Happy Meeple forum
Our games => Cartographers => Topic started by: Geraldine M. on 12/10/20, 06:25am

Are the ELO ranking changes weighted differently for this game?
I know we are getting +6 points for the first 20 games, but I mean something different.
For example, I just lost to someone with a ranking over 400 points ahead of mine, and my ranking was reduced by over 8 points as a result (which presumably means that my ranking would have dropped over 14 points without the 6point bump). But in other games, like Lost Cities, when the starting rankings are very different, the penalty for the weaker player losing seems much smaller.
Please note I'm not complaining at all about my ranking. I'm just very confused about how it is being calculated.

Hello,
This is an interesting topic. I also have analyzed the ELO ratings in our website for quite a long time.
It has come to me to find our the formula to calculate our ELO as following:
New_ELO = Old_ELO + K*( Result  1/(1+10^((Opp_ELOOld_ELO)/800)))
Where:
New_ELO = your new ELO after the game
Old_ELO = your old ELO before the game
Opp_ELO = your opponent's ELO
Result = 0 if you lost, Result = 1 if you win
K = Kfactor, which will be discussed below
So, basically, the difference between this formula and the popular formula you can find in the internet is the number 400 (such as in FIDE chess ELO calculator) is replaced by 800.
I also do believe that our website has some cap for the ratings you can gain/lose in a game, to avoid extreme cases (e.g: a 2000player lost to a 1000player will not lose too much ratings as in the formula).
The problem you see is because the Kfactor is not fixed in our website. It is depending on the number of games you have played before (i.e Kfactor changed as you are more experienced with the game).
For lost cities, I guess you have played 300+ games (or even more?), hence the Kfactor is set at the minimum value, i.e 23 (or somewhere near that). But for new games like Cartographer, while you have played less than 50 games, the Kfactor is very high (somewhere from 80 downto 28?, it will eventually reduce to 23 at some points when you play an enough number of games).
That's why in new games like Cartographer, your rating jumps quite fast after a single game, unlike in the games that you have played for long time such as lost cities.
___________________
All of the above is just my observation and reasoning. I do make an excel file to calculate and keep track of my ratings to guess the value of Kfactor used in each game, but it is rather messy because of the number rounding problem, and somehow lost and draw at early messed up my formular (?!). Hope that someone can give a full detail about how the value of Kfactor is applied.

That makes sense. Thanks.

This is almost perfect. Thanks!
K is 28. That means, if you lose against a player that has the same rating, you lose K/2= 14 points. And if you win, you win 14 points.
K is bigger for your first 25 games, from 28*2.5 down to 28*1. So for the first game, K=70 and it goes down from there.
We have two other modifiers.
1) Slower changes for experienced players
if (nb_of_games>300) then K=K*0.83;
else if (nb_of_games>100) then K=K*0.88;
2) Faster moves for low rated players to give them a chance to come back a bit faster
if (rating<=1200) then K=K*1.6;
else if (rating<=1300) then K=K*1.3;
else if (rating<=1400) then K=K*1.15;
That's all.

This is almost perfect. Thanks!
K is 28. That means, if you lose against a player that has the same rating, you lose K/2= 14 points. And if you win, you win 14 points.
K is bigger for your first 25 games, from 28*2.5 down to 28*1. So for the first game, K=70 and it goes down from there.
We have two other modifiers.
1) Slower changes for experienced players
if (nb_of_games>300) then K=K*0.83;
else if (nb_of_games>100) then K=K*0.88;
2) Faster moves for low rated players to give them a chance to come back a bit faster
if (rating<=1200) then K=K*1.6;
else if (rating<=1300) then K=K*1.3;
else if (rating<=1400) then K=K*1.15;
That's all.
Well, actually for the first game, player always start at 1400 rating, so K=70 is also multiplied by 1.15, which gives 80.5 (which is close to my calculated value of 80).
The rest makes sense to me!