Tweet
Re: Reasons to believe there's FIDE rating "inflation"?
I'm not sure you actually addressed what I was commenting on in the previous graph.
What I noticed was not a comparison between any of the actual results curves and the expected results curve. It was that the actual results curve for the 2400s looked steeper on both sides of 0 than the actual results curves for the other three groups you plotted.
As I look at it, your data shows the 2400s doing worse against higher rated opponents than the other three groups did, and doing better against lower rated opponents than the other three groups.
Is this not what the graph shows? Isn't the purple line mostly above the other three on the left and mostly below them on the right?
As you say, there's a lot of noise in the data -- there can't be that many data points for CFC 2400s playing opponents rated 250+ points higher, or for CFC 800s playing opponents 200+ points lower.
I suggested graphing the results of 2000s and 2300s because I expected there'd be more data points and less noise. But a graph that averages all results and plots it against the expected results curve does not allow a comparison of players from different K factors.
I'm not sure you actually addressed what I was commenting on in the previous graph.
What I noticed was not a comparison between any of the actual results curves and the expected results curve. It was that the actual results curve for the 2400s looked steeper on both sides of 0 than the actual results curves for the other three groups you plotted.
As I look at it, your data shows the 2400s doing worse against higher rated opponents than the other three groups did, and doing better against lower rated opponents than the other three groups.
Is this not what the graph shows? Isn't the purple line mostly above the other three on the left and mostly below them on the right?
As you say, there's a lot of noise in the data -- there can't be that many data points for CFC 2400s playing opponents rated 250+ points higher, or for CFC 800s playing opponents 200+ points lower.
I suggested graphing the results of 2000s and 2300s because I expected there'd be more data points and less noise. But a graph that averages all results and plots it against the expected results curve does not allow a comparison of players from different K factors.
Comment