Hi,

I was recently speaking with a fellow p_layer_ who made a comment that raised an eyebrow. He claims that todays rated p_layer_s are not of quite the same caliber as those equivalent ratings 10-20 years ago. He seemed to imply there are many more 2000+ p_layer_s today than ever before (quite possible) and that it is easier to attain a rating at or about that level now than it was 'back in the day' so to speak. His argument didn't limit itself just to 2000+ p_layer_s either. Simply put...a 1750 p_layer_ from the past would (generally) be better than a 1750 p_layer_ in the present.

I agree there may be more people who play our sport due to a rise in popularity (which would produce more high level p_layer_s just due to the numbers), but is there any merit to his claim? Or are these the senile ramblings of a TT p_layer_ living in the past?

Have there been any significant changes to the mathematical or statistical process done to determine ratings in recent years? If so, when were these changes implemented, why were they implemented, and what was changed exactly?

Any thoughts appreciated!:_b_link__:

Post edited by: Flying Poo, at: 2005/10/10 11:18

Great question!

There has not been a significant change to the ratings formula that would introduce 'ratings inflation', which is the shorthand term for the situation you describe.

However, the USATT rating system is not, completely, a 'zero-sum' equation, and over time there is a tendency for ratings to creep upward. I'll describe one example of this in a moment.

In Canada, the ratings system is deliberately inflationary, and points won and lost in a match are not neccessarily equal. The Canadian thinking is that this encourages p_layer_s to stay tournament active with no fear of loss of prestige if they lose a few matches.

In the US, adjustments can happen. Say that a p_layer_ arted 1700 has a good day, and beats one 2000 p_layer_, loses to 2 other 2000 p_layer_s. In his two losses, no points are exchanged because the rating differential is so steep. His 2000 level win, however, results in a 50 rating point loss to his opponent and a 50 rating point gain for the victor. Very straightforward.

But...

If that 1700 p_layer_ beats all three 2000 level p_layer_s, its assumed he is not really 1700. His rating is adjusted upward to nearly 2000 and the results reprocessed. Now his opponents all lose 8 points apiece and he gains 24. This is done to protect higher rated p_layer_s from being beaten by extremely under-rated opponents. But the net result, over many years, is for the total number of points in the system to increase, and a slight inflationary increase.

So in my opinion, you elder friend is correct. 2000 is not quite what it was 20 years ago. As a relative measure, between currently active p_layer_s, its very accurate. As a means of comparing 1980 to 2005, its not entirely reliable.

Alan

I agree with Alan, the focus on absolute numbers is always a bit misleading. One way to perhaps answer your question would be to find out where the median (that is, half the p_layer_s have a rating below and half have a rating above this "mmagic number") of p_layer_s are now as compared to, say, 10 years ago. The median today is about 1620 or so. I don't know what that number was 10 years ago.

andreas

I agree with everything in this post. However the one factor that I don't know the impact of is the case that as p_layer_s ability starts to decline, they tend to drop out of competition. If the pool of p_layer_s remained constant, then those p_layer_s who were getting worse would be losing points to those p_layer_s who were getting better. Instead rather than lose points, many p_layer_s would rather drop out of competition. It seems to me that that effect would tend to offset the effect that you mentioned, and reduce or even eliminate the "rating inflation" that you mentioned.

Any thoughts?

I agree with everything in this post. However the one factor that I don't know the impact of is the case that as p_layer_s ability starts to decline, they tend to drop out of competition. If the pool of p_layer_s remained constant, then those p_layer_s who were getting worse would be losing points to those p_layer_s who were getting better. Instead rather than lose points, many p_layer_s would rather drop out of competition. It seems to me that that effect would tend to offset the effect that you mentioned, and reduce or even eliminate the "rating inflation" that you mentioned.

Any thoughts?

You are correct that a p_layer_ who stops tournament play 'takes his points off the table' and reduces the pool of total points available. However, USATT numbers show that tournament play has increased since 2002, rather dramatically! There are *more* points in the pool now than previously.

Don't underestimate the effect of those 'adjusted' p_layer_s, either! Having your rating adjusted from 1300 to 1800 _base_d on evidence of new playing level is introducing 500 rating points at no one's expense.

In general, the US System is a 'zero-sum' system (points gained equal points lost) but in practice it has a slight inflationary tendency.

Alan