Trying to understand ELO calculations...

Jaxel

Administrator
So some of you know I am writing an ELO ranking ladder... and I'm having issues understanding some of the concepts. The programming is working fine; I'm just trying to figure out WHY things are programmed the way they are...

I am following the article on wikipedia:
http://en.wikipedia.org/wiki/Elo_rating#Mathematical_details

1.png


As you can see, the calculations use several constants. I understand what K(factor) is... but the problem is that the constants for "10" and "400" aren't really explained. Why are these numbers what they are? What do they do?
 
Well I did the math with different "10" and "400" values. And I can now see what they do...

Basically, these numbers calculate the strength of expected wins/losses. Basically, if I had the numbers set to "1" and "200", and a 1600 ELO player goes up against a 1400 ELO player... the 1400 ELO player only has about a 1% expectation to win; and the 1600 ELO player has a 99% expectation. So if the 1400 player wins, he will gain 99% of the K-factor rating. If the K-factor was 100, he would go up 99 points... and the 1600 player would lose something similar.

Basically, "400" is the measurement of "skill gap". While "10" is the basis for how hard you want to calculate against the skill gaps. It also has the side-effect of being the minimum win expectation for a player. A player can never have less than a 10% expactation, or more than a 90% expectation to win.

The skill gap actually follows a curve, the further away from the median the two players get from each other, the more an expectation of win skews in someone's favor.
 
For those who have been asking, this new ELO stuff I am working on is for a NEW section to the forums, for an online ranking ladder. Where people can challenge others on PSN/XBL and report their scores here on 8wayrun. Very similar to what we saw back in the day with GameBattles, but here on 8wayrun.

The current offline tournament rankings located here, will NOT be moving to ELO and instead will remain as ATP-ERS. There are several significant reasons for this... allow me to explain.
  • Not everyone who attends offline tournaments are registered here on 8wayrun... or, they don't sign up with their 8wayrun usernames and it becomes hard to track player's performances. ELO requires EVERYONE who attends a tournament to be registered in order to get ranked. ELO is calculated based on the registered rankings of your opponent... but if someone isn't registered, then there is no way to gauge their skill.
  • ELO is not only based on players, its also extremely time sensitive. It not only gauges WHO you played, but WHEN you played them. You could beat someone in a tournament last week, and they could have beat you in a tournament, last month. But if the results for the tournament last month get submitted AFTER the tournament from last week; it skews the ELO rankings.
  • The ATP-ERS system awards flat points, based on the size and relative importance of an event, rather than who competed in said event. Because of this, we can leave unknown players in the rankings and they don't skew point awards that much. Then, if later on in time, the unknown player registers on 8wayrun, they can "claim" their missing points from said event and add it to their performance.
  • Because ATP-ERS puts more of an emphasis on the EVENT, rather than the PLAYERS of the event, it helps better promote the idea of attending more events in order to increase rankings. With an emphasis on players, it actually deters the idea of attending more events in an effort to avoid point spoilage and putting your rank at risk to upsets.
Basically ELO only works in a closed, highly organized system. If we wanted to use it, every event on the ladder would have to be pre-approved for accuracy, and every person who goes to said event must sign up using their 8wayrun username, or must immediately create an 8wayrun account when signing up.

Since tournaments are run by hundreds of different people, there is no way to force them all to do that. When these TOs are running majors for much larger games, there is no way they would go along with such demands for a community as small as ours.
 
NO ISSUES WITH THE PROBLEMS YOU MENTIONED REGARDING THE ELO SYSTEM...BUT I HATE THE ATP-ERS ONE JUST AS MUCH. WE NEED SOMETHING THAT HAS A MORE SENSITIVE MEASURE FOR SKILL. WHAT DO YOU THINK ABOUT RANKING PEOPLE BASED ON THEIR AVERAGE PLACEMENT IN TOURNEYS? WHEN YOU COMPARE THE PLACINGS, YOU'LL HAVE TO LOOK AT THE PERCENTILE SCORES BECAUSE NOT ALL TOURNAMENTS HAVE THE SAME NUMBER OF ENTRANTS. SO BASICALLY YOU FIGURE OUT WHETHER 8TH PLACE WAS TOP 10 PERCENT OR TOP 50 PERCENT IN A PARTICULAR TOURNEY, BASED ON THE NUMBER OF ENTRANTS.
 
IAB, thats exactly what the ATP-ERS does. It gives more points based on the number of entries in a tournament. The ATP-ERS system treats everybody equally; every player contributes 150 points to the event value. Events do have maximum values however, based on the type of the event; for instance, a "major" maxes its value at 5000 points, which is about 33 people.

Also, from now on I will be completely ignoring your posts.
 
Last edited:
Back
Top Bottom