this partner, so that boards in a "first time partnership" potentially count less than those in a long-term pair, which also has the nice effect of encouraging people to play in "one off" partnerships without worrying much about how it will effect their rating). Just looking at the discussion on this thread, while I don't know how to measure the relative caliber of Gnasher and Frances (who I'm sure are both fine players), it sounds like Gnasher plays a lot more "one off" partnerships than Frances does which would explain somewhat why his rating is so much lower.

[dcrc2]

I think it will take forever. For clubs with few points of contact with the wider EBU, the occasional point of contact will cause wild fluctuations one way or another. Suppose a top pair pop into the a little club somewhere in Cumbria, whose members rarely, if ever, play outside that club. If the top pair have an off night and score poorly, this little club will be deemed a very strong field and the members' ratings will be inflated in perpetuity. Obviously the opposite effect is also possible.

I think you are wrong about the "wild fluctuations". Indeed there's an entire section of the document on "diffusion" which explains that the problem is pretty much the opposite: an isolated club with occasional visitors will have an average rating very close to 50%, and it will take a long time for this to shift. I've just worked through the maths myself and assuming I understand the document correctly -

 

Say that the visitors do about 10% worse than expected, and that the club has 5 tables; then each of the regular club members' session grades is about 1% higher than normal. Then the following week, assuming that the same regulars turn out at the club, they will find everyone's rating has increased by just under 0.05% (the most recent session is weighted by about 1 in 20), and the average session score that week will therefore be 0.05% more than before the visitors arrived. The week after that, the average rating has actually increased a little more (the weighting of the week the visitors played has decreased, but this is offset by the small increase in strength last week), but only very slightly. Following the calculations through, I'm finding that in the long run the average rating converges to about 0.07% more than before the visitors arrived.

 

So, yes, the regular members' ratings have increased permanently, but only by a tiny amount. Indeed, the change in average rating will be (almost) proportional to the total number of matchpoints above average won by all visitors to the club. The more visitors you have, the more the club's rating changes. A single visiting pair has almost no effect.

[Vampyr]

In a Swiss Teams event, Butlers (or cross-IMPs) simply won't produce the right answers, because the results are strongly dependent on the strength of the team you are playing.

 

I suppose that in a Swiss Teams, they would note who your opponents are and then cross-IMP the results. This would presumable be possible where Bridgemate II was available; I don't know how it would be managed if there was only Bridgemate I. For privately-played matches, I doubt anyone would go to the trouble of sending in the necessary information.

 

In an event with seating rights, Butler or cross-IMP scores are also of questionable value, because the stronger pairs tend to pick each other. And it creates a perverse incentive to pick the weaker opposing pair so as to inflate your grade.

 

Do you think people would be that bothered, though? And in fact, they might prefer to deflate their grade, because...

 

One wonders if these ratings might one day be used as a basis for international selections...

 

My understanding is that these ratings are primarily for use in handicapping.

[gnasher]

Do you think people would be that bothered, though?

Probably not, because the only events which have both duplicated boards and seating rights are things like the Spring Fours, the Crockfords final, and the final of the Brighton teams, where there are strong incentives to do well as a team.

[Finch]

 

In a Swiss Teams event, Butlers (or cross-IMPs) simply won't produce the right answers, because the results are strongly dependent on the strength of the team you are playing.

 

Isn't this true of Swiss Pairs events as well? The NGS data already includes matchpointed swiss pairs (e.g. about 1/7 of my boards are from the Brighton swiss pairs when, in theory, we were playing the strongest pairs in the field for most of the event).

[gnasher]

Isn't this true of Swiss Pairs events as well? The NGS data already includes matchpointed swiss pairs (e.g. about 1/7 of my boards are from the Brighton swiss pairs when, in theory, we were playing the strongest pairs in the field for most of the event).

I was wrong about this. For Swiss pairs events they adjust for the strength of the people you play:

 

Things are different for a Swiss Pairs Movement. Here we take each match as a separate stanza within the event as a whole. For each match, we are playing against a single pair of opponents. Here, as in Bridge Club Live, it is by far best to take the grading values of just your two opponents. The SOpp factor for one match is just the average current grade of your two opponents.

Your SOpp for the event as a whole is the average current grade of your opponents in each of your matches.

("SOpp" = "Strength of opponents")

 

As Stefanie says, they could do the same for Swiss Teams. They'd have to know which half of the opposing team you were playing.

[hatchett]

I agree with much of AWMs post. They are many flaws in this rating system. It seems to me a very good player playing

with a very weak player may have an advantage. Imagine a 70% player playing with a 30% player in a 50% field. The 70% player hogs the bidding

so he gets to play more than his fair share of hands which when he plays will often lead to a good score (I know this example is oversimplified

because there might be issues bidding the right contract but the point is there). If he plays with another 70% player with the same declarership skill, this tactic is not available. Effectively the partnership is playing as a 52/53% partnership but is rated 50% by the system.

[Oof Arted]

Seems to be just another waste of time and money :blink:

[SimonFa]

I've been playing for about 18 months and play at 3 clubs with a number of partners and am quite pleased to see my grade is a a shade over 50%.

 

Interesting, after about 100 GIB MP tourney's my average is also a shade over 50%.

[gnasher]

Seems to be just another waste of time and money

That remains to be seen. If it gives people another reason to play more face-to-face bridge, that's a good thing. If it encourages them to play better, that's a good thing too. Those benefits may not materialise, or they may turn out to be insufficient to justify the cost, but I think it's premature to assume this.

 

There have been several posts pointing out flaws in the system. I don't think it matters that it won't work perfectly - for differentiating good players from bad, it's a vast improvement on masterpoints. For players near the top, we can continue to assess them by asking the question "What has he won?" (This would, by the way, give a quick, easy and correct answer to the question mentioned in awm's final paragraph.)

 

I can think of one diasdvantage of this scheme: some people may not like being told that they're not very good. However, I doubt if that would cause many people to stop playing: either they'll see it as a reason to try to improve, or they'll dismiss the grading system as being inaccurate, irrelevant, and not nearly as meaningful as masterpoints.

[mr1303]

I come in at 62. something % (but strangely only if I search for myself, rather than search down the list of top players). This is strange, since apart from my last two games which I've played down in Cornwall after I moved here, 99% of all my games in the last 3 years have been with the same partner in a rather low ability club near Bristol. And when I search for her ranking, I'm an ace of clubs, whereas she's an 8.

 

I think that it needs a little work before it becomes particularly meaningful, but it's a nice idea.

[IanPayn]

I come in at 62. something % (but strangely only if I search for myself, rather than search down the list of top players). This is strange, since apart from my last two games which I've played down in Cornwall after I moved here, 99% of all my games in the last 3 years have been with the same partner in a rather low ability club near Bristol. And when I search for her ranking, I'm an ace of clubs, whereas she's an 8.

 

I think that it needs a little work before it becomes particularly meaningful, but it's a nice idea.

 

++++I agree that it might need a bit of work, but despite my absolute lack of interest in any form of ranking on a personal level, I can see that this is an improvement on "Master Points". It's the difference between a Form Guide and the Dictionary of National Biography. If anyone cares, I'm chairman of the EBU's Tournament Committee, and so I sort-of have a vested interest in this working. But that's wearing my business hat: wearing my other hat, as a part-time member of the human race, I think this is A Good Thing and already stimulating interest. Why, only today someone talking to me declared himself a Queen.

 

It took a moment...

[Vampyr]

in a rather low ability club near Bristol.

 

I imagine that clubs (along with all subsets of the population) were all assigned a 50% average, and that that hasn't moved much.

[blackshoe]

Sure the tactic is available. It's just not nearly as effective. B-)

[mchristie]

I'm Mike, the team leader for the NGS and have just found this thread. Interesting comments. I'll try to give some answers.

1) Andy Bowles now restored to grade of 60+.

2) Grade accuracy standard error of at best 2%, but you need a good mix of partners, and see below for caveats.

3) Biggest limitation: Ann usually plays with Bob and they score 50% (v average opps), but occasionally Bob plays with Dave and they score 45%? Who's the strong player? No grading system can know. NGS will (remarkably quickly) decide that Ann is the strongest getting a grade probably near 55 (but depends on starting conditions). From our unpublished analysis of playing patterns, we think this and similar problems significantly affect the grade of between 2 and 5% of the EBU 50,000 members. (based on Yorkshire test bleow)

4) We have lots of unpublished maths behind the NGS, for example:

test of more complex estimation(Kalman filters/Least Squares estimatation) for the 4000 Yorkshire members over six months, which would give each player a personal grade accuracy as well as grade, but we rejected as it means your grade would change when you don't play, and takes very much more computation.

measures of degree of mixing between clubs (around 50 clubs are sufficiently isolated that it will take over 5 years for their grades to be nationally standardised.)

mathematical behaviour of the discrete filter that the NGS system uses.

etc, etc.

5) NGS does use IMP scored pairs events not just MP.

6) Want to use pair-wise scores of teams events sometime in the future, but we need 100% reliable data on which players were which pair for each match of an event.

7) National/Regional events are weighted threefold. Thsi is primarily to assist in standardising grade averages between clubs ("diffusion"), but has the side effect that for strong tournament players, their grade is more dependent on these events than local club duplicate sessions. So I hope that in a couple of years, many strong players will get a more accurate grade, but I agree it would help if we could use at least some team events for such players. Some professional players suffer from the Dave effect (above) more than others.

8) Partnership grades are more meaningful, but players want individual grades. We're trying to improve the way P'ship grades can be displayed and searched.

If anyone is interested, and has maths or software backgroud, we'd welcome some volunteer assistance with improving and monitoring the system. Contact ngsqueries at ebu.co.uk

[WellSpyder]

5) NGS does use IMP scored pairs events not just MP.

Thanks, Mike, for posting this interesting reply. The above point 5 is the only one on which I still have some doubts. I know that the documentation for the system is clear about doing this, and about how it is done, but I notice from the indivudual session records for what I have played in recently that none of the imp-ed pairs events have any grading information alongside them - and there are actually quite a lot of these on my record since the county I play in holds an imp pairs event every month.

[gordontd]

Thanks, Mike, for posting this interesting reply. The above point 5 is the only one on which I still have some doubts. I know that the documentation for the system is clear about doing this, and about how it is done, but I notice from the indivudual session records for what I have played in recently that none of the imp-ed pairs events have any grading information alongside them - and there are actually quite a lot of these on my record since the county I play in holds an imp pairs event every month.

We have a Cross-IMP pairs at the Young Chelsea every Friday night, and those events all have grading information alongside them. The only event that caused a problem was the Swiss Pairs with IMP scoring, which has already been noted and was apparently the only one of its kind so far submitted (though we are having another on April 1st).