Sorry if this has been asked before - I am not a chess player and I am a bit confused.

I understand that if you beat someone much lower than you, you only gain a little (you were expected to win). And the opposite applies - if you lose against someone much higher than you - you gain a lot and they lose a lot.

But in a game of similarly ranked players, there also seems to be a lot of movment - yet isn't it expected that either could win - and thus the ranking not move much? For example, in a recent streetsoccer game a 1620 ranked player beat a 1649 ranked player. The change to both was 17 points.

This seems excessive?

Basically, it is ELO Rating System. You can learn more at following link.

http://en.wikipedia.org/wiki/ELO_rating_system

If you beat a player with the same rating as yourself, you gain 16 rating points (and your opponent loses the same). The change is 32*(R-E), where R is the result, and E is the “expected result”. The expected result is M/(1+10^((O-Y)/400)). Here, M is the maximum result for the game (5 for street soccer, 2 for all other games), Y is your rating and O is your opponents rating.

Thanks for the info - so I tried to create a spreadsheet showing what the result would be - and I cannot get the values to work… What am I doing wrong:

Let's say you are playing a player of equal value in streetsoccer. You win.

change = 32*(R-E)

E = M/(1+10^((O-Y)/400))

You won, so

R = 5 (assuming streetsoccer)

and

M = 5 (max possible result in streetsoccer)

Lets break this term M/(1+10^((O-Y)/400)) down

(O-Y)/400 = 0 (since you have the same rating)

thus 10^0 = 1

and thus E = 5/(1+1) = 2.5

So

change = 32*(5-2.5) = 80

It should be 16. What am I doing wrong???

I think the change should be 32*(R-E)/M instead of 32*(R-E).

But in a game of similarly ranked players, there also seems to be a lot of movment - yet isn't it expected that either could win - and thus the ranking not move much?

Following that logic, if two similarly ranked players played each other 10 times, they'd still be similarly ranked, even if one player beat the other one all 10 times. A certain amount of volatility has to be expected.

Yes, Tasmanian is right. I forgot to divide by M. The correct rating change is 32*(R-E)/M. Sorry.

I understand there's no weighting factor how 'stable' a rank is. Meaning it does not matter if one played 100 games already or just starts with a random 1500 initial setting.

The expert's rank changes as much as the newbie's ?

Very strange to me. Or I just do not see it correctly?

That is not correct. If a new player N plays against an established player E, N's rating changes as much as if he were established, but E's rating doesn't change as much. There may be exceptions if one player loses on time before making a move (?).

Well, I am fairly new here, and my 8th game (only 6th ranked game I think) was against a player with over 400 games. I won. This is the message from the system

TnT old rating: 1488 newrating: 1509

old rating: 1608 newrating: 1591

Seems like both moved the same amount

\* = name removed to protect the innnocent :-)

Rating change TnT=1509-1488=21.

Rating change opp=1608-1591=17.

So the ratings did NOT change by the same amount. Like TD said, with one experienced player and one newbie, the rating of the experienced player moves less than that of the newbie. How much less depends on the number of games the newbie has played. With two newbies, the rating change the same, just as with two experienced players.

If I apply the numbers to the formulas given on 28-Aug, I get for TnT a change of +21.03 and for his poor opponent a change of -21.03.

And I cannot see how the experience influences the calculations.

FAQ tells the real formula is not interesting.

This is not true ;)

It's not the experience what change the formula, but the likelihood of the rating.

I mean, if you are a 2100 points gomoku player and register with another nick (or just you are a very good player, but new to LG), then it's better to set your rating up fast without messing down everybody's rating :)

I have not yet figured out exactly how the experience influences the calculation, but basically, if one experienced player playes against an inexperienced one, the rating of the inecperienced player is calculated normally, while the rating change of the experienced player is multiplied with a correction factor, c. c is less than one, and increases with the number of games the inexperienced player has played, until it reaches one. If you really want to know how c increases with the number of games, send me a message, and we can conduct an experiment to find this out (although it would be easier if Richard just told us! :-) ).

Another complicating factor is when two inexperienced players play against each other. If they have played the same number of games, the rating change is calculated normally (no correction factor). If one of them has played 1 game, wile the other has played 4, I have no idea how the correction factor is calculated.

… conduct an experiment to find this out (although it would be easier if Richard just told us! :-)

I understand rating does not apply to private games ?

Hard to exactly meet in a tournament ?

BTW:

Who is Richard?

or rather:

Who is LittleGolem in general?

Who refuses to receive donations currently?

Why?

Richard makes it possible for us to play our games and to meet some great people here. He is the creator and webmaster of littlegolem.net.

Forum duscussion about the donations can be found here:http://www.littlegolem.net/jsp/forum/topic.jsp?forum=1&topic=1049

Richard makes it possible for us to play our games and to meet some great people here. He is the creator and webmaster of littlegolem.net.

Forum duscussion about the donations can be found here: http://www.littlegolem.net/jsp/forum/topic.jsp?forum=1&topic=1049