Can you explain how these rankings work?  What are the scores based on?

Yes. It's based on the ratings of the opposition.

This is the basic premise of it...

Diff in ratings   % chance of victory
+400               100
+300                 98
+200                 94
+100                 77
+50                   65
+25                   58
+10                   53
0                       50
-10                    47
-25                    42
-50                    35
-100                  23
-200                    6
-300                    2
-400                    0

So these are the chances of various players beating Boggo...
Andy Murray    97%
Jamie Baker    44%
Bloomfield       36%
Goodall           29%
Skupski           13%
Ward                 7%
Seator               1%

To calculate a player's new rating the formula is...
    13 * (Result - Chance of Victory)

(It's a bit more complicated than that in that it takes into account the score, surface and the player's age too.)

Eg. if Boggo played Skupski
... and Skupski won - Skupski's rating would increase by 13*(1-0.13) = 11.3, and Boggo's would fall by 11.3
... and Skupski lost - Skupski's rating would fall by 13*(0-0.13) = 1.7, and Boggo's would fall by 1.7


Although there is no limit of 1000. I calibrated it around that. Federer's current rating is 981.

Very interesting - looks like a slightly simplified Elo rating system with surface adjustments, though I'm intrigued as to how you adjust for age. Sad as it looks, particularly re. the Boggo-Josh line-up, it's probably quite a lot more accurate than the real rankings and probably gets more accurate the further down the list you go as well.

Even more intriguing is: how on earth did you manage to collect all the data to do this? I've been interested in the idea of doing something similar for a couple of years, but not being able to get hold of the massive amount of data that would be needed in an easy to use format has always put me off.

DJ wrote:

---

Does that mean that Joshua Milton's victory this morning, over a guy ranked almost 450 places above him, means he rises by 13*(1-0)=13 points, to 269.1 - or have I missed something? Do I need to know his opponent's ranking in YOUR table to make the calculation, rather then his ranking in the ATP lists?

---

It's the opponent's rating (not ranking) in my table it uses.

So Milton/Marek example...
Milton was rated 256.1.
Marek was rated 390.6. 
That's difference of 134.4, which equates to a 16% chance of a Milton victory.

So Josh Milton's rating would rise by 13 * (1-0.16) = 10.9.



 

Right, got it, thanks.
It does mean, however, that we out here can't make sensible use - for our own interest, that is - of your table without having the whole of it, not just the Brit bits, accessible to us; equally, only the up-to-date version would be of use. Do you update it every week? Would it be something you would care to share with us on a regular basis?
I'm looking forward to reading your response to Steven's queries, too.
Thanks for providing us with us, as well as your African fantasies!

And Steven, yes it is based on the ELO system.

The age adjustment, is there so that the ratings rise (or fall) quicker for under 21s. It's that constant of 13 that changes.

18 & Under - 26
19 -             22
20 -             17
21 & Over -  13

As that age their ability changes far quicker than when older.

Also it adjusts that constant to make matches played when under 18 less important. eg. If someone plays a match at 14 years old, and didn't play again until he was 18, that initial match's constant is 6 times less important.

Because a player's ability at 14, bares little resemblance to a player's ability at 18.