EPTOn November 28 2011 10:33 Teddyman wrote:
It's just inflation. When a player retires, his rating is removed from TLPD. If the rating of a retiring player is low enough, the average rating of the active player pool will increase with his retirement. As games are played, this extra rating will eventually end up in the hands of the top players. Naturally there are many more retired BW players than retired SC2 players, so the curve will be steeper in BW.
It's just inflation. When a player retires, his rating is removed from TLPD. If the rating of a retiring player is low enough, the average rating of the active player pool will increase with his retirement. As games are played, this extra rating will eventually end up in the hands of the top players. Naturally there are many more retired BW players than retired SC2 players, so the curve will be steeper in BW.
Isnt that a zero sum course of events? In order for someone to lose, someone else has to win. And in the end - both players will quit, the winners and the losers. The winners might have longer careers with more overall games, but to balance that off, there must be more losers than winners.
![[image loading]](http://i.imgur.com/t3Cib.png)
The international SC2 ranking does appear to have what could be called ELO inflation, but if one normalizes the SC2 ranking data to the same range of SC2 korean/BW korean, then it could be comparable. The SC2 international appears to have an overall smaller dynamic range than both the other lists.
On November 28 2011 10:56 Takkara wrote:
One major variable that is not being adjusted for in this "analysis" is the effect of the different league styles on the data. For BW we have PL, MSL, and OSL. For SC2 we have mostly GSL. The structure of PL in particular is vastly different than anything that exists in SC2. Because the ways that the gamers are amassing ELO is different, we can't directly compare the relative values of high and low ELO and assume the two games will be the same if they are equally skilled.
There's so much noise in the "data" that you can't draw any conclusions at all from what you've presented. Not the least of which you need to prove that there is a connection between larger gaps in ELO between pro players and the skill required to play the game. There's so many other, logical explanations both for why such a gap would exist and why it would not exist that it doesn't clearly follow from your chart.
Secondly you need to find a way to normalize the ELO between the two games to turn your apples and oranges comparison into an apples to apples comparison. Of course, finding this normalization is probably not easy and may not even be solvable.
Thirdly, you'd be more intellectually honest to compare such ELO data 5-10 years ago for BW to see what, if any, effects time has had on such ELO graphs with just BW. It would be important to note if BW has gotten more distinguished, less, or stayed relatively the same.
There's other issues, but this is a good start if you honestly want to try to prove something with your numbers.
One major variable that is not being adjusted for in this "analysis" is the effect of the different league styles on the data. For BW we have PL, MSL, and OSL. For SC2 we have mostly GSL. The structure of PL in particular is vastly different than anything that exists in SC2. Because the ways that the gamers are amassing ELO is different, we can't directly compare the relative values of high and low ELO and assume the two games will be the same if they are equally skilled.
There's so much noise in the "data" that you can't draw any conclusions at all from what you've presented. Not the least of which you need to prove that there is a connection between larger gaps in ELO between pro players and the skill required to play the game. There's so many other, logical explanations both for why such a gap would exist and why it would not exist that it doesn't clearly follow from your chart.
Secondly you need to find a way to normalize the ELO between the two games to turn your apples and oranges comparison into an apples to apples comparison. Of course, finding this normalization is probably not easy and may not even be solvable.
Thirdly, you'd be more intellectually honest to compare such ELO data 5-10 years ago for BW to see what, if any, effects time has had on such ELO graphs with just BW. It would be important to note if BW has gotten more distinguished, less, or stayed relatively the same.
There's other issues, but this is a good start if you honestly want to try to prove something with your numbers.
Doesnt the "noise" you describe actually define what we see in the games as skill? Because that is what the players have to do, their mission to play in the conditions that the league presents them. And those who are the most successful at that, could be described as being the most skillful. IE Bisu being very skillful as he plays superbly in proleague despite lackluster individual league games.
One can solve the normalization problem by looking at the actual dynamic range of the Elo rankings, rather than the absolute values. Nonetheless, this appears to not be an important issue as the BW and SC2 korean charts appear to be in the same range of absolute values.
The age of the game and metagame development differences may be a relevant argument, but it is so hard to quantify in numbers that i cant imagine how one would include it in the analysis
