Isnt current ranking system too easy to manipulate? (to gain high rank with low skill)

[thankforpie]

What i mean is exlusively the fact that you with rank such as 1600 or 1800 can play vs people rated 1200 and still gain a lot of points?

 

I understand this could be implemented because there wasnt enough players but isnt it making the rating system much less purposeful? if high rating == high skill, then you shouldnt get points from low rated players,because that makes the entire list much less consistent with actual player skill in reality

 

for example i saw some guy 1600 drop to 1500 (some games vs me, some vs others) and he quickly made it back to 1600 on people who have default rating (1200) or near (1100) and have absolutely 0 clue about game.

 

in current ranking system you can easily get to 1600 (which means become to get in top 100 of players) just by playing with people who are not harder than medium petra AI if you are willing to spend the time

[Guest]

ELO does take this into account. A high rated player winning against a low rated player would yield a smaller gain depending on the point difference. With a 600 point difference yielding no points gained and the loser with no points lost. Easy test; play vs borg.

[thankforpie]

4 minutes ago, (-_-) said:

ELO does take this into account. A high rated player winning against a low rated player would yield a smaller gain depending on the point difference. With a 600 point difference yielding no points gained and the loser with no points lost. Easy test; play vs borg.

is the 600 points difference factual limit? because it means exactly what i said, that you can play vs new players to get in top 100 (1600) (only 400 points difference here, because new players are about 1200) or even up to 1800 (600points difference, if gives no points then just play vs 1250 rated player and ranking rises...) and that makes whole rankings at least up to 1800 very untrue

Edited October 29, 2018 by thankforpie

[Stan`]

5 minutes ago, (-_-) said:

EASY test; Lose vs borg.

Fixed

[Guest]

Well, a 400 difference would give only a few points. Anyway, being a widely used form of ranking I am sure there is reasoning behind this.

If the values are that broken, I suppose the formula could be tweaked. But not convinced yet.

Edited October 29, 2018 by Guest

[Imarok]

For sure there is one issue: If a player always plays against bad players and does that often he can get a very high rating.
So our system favours players that play often and always against weaker players.

The only solution I see would be some kind of matchmaking, where wfg picks a random opponent for you and you are forced to play that game. But that seems difficult to realize.

[user1]

It appears that the difference between 1400 and 1600 is an order of magnitude and the difference between 1600 and 1800 is another order of magnitude. This is shown by the fact that ~1600 is top 100 but ~1800 is very rare in comparison (top 10), and >2000 is ultra rare. 

IMO it would be pretty hard to get much higher than 1600 by playing low rated players.

It's also not difficult to get into the top 100 because there are not thousands and thousands of active rated players; just hundreds. If you assume there are only 200 active rated players then top 100 would really only be just somewhere above the 50th percentile of active players. Top X is always arbitrary and can only be considered a measure of skill as it compares to the total # of competitors. The total # of competitors is kind of unknown and many(most) rated players are inactive and/or have only played a very small # of games.

In short, the value of the rating system is increased with more participation and participation is relatively low.

Also, interpreting a player's rating as a measure of skill requires knowledge of certain variables that aren't exactly easily known, but for instance, make sure you at least consider the # of rated games the player has participated in.

[Guest]

If all the data from a game report is saved, then I guess a fancy new profile lookup system could be implemented. Judging a player based on scores etc would be easier than just looking at the rating.

[DarcReaver]

Not sure where the problem is. The game is barely more than a tech demo at this point, and the rating - if at all - is only to be used as an identifier if you're playing against a "newbie" or someone who already played lots of games. To prevent that the majority of online tests are noobstomps.