I was always under impression thet once you hit that 1700, and just play well enough to keep your winrate around 50%, you'll gradually get higher on ladder, despite you're perhaps not good enough. [end of opinion]
Quote from: div.ide on August 01, 2012, 15:27 I was always under impression thet once you hit that 1700, and just play well enough to keep your winrate around 50%, you'll gradually get higher on ladder, despite you're perhaps not good enough. [end of opinion] That's why there is a lot of noobs in games like CD // PD.EU / 2000+++++++ ?
I'm not sure if exact alrogithm is available.But to summarize my observations and some post I read in past but can't find now:- base PSR gain/loss is dependant on average PSR in given game, with higher ranked games giving lower score gains (that's why around 2000 they have to really bust their balls to get any points froim game at all, usually it's something like +1/0. Ask stalker for more details.)- Actual gains/losses for each player are modified from base gain depending on their relative PSR compared to other players in that game. Having your score 150+ above average will result in estimated score being reduced to 0/0 (basically, -100%), while getting below will increase gain (again cap is met at 150 difference, which results, if I am not mistaken, in +100% gain/loss boost, that's why below 1500 you'll usually have !r score around +19/-17 levels, or more)- Then it's further affected by estimated percentage chance for team to win. Having lower chance to win gives you more gain for win while smaller loss for defeat so if you're balancing it at 60/40, sent might have scores like +7/-9, while for scourge it would be opposite: +9/-7 (Basically, thhat's what shpura just said). Familiar uber pro pub stack (pardon my sarcasm) sometimes used my noobscore in game to get it balanced around 70/30 (or further), sometimes ending up having +21/-7 or so (właśnie tak marian nabił 1800 PSRa)- Note to above point: while !balance command tries to match players so average PSR score on each side is equal, chances for each to win are not matching with estimated chance of victory. I am not sure how exactly is it calculated, but algorithm expects that team with higher standard deviation is more favorable.To explain what I just said, consider this:Sentinel has 5 players with exactly 1500 points eachScourge have 3 players with 1500, one with 1600, and one with 1400.Both teams have average of 1500, but since standard deviation is higher in scourge team, scourge will have higher expected chance to win (which also implies slightly less favorable point gains for that match).I can only speculate that having some higher rated players is supposed to outweight disadvantage of having lower rated ones...- These scores are capped at +30/-30, you can never gain or lose more from single match.- Obviously, since scores are integers, whatever algorithm calculater, will be rounded. Point is, it usually seems to round it if favor of player, which is significant in higher ranked games. [following part is pure opinion and might be wrong] This effect seems to really start to kick in above 1700, where players frequently get +4/-3 and such, despite game being balanced 50/50. I was always under impression thet once you hit that 1700, and just play well enough to keep your winrate around 50%, you'll gradually get higher on ladder, despite you're perhaps not good enough. [end of opinion] - EM games, while displaying full score trough !r command, only apply 50% of it after game. Wanna more points? Stop playing -em.I guess that's all I can tell.
Author
Topic: [Question] about the rating system (Read 2013 times)
WiKi
How To
Rules
Ban
Unban
LA Staff
Maps
Streams