The best explanation I can use is to copy/paste what PAPA did with their PARS system, since we pretty much copied what they did using tournament results as the data rather than head-to-head matches:
It uses the Glicko algorithm, a sophisticated method of interpreting competitive play and determining how accurate player ratings are at any given time. The same algorithm is widely used in the chess community, as well as for many online gaming systems. The essence of the system is this: When you win, you gain rating points. When you lose, you lose rating points. How many points each player gains or loses depends on the quality of each opponent. If you beat a great player, you stand to gain a lot of points. The more you play, the more accurate everyone's rating becomes.
If your question is currently how we're using the IFPA tournament data to calculate that Glicko Rating, it's easiest for me to use an example:
For purposes of the IFPA Rating calculation, it looks at these results as a bunch of simulated head-to-head matches, so it sees you in first place and does the following:
Erik Wurtenberger d. Larry Smith (1-0)
Erik Wurtenberger d. Casey Rice (1-0)
Erik Wurtenberger d. Rob Uzzolino (1-0)
So you in essence "won" 30 matches that day. Your IFPA Rating increased by 58 points that day.
If I look at Larry Scott from the same event, he finished tied for 8th. The system looks at that as him losing to the 7 players above him, finishing with a draw to the two other players that finished in 8th, and defeating the 20 players below him. (20-7-2 record)
Larry's IFPA Rating increase by 3 points that day from that activity. Even though he went 20-7 as a record, he only beat one player rated higher than him while losing to 3 players rated lower than him. The Glicko Algorithm values all of those matches individually so the credit Larry got for defeating a bunch of players rated in that 1000-1200 range didn't do much to boost his own rating for those wins.
We actually limit the number of simulated matches that count for any one tournament result. It's limited to the closest 32 players that you finish higher and lower than. So if your Cincy League had 100 players and you won, your record would be 32-0 with "wins" coming from the players that finished in 2nd through 32nd place. You wouldn't get any credit for "beating" the players that finished in 33rd through 100th place. This was to limit the volatility of the Ratings calculation from any one result.