Well, that's part of the reason I'm skeptical. Aside from still not seeing exactly how things are calculated (waiting on Bob Jewett to link me), I've seen my rating fluctuate rather drastically (about 100 points) when I haven't played in over a year. I guess the people I've played have got worse?
Cleary I don't know what you are looking for. If you think there is a formula that can be put into a spreadsheet to calculate your rating, get used to disappointment.
Here is a serious answer. Please take the time to try to understand this. If you do understand this description, you understand the essence of what we do.
What is the chance of getting heads twice in the next two flips of a fair coin?
What is the chance of the next card being a King with a fresh shuffled deck?
What is the chance of rolling a 7 with two die?
In each case, the chance is called a PROBABILITY, and we can compute the probability because we start with key knowledge that we all accept (heads and tails equally likely on a flip, each of six sides of a die equally likely, 56 cards in a deck, etc)
Now switch to a pool match.
What is the chance Adam Kieler beats Joey Testa in a race to 5?
If we knew and all agreed on the key core knowledge--the chance each wins a single game--then this would be like the above questions. We could answer it, and the answer would be called a PROBABILITY.
But we don't know the key core knowledge. Instead imagine we start with guessed ratings for each player that gives us a guess for that key number--the chance each wins a single game. Armed with this, we compute the chance each wins a race to 5. But because it is based on uncertain core knowledge we call it a LIKELIHOOD rather than a PROBABILITY. The key to what we do is we move the player ratings around until the LIKELIHOOD is as large as possible.
There are nearly 2 million matches in our system. and the LIKELIHOOD all two million of them turned out the way they did is the likelihood each match turned out like it did multiplied together--a product 2 million terms long. Because each match likelihood is less than 1, this product is an extremely small number.
Here is the critical feature for our computed ratings: If you change the rating for any single player out of tens of thousands of players and compute that LIKELIHOOD again, the result will be smaller. Our optimized ratings are the ones that maximize the LIKELIHOOD. And when a single new match is put in, every single person's rating is tickled.