[Cyberyeti]

One thing that would affect me if I cared.

My normal partner and I were thinking of going down to the local club and playing Fantunes (which we don't know that well as opposed to our normal system) for a bit of a laugh. This will potentially score us a much lower score than our normal system having 2 potential consequences:

1. People playing us when we're playing Fantunes will do much better than they "should" and their ratings will rise, ours will drop.

2. People playing us when we're playing our normal system will be playing against a pair much better than their rating.

The fairest way round this would be to allow people doing this to opt out of the NGS for certain sessions, but I don't think we can do this without losing the rights to magazine points and masterpoints.

[WellSpyder]

Cyberyeti, on 2012-October-29, 10:39, said:

One thing that would affect me if I cared.

My normal partner and I were thinking of going down to the local club and playing Fantunes (which we don't know that well as opposed to our normal system) for a bit of a laugh. This will potentially score us a much lower score than our normal system

C'est la vie! I had a go at playing Fantunes at the EBU Autumn Congress, knowing (like you it seems) that my partner and I didn't really know the system well enough to match what we would expect to score playing a more familiar system. Sure enough, I have dropped a grade in the NGS as a result. Never mind - it was fun, and it was also very refreshing to see that almost everyone we played also thought it was fun to play against a completely different system. No doubt my grade will go back up again in due course if it was a genuine reflection of my normal game, and won't if it wasn't.....

[gordontd]

WellSpyder, on 2012-October-29, 10:57, said:

C'est la vie! I had a go at playing Fantunes at the EBU Autumn Congress, knowing (like you it seems) that my partner and I didn't really know the system well enough to match what we would expect to score playing a more familiar system. Sure enough, I have dropped a grade in the NGS as a result. Never mind - it was fun, and it was also very refreshing to see that almost everyone we played also thought it was fun to play against a completely different system. No doubt my grade will go back up again in due course if it was a genuine reflection of my normal game, and won't if it wasn't.....

I agree with this attitude. Most of my bridge is played with random pick-up partners, and I could opt out as a host, but I decided not to when this first came in, and I've stuck with it - even getting 27% with a partner with dementia. My grading has jumped around a lot, but if I do get an unexpectedly good result with a low-graded partner, I get the full benefit of it.

[Cyberyeti]

gordontd, on 2012-October-29, 11:07, said:

I agree with this attitude. Most of my bridge is played with random pick-up partners, and I could opt out as a host, but I decided not to when this first came in, and I've stuck with it - even getting 27% with a partner with dementia. My grading has jumped around a lot, but if I do get an unexpectedly good result with a low-graded partner, I get the full benefit of it.

Perhaps I wasn't clear enough, I don't care about my own grade (or indeed know what it is), I was concerned about the effect of my artificially high/low grade (depending on what I happen to be playing) on the ratings of people who play me and who do care about theirs.

[Vampyr]

MickyB, on 2012-October-29, 05:43, said:

I like the idea of reducing the par score for pick-up partnerships.

gordontd, on 2012-October-29, 11:07, said:

I agree with this attitude. Most of my bridge is played with random pick-up partners, and I could opt out as a host,

It would be a simple change to allow pick-up partnerships other than the actual host to opt out. I told the developers before the scheme was implemented that this would be an improvement, and do not understand why they disagreed. Now they are considering it... LOL

[TimG]

Cyberyeti, on 2012-October-29, 10:39, said:

One thing that would affect me if I cared.

My normal partner and I were thinking of going down to the local club and playing Fantunes (which we don't know that well as opposed to our normal system) for a bit of a laugh. This will potentially score us a much lower score than our normal system having 2 potential consequences:

1. People playing us when we're playing Fantunes will do much better than they "should" and their ratings will rise, ours will drop.

2. People playing us when we're playing our normal system will be playing against a pair much better than their rating.

The fairest way round this would be to allow people doing this to opt out of the NGS for certain sessions, but I don't think we can do this without losing the rights to magazine points and masterpoints.

One session (and especially the 2 or 3 boards you play against any other pair) will have little effect on ratings in the long run. Yes, your opponents may get a small bump and you may drop a bit, but it shouldn't be significant when mixed in with your 100+ sessions over a year.

[awm]

While the non-linearity issue is most obvious with extreme cases, it may have effects elsewhere too. Here's an easy test to see whether this is happening.

Hypothesis: If we take the sum over all games where predicted score > 50% of [actual score - predicted score] then the value will be negative. The higher the threshold (i.e. if we take all games where predicted score > 55%) the more negative the value. Conversely, if we take the sum over all games where predicted score < 50% of [actual score - predicted score] then the value will be positive.

The implication would be that the real distribution of actual scores is closer to the mean than a strictly linear system would predict. The argument about predicted scores > 70% makes clear that this must be happening to some degree; the interesting question is whether the effect is enough to substantially bias the method.

The effect of such a bias would be that a player who wants to raise his rating is best off playing in the hardest events possible (such that his predicted score is quite low, but his actual mean score is better). This might seem good, but if many players pursued this strategy it would tend to dilute the strong events (which often depend on some self-selection to keep the field reasonable and make them more fun for the best players).

[Vampyr]

awm, on 2012-October-29, 22:55, said:

This might seem good, but if many players pursued this strategy it would tend to dilute the strong events (which often depend on some self-selection to keep the field reasonable and make them more fun for the best players).

Most of the strong events require pre-qualification or selection, so this should not be a problem.

[gnasher]

Vampyr, on 2012-October-29, 18:23, said:

It would be a simple change to allow pick-up partnerships other than the actual host to opt out. I told the developers before the scheme was implemented that this would be an improvement, and do not understand why they disagreed. Now they are considering it... LOL

I don't think they are considering that, or they haven't said they are anyway. Mike said that they could reduce the expected score for an unfamilair partnership, but I can't see any suggestion that they'd let such a partnership opt out entirely. He also said that they're not going to do it in the foreseeable future.

Edit: FWIW, I don't think one should adjust for irregular partnerships. Forming a good partnership is a bridge skill. It seems wrong to penalise people for success in this aspect of the game.

This post has been edited by gnasher: 2012-October-30, 02:58

[gordontd]

Vampyr, in post 165, you've taken part of my earlier post, edited it, and put it under someone else's post to which it had no relation, then drawn a conclusion with which I don't agree. Please don't do this.

[Vampyr]

gordontd, on 2012-October-30, 03:09, said:

Vampyr, in post 165, you've taken part of my earlier post, edited it, and put it under someone else's post to which it had no relation, then drawn a conclusion with which I don't agree. Please don't do this.

I didn't draw a conclusion; I was just quoting things that were relevant to my own comment for the sake of context. But I see now that your "I agree with this attitude" appeared very inappropriately under another post. I was careless and did not intend to mislead. I am sorry.

Anyway when I said I thought they were considering allowing people to opt out, I had misremembered an earlier post. I am disappointed, I guess because I don't entirely agree with gnasher that forming a good partnership is a bridge skill. If it is, then being adaptable and able to play with people of all different temperaments and levels of skill is one too, so perhaps they are both being rewarded and it is not a big deal.

Still I feel that as people play with widely varying combinations of regular, semi-regular and casual/pick-up partnerships, one metric does not really measure all there is to know about comparable skill levels. On the other hand, the rating is a measure of performance, not skill, so maybe it is OK. The thing that I worry about, as I have said before, is for casual partnerships to be made less attractive, because that would be damaging both to bridge culture and to attendance numbers. There is no way of knowing whether this is the case, but in any case this is why I am interested in changes that might mitigate the effect. Which may not exist.

[TimG]

gnasher, on 2012-October-30, 02:49, said:

Edit: FWIW, I don't think one should adjust for irregular partnerships. Forming a good partnership is a bridge skill. It seems wrong to penalise people for success in this aspect of the game.

I don't think a true and accurate rating system is supposed to penalize or reward anyone. It's supposed to simply be a decent predictor of results.

If you are good at forming a partnership and have done so, this will be reflected in your results and your rating.