Northern League R4

[Cantabrigian]

https://blog.vexillia.me.uk/2022/09/adlg-v4-super-armies.html
https://blog.vexillia.me.uk/2022/09/adlg-size-matters.html
Live dashboard

Fascinating stuff, thanks! 

I think that a lot of what you're seeing is regression to the mean - with small numbers of samples you see a lot more random variation.  For example, take your V3 chart, and put your finger on it to cover up any results with less than 125 games, and the relationship looks a lot less clear.

[vexillia]

I think that a lot of what you're seeing is regression to the mean - with small numbers of samples you see a lot more random variation.  For example, take your V3 chart, and put your finger on it to cover up any results with less than 125 games, and the relationship looks a lot less clear.

I agree.  I made the same point myself, many times.  With a small number of games (and probably players and opponents) the average is all over the place.

The important thing is that the mean is 50% suggestive of few, if any, anomalies in the armies once player & opponent factors have been smoothed out.

[Cantabrigian]

I tried to get over to the regression to the mean thing by separating the armies into buckets depending on how many games they'd been used in, then taking the average efficiency for each bucket.

I'm not sure if any meaningful conclusions can be drawn from this graph.

And that's hoping that I got the right data source...

[vexillia]

And that's hoping that I got the right data source...

Looking at the results, you've analysed the combined results from V3 and V4.

I think they belong to separate data sets as the rules changed between versions.

[Cantabrigian]

Looking at the results, you've analysed the combined results from V3 and V4.

I think they belong to separate data sets as the rules changed between versions.

I'd imagine that you're right.