Morax wrote:
Our rating system is fairly simplistic in that it doesn't take into account what variety of maps a person plays, eg: you could find someone with a 1700 rating that only plays GAP and nothing else. You could find a 1700 that only plays Sentons. You could find a 1700 that knows how to play most of the common maps.
It's really a load of crap that our tournaments even use global rating for this reason. It's pretty much why I've always used 1v1 rating to judge the WWPC since you have to play a variety of maps and it is therefore a better assessment of game knowledge or skill.
You guys probably don't care about that, though, as I'm getting the hint that you like shiny, big numbers as opposed to what they truly mean.
I'm all for that. Yes, if you want to be technical, for a statistical sample to be relevant it needs to be randomised. Global rating isn't randomised because the sample is determined by which maps the player chooses to host or play. There was however that annoying tendancy for players to play a lot of unranked matches - thus allowing for their rank to be determined by a statistically small and favourable game sample. The rank it all approach was designed to counter that.
The problem that I am seeking to address here is that not only is Global Rank a measure of legitimate game outcomes. It is also a measure of desyncs and crashes. By allowing for a game to be drawn, we can limit the sample of ranked games to those which are actually played to completion and have legitimate outcomes. It is immune to abuse because it requires the agreement of both teams.Statistics: Posted by Hawkei — 01 Jul 2016, 03:15
]]>
Either way you look at it, having an incorrect rating isn't fun for anyone.