Jump to content
  • Sign Up

Rating Volatility & Impact on Player Frustration


Ragnar.4257

Recommended Posts

Alot of players have the impression that the current ranked ladder is just "random luck". They get upset when their rating gets tanked by being placed in a match that they straight-up have no hope of winning.

Generally, this gets attributed to the problem of personal rating being assigned in a team based game.

Now this isn't actually a problem at all. Sure, in any one game, you might be unlucky and placed on a team that is totally un-carryable. But over the course of 100+ games, this will be off-set by other games where you get lucky, and the enemy team is the unlucky ones, and you have a basically free win.

So the randomness in team rosters is supposed to be off-set over the course of a season by "averaging out" the random-factor. Over the course of 100+ games, the games where you get a bad team cancel out the games where you get a good team, which means the only remaining factor is your own personal skill.

This is called "increasing the sample size". If you conduct an experiment, you repeat it to make sure it wasn't just down to luck. If you conduct a survey, you include as many people as possible, so that your result isn't impacted by random variance in the population.

However.

The volatility in ratings is where things become problematic. In the current setup, it only takes 3 losses in a row to tank your rating down not just one, but TWO whole divisions. It doesn't matter that for the last 100 games you've consistently held a rating of X. Just those 3 games lost is enough to throw you waaaaay down the leaderboard. And likewise, it only takes 4-5 lucky wins to bump you up not one, but TWO divisions higher than you should be.

THIS is why people get frustrated with the ratings/leaderboard, and get toxic with their team-mates, because just 1 or 2 losses is enough to send them tumbling down the ratings. This is why top players camp rating, because it doesn't matter if you've been holding a top-5 spot all season, it just takes 3 unlucky games in the final week to throw you out of even the top-50. This is why after only 2-3 losses, players are liable to throw up their hands in despair and blame "the system".

Effectively, the leaderboard only cares about the last ~10 games you played, and this means that the sample-size solution to the random-team problem is being negated.

Now consider, if we increased the number of games required each season by 50% (to 180), and cut rating gain/loss by 50%. So where a win/loss before might give +10/-20, now it would give +5/-10. Now your performance over the season as a whole has greater importance, and the impact of a few lucky/unlucky recent games is diminished. Players become less anxious about winning or losing a particular game. There is less toxicity in matches. And players are forced to pay attention more to their rating as a trend over the season, rather than as simply the random result of the last ~10 games, and will therefore spend less time whining about matchmaking, duo-queue, match-manipulators, etc etc.

Why isn't this a thing?

Speaking for myself, I don't care much about by absolute position on the leaderboard. But I do care about my rating in so much as I care about getting good games. I know that if I drop 100 rating then I'll start getting put in matches that just aren't fun at all, because both my team-mates and the opposition are laughably incompetent. I have therefore, on occasion, become somewhat toxic and ragey when on a 5 game losing streak. Because just those 5 losses is enough to drop me WELL down below where I should be playing.

It doesn't have to be this way.

Link to comment
Share on other sites

Team rating would be great if teams were actually a thing. They aren't though. Everyone is thrown into the same ranked queue together and expected to just be friends and get along, even though your teammates rating effects the volatility of your rating gain/loss in matches.

Like for instance; getting a highly rated or smurfing DuoQ on your team brings down the ratings of everyone in your match; particularly your own team, so you gain less or lose more. Another thing that happens a lot is getting a teammate that just afks in spawn without a word. With the current rating system that afker is set to lose an equal if not lesser amount to everyone else in the lobby totally based on their own rating compared to the other players in the game and nothing else.

No amount of analyzing or further tinkering the numbers is going to change that. It's a very dated and broken system compared to other games that use a combination of personal contributions to a match, streaks, and ratings. Imo, it needs total restructuring to be like that. A more sophisticated, more modern algorithm. Increasing the amount of games required is just going to deflate people's ranks which will probably result in even worse matchmaking than there is already, AND having to play another 60 games just to keep up with the LB sounds like actual torture.

Maybe it would be better if it were; y'know, fun and rewarding. Where good players are rewarded for good play, and every step forward isn't immediately followed by two steps back because of factors totally outside their control.

Link to comment
Share on other sites

I have talked about this a couple of months ago - actually, GLICKO-2 allows this reduced rating volatility already. There is a value... uhm, I am too lazy to look for it, something like "minimum rating deviation per game". At some point, the games do not have reduced rating changes, even though - as you say - the sampel size gets smaller and GLICKO-2 becomes generally more secure about your personal rating.

So: I am with you and I would like to see it implemented. It shoud also be not that hard - the first few games player per season already show this behavior: The first 15ish games have high volatility, after that the changes become more or less stable. This is probably exactly this "minimum rating deviation per game".

However, it needs to be considered that this also means:

  • if you troll in your first games, you will have a very hard time climbing in your later games,
  • if you switch classes or your class gets nerfed/buffed in the balance patch, few things will change,
  • and if you for some reason improve significantly over one single season, this will mainly affect the next season played.

However, I would not like to see 180 games required per season. GLICKO-2 becomes very secure about ones rating very quickly (generally speaking). And I already struggle with 120 games rather often, because I do not really have that much time. It is not really needed for the system too.

€: Here it is. Is this the same thing you propose? Then it might already be implemented in the "confidence level" of GLICKO-2 and would probably be rather simple.https://en-forum.guildwars2.com/discussion/43866/suggestion-to-make-people-play-more-games

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...