ESL Ladder rating system update
Posted: Tue 20 Apr, 2021 6:15 pm
This update was done a few hours ago by WiseWindu, here are some details.
-Update ESL formula for the following values:
C value determining the rate at which r/d applies after a time period of inactivity increased from 24 (or 18, check) to 65. c = 65 is equivalent to going from 60 r/d (reliable confidence interval as per professional chess) to 250 (no longer on leaderboard due to low confidence interval) in 30days of activity in a row, aka 15 time periods.
T value changed from t = 86400 (seconds) to t = 172800 (seconds) [aka 48hours/2days], meaning you only start to have your r/d and thus elo decay if you fail to play a game once every 48hours. At the moment this value is 1 day.
R/d elo Multiplier bonus reduced from 50% to 15%.
If one player has an r/d of less than 100 the r/d elo multiplier increases from 15% to 30%. This rewards you extra for beating players with low r/d. This should make it less valuable to farm noobs, especially alongside the increases rate of decay and overall reduced r/d multiplier, and makes reliable players just under top tier prime targets, but that in turn means they'll get loads of practise potentially setting them up to become top tier players in the near future.
ESL XP modifier for winning increased from 125% to 150%.
Following ESL modifiers will not be modified:
The Minimum allowed r/d value remains at 30, with an increased time period it is possible that this may need to be increased because the lower the games played to time passed ratio, the less accurate the elo assessment is, therefore accuracy decreases as T increases, but for now we are unsure if it is necessary to increase this value; we'll have to see how many people reach elos around 60, atm only Lacia achieved it for more than a day or two and I don't see the doubling of the T value to change that too much.
The maximum r/d you can have before being removed from the ladder remains at 250. 250 is already a gigantic amount of r/d and makes your elo rating very questionable. Anything higher than this is not worth its salt in terms of accuracy and thus you have no right to be on the ladder. Especially with a bigger time period. Again, the larger time period if anything might mean we have to decrease this number. It all depends on how many people fail to stay below it - if 95% of active players stay under 250 r/d then we will decrease this value.
Ladder "safety time" remains at 600. Games are not calculated that are newer than this. To be frank I do not know what this value refers to and that alone is why it has not been changed - there's no other logic behind it.
ESL XP modifier for losing remains at 100%. Punishing people for losing is just too brutal and rewarding people for losing is dumb. Not really contentious.
-Update ESL formula for the following values:
C value determining the rate at which r/d applies after a time period of inactivity increased from 24 (or 18, check) to 65. c = 65 is equivalent to going from 60 r/d (reliable confidence interval as per professional chess) to 250 (no longer on leaderboard due to low confidence interval) in 30days of activity in a row, aka 15 time periods.
T value changed from t = 86400 (seconds) to t = 172800 (seconds) [aka 48hours/2days], meaning you only start to have your r/d and thus elo decay if you fail to play a game once every 48hours. At the moment this value is 1 day.
R/d elo Multiplier bonus reduced from 50% to 15%.
If one player has an r/d of less than 100 the r/d elo multiplier increases from 15% to 30%. This rewards you extra for beating players with low r/d. This should make it less valuable to farm noobs, especially alongside the increases rate of decay and overall reduced r/d multiplier, and makes reliable players just under top tier prime targets, but that in turn means they'll get loads of practise potentially setting them up to become top tier players in the near future.
ESL XP modifier for winning increased from 125% to 150%.
Following ESL modifiers will not be modified:
The Minimum allowed r/d value remains at 30, with an increased time period it is possible that this may need to be increased because the lower the games played to time passed ratio, the less accurate the elo assessment is, therefore accuracy decreases as T increases, but for now we are unsure if it is necessary to increase this value; we'll have to see how many people reach elos around 60, atm only Lacia achieved it for more than a day or two and I don't see the doubling of the T value to change that too much.
The maximum r/d you can have before being removed from the ladder remains at 250. 250 is already a gigantic amount of r/d and makes your elo rating very questionable. Anything higher than this is not worth its salt in terms of accuracy and thus you have no right to be on the ladder. Especially with a bigger time period. Again, the larger time period if anything might mean we have to decrease this number. It all depends on how many people fail to stay below it - if 95% of active players stay under 250 r/d then we will decrease this value.
Ladder "safety time" remains at 600. Games are not calculated that are newer than this. To be frank I do not know what this value refers to and that alone is why it has not been changed - there's no other logic behind it.
ESL XP modifier for losing remains at 100%. Punishing people for losing is just too brutal and rewarding people for losing is dumb. Not really contentious.
