Big problem with 5.0+ rule

[sam_p]

Interesting hypothesis. To there are actually an increased number of players (a cluster) at the boundaries, I took a look at the count of players at each rating down to the tenth that I have. See blog post here with the chart and details.

Now, perhaps you weren't necessarily saying the clusters around the boundaries were larger, but from my data at least, that doesn't seem to be the case.

It actually seems according to your data to be quite a nice normal bell shaped distribution, which in and of itself is kind of interesting. Clearly this corresponds to the common perception that as you get higher and higher in rating there are less and less players at the same rating around, but I would have guessed that there would be a bit of a plateau at the lower ends since there should be many players who are at lower levels and stay there - for any number of reasons. It seems though that perhaps people who stay at really low levels just stop playing USTA ultimately so they drop out of your data (based on people who are playing in leagues right?).

 

[schmke]

It actually seems according to your data to be quite a nice normal bell shaped distribution, which in and of itself is kind of interesting. Clearly this corresponds to the common perception that as you get higher and higher in rating there are less and less players at the same rating around, but I would have guessed that there would be a bit of a plateau at the lower ends since there should be many players who are at lower levels and stay there - for any number of reasons. It seems though that perhaps people who stay at really low levels just stop playing USTA ultimately so they drop out of your data (based on people who are playing in leagues right?).

I did look only at players with ratings calculated this year, so I am missing anyone that would have played in the past but quit playing this year or prior. I'll try to look at that later.

 

[anubis]

It may be different for other players, but for me, the USTA league tennis is where I consistently find the best competitive play without having travel or take time off work.

I agree, I use the league play for more than half of my competitive matches as well. I'm just trying to find some options for the OP

 

[rod99]

I agree, I use the league play for more than half of my competitive matches as well. I'm just trying to find some options for the OP

tournaments aren't that big around here. they have them periodically but draws aren't very big and entry fees aren't cheap. most people play leagues.

there are some age group tournaments but i couldn't play the 35's until next year. again, draws aren't that big.

i did play 9.5 combo and 9.0 mixed (with a 4.0) for usta this year. i have plenty of guys i know that i play competitive matches with, they just aren't in usta leagues.

 

[asimple]

Interesting hypothesis. To there are actually an increased number of players (a cluster) at the boundaries, I took a look at the count of players at each rating down to the tenth that I have. See blog post here with the chart and details.

Now, perhaps you weren't necessarily saying the clusters around the boundaries were larger, but from my data at least, that doesn't seem to be the case.

I think the issue is that the formula which creates the ratings will create a normal distribution by design (or CLT if not by design). I'm not sure, but I think the proper way to determine if this is true would be to check the accuracy of match outcomes based on dynamic NTRP. I think you would see more accurate results around the 4.0 level and 4.5 level than at the intermediate points. I would need to think more about this though.

I definitely see two distinct levels in 4.5 although this could be biased.