Establish Rating System for My Community

Question

I want to establish a rating system for my community.

1. I've decided to publish the rating every 1st January, April, July and October.
2. Every new player will be given a rating of 1200.
3. Using result submitted to me by tournament organisers, I will get every players' Rating Performance for any tournament.
4. Every time I want to publish a new rating, I will calculate the Average Rating Performance for every player.
5. Let say between January to April, Player A's Rating Performances are 1300, 1200, 1400 (Player A participated in 3 tournaments)
6. So by April 1st, Player A's rating is (1300 + 1200 + 1400)/3 = 1300

Is this correct?

Edit

To put it simple, am I correct to use Player's Rating Performance to base my rating system?

Or should I calculate players' rating based on every player's encounter using calculator like this: https://ratings.fide.com/calculator_rtd.phtml ?

Answer 1

It is better to look at every game. Only looking at tournament performances causes you to lose some important information.

Should a 4 round tournament count the same as a 9 round tournament? Under your system, they count the same. That doesn't seem right.

I had a tournament recently where I won against 3 low-rated players and drew against a player rated over 2000. My performance rating was 2041. But there's a difference between a 2041 performance done this way, and a 2041 performance achieved by scoring 2/4 against players rated over 2000.

It's worse when a player wins (or loses) every game in a tournament. Are you going to make a player's rating go down after they win every game, if their opponents were too weak and thus the performance rating for that tournament was low? That doesn't seem fair.

Also, what happens during the next rating period? How do you account for the fact that the player already has a rating? Are you averaging in the old rating? But if a player with a well established rating plays in one tournament in a period, simply averaging the old rating and the one new tournament seems to put too much weight on the one new tournament.

Answer 2

The way elo works is that they take current rating and tournament performance, and then adjust a new rating based depending on k-factor. For all intents and purposes they count every game during a tournament as happening at the same time. So your step 6 is wrong because it doesn't weigh the latest result more than the previous tournaments. If somebody plays their first 10 tournaments with a performance of 1200 and then their last 5 tournaments they played at a 2000 level, they are clearly much closer to 2000 strength than 1200 now.

Answer 3

Starting new players at 1200 will distort the ratings and make it take longer , with a LOT more games played in total, for the ratings to settle down to anything near accurate.

You should first have a big tournament where most all local players play a lot of rounds. Swiss would work best if you have more players than you could have enough rounds for a RR.

Then set initial ratings based on results not some PC approach where everybody starts with 1200. Some players will be experienced and have very high actual ratings while some will be beginners whose actual strength is very low. Starting them off equally is nonsense.

Statistically far better to rate each player based on each specific game rather than some so called average performance that might not reflect reality as well as each game individually does.