amirbachar

04-01-2011, 10:17 PM

Since I finished working on that, I'll post my rankings here too:

http://simtheworld.blogspot.com/2011/03/all-time-atp-rankings-update.html

The methodology used:

I'll use the notation A to speak about a player with current ratings A.

I start by assigning arbitrary rankings for all players (let say everyone start with 1000).

Then, for each set between A and B, the expectancy for A is A/(A+B). For each player we can calculate the expectancy sets won according to the sets he actually played.

If the expectancy is lower than the number of sets he actually won, we increase his ratings.

If the expectancy is higher than the number of sets he actually won, we decrease his ratings.

That system is called KRACH.

Now, to adjust it to all time tennis rankings:

I assigned set factor, according to the importance of the set n the match (so final sets worth more).

I assigned tournament factor, so GS for example worth more.

And finally, the latest addition - for each player I ranked all the years he played in according to the sets won/expectancy (like before).

For the best year for each player, there is a factor of about 1.9, for the 2nd best about 1.38 and so on... So I factored each match by the minimum of both players factors in that year.

http://simtheworld.blogspot.com/2011/03/all-time-atp-rankings-update.html

The methodology used:

I'll use the notation A to speak about a player with current ratings A.

I start by assigning arbitrary rankings for all players (let say everyone start with 1000).

Then, for each set between A and B, the expectancy for A is A/(A+B). For each player we can calculate the expectancy sets won according to the sets he actually played.

If the expectancy is lower than the number of sets he actually won, we increase his ratings.

If the expectancy is higher than the number of sets he actually won, we decrease his ratings.

That system is called KRACH.

Now, to adjust it to all time tennis rankings:

I assigned set factor, according to the importance of the set n the match (so final sets worth more).

I assigned tournament factor, so GS for example worth more.

And finally, the latest addition - for each player I ranked all the years he played in according to the sets won/expectancy (like before).

For the best year for each player, there is a factor of about 1.9, for the 2nd best about 1.38 and so on... So I factored each match by the minimum of both players factors in that year.