104th placed player points go up before GS [Archive] - MensTennisForums.com

104th placed player points go up before GS

amirbachar
03-18-2011, 11:32 AM
Hi, check out the stats here:
http://simtheworld.blogspot.com/2011/03/avergae-point-for-place-104-by-date.html

n8
03-18-2011, 11:36 AM
Interesting. Putting vertical lines when the cut off for Grand Slam entry is would add clarity.

duong
03-18-2011, 11:39 AM
yes, interesting :lol:

amirbachar
03-18-2011, 12:04 PM
Interesting. Putting vertical lines when the cut off for Grand Slam entry is would add clarity.

You are right - updated with lines...

ApproachShot
03-18-2011, 12:17 PM
Good work. Couple of questions though.

What is the periodicity of your moving average? Or did you just take the absolute value instead of moving average? Also, I was wondering if it might be possible to expand the graph with a data set for a wider period. Would be interesting to see for instance if the cyclical trend is consistent throughout years and how the threshold number of points for the 104 ranking has changed (if at all significantly) over time.

Edit: Aha, I think I see what you have done - did you aggregate each of the years between 2000 and 2008 and use the average points at each week of the year? If so, I still think decomposition of the values across a time series would be of merit as it would allow us to see ranking patterns over time.

Action Jackson
03-18-2011, 12:18 PM
Well the points were doubled not long ago.

n8
03-18-2011, 12:23 PM
You are right - updated with lines...

That was quick! Thanks, much better. You can definitely see it for the US Open.

Well the points were doubled not long ago.

Data period ends December 2008 and the ranking points were doubled in January 2009 so it's OK.

ApproachShot
03-18-2011, 12:24 PM
Well the points were doubled not long ago.

As StatRacket pointed out, the data set excludes the period when the points were doubled. But even so, that shouldn't necessarily affect the results as presented in these graphs though (unless that means that recent results are weighted more than past ones). Sure, there might be a magnitude effect but the point here I believe is to illustrate that the points total of the person ranked at #104 increases prior to a Grand Slam tournament. That relationship can hold whatever the ranking structure.

duong
03-18-2011, 12:28 PM
Aha, I think I see what you have done - did you aggregate each of the years between 2000 and 2008 and use the average points at each week of the year?

exactly.

Actually I understand it's quite complicated for people who are unused to stats to understand that : I think it should be more explained ;)

the topic is not actually the evolution over the years but the evolution of points within a calendar year : precisely within a year, the points are higher at the precise moment when the direct entries for slams are gifted.

You could do it within only one year like 2008, but the stat is more precise if you make an average over several years

(the concept of "moving" average here being not necessary : imo you should cut it within one calendar year and not writing any years here but rather only the dates ;) )

As StatRacket pointed out, the data set excludes the period when the points were doubled. But even so, that shouldn't necessarily affect the results as presented in these graphs though (unless that means that recent results are weighted more than past ones).

exactly again :yeah:

the graph not being about the evolution over the years but within one year, the doubling of the points doesn't affect anything (except a higher weight for recent years but it's completely marginal)

ApproachShot
03-18-2011, 12:37 PM
Completely correct, duong :)

OP, I admire your work and I would just suggest making some minor improvements to it:

Label the x-axis as 'weeks' or 'date' and have it going from 1st Jan to 31st December
Replace 'moving average' with just 'average'


Do you have the raw data and if so, would you mind sending it to me (can contact me via PM to discuss how best to do that)? Although eyeballing the data makes shows that prior to slams the ranking points do seem to go up, I'd like to do some statistical significance tests on it.

Also, can anyone come up with some plausible reasons why the average points should decrease so dramatically after the AO cutoff?

amirbachar
03-18-2011, 02:16 PM
Completely correct, duong :)

OP, I admire your work and I would just suggest making some minor improvements to it:

Label the x-axis as 'weeks' or 'date' and have it going from 1st Jan to 31st December
Replace 'moving average' with just 'average'


Do you have the raw data and if so, would you mind sending it to me (can contact me via PM to discuss how best to do that)? Although eyeballing the data makes shows that prior to slams the ranking points do seem to go up, I'd like to do some statistical significance tests on it.

Also, can anyone come up with some plausible reasons why the average points should decrease so dramatically after the AO cutoff?

The moving average is of 2 weeks, since the actual date changes from year to year.
The raw data is here:
http://stevegtennis.com/rankings/'

Maybe the average points decreases so dramatically after the AO cutoff, because the first tournaments' cutoff is very high, therefore some players around that spot simply lose points without being able to defend them...