ADLG: Are There Super Armies? (2020)

In 2018, my peak ADLG period, I analysed all the tournament data available from the Art de la Guerre (ADLG) website. I was looking for quantitative evidence for so called super armies. In short, I didn't find any as this 2018 chart showed:

It's been two years since I last looked at the data so I thought I'd see if anything had changed.  The first thing to report is that ADLG has remained an active tournament rule set.  The latest data set contains 40,545 entries up from 27,670 in 2018.

Please bear in mind what follows is an analysis of the full data set available as of July 2020. The data set includes that used in 2018 plus more recent results.  When comparing 2018 with 2020 I am comparing the data sets available that year not the data for just 2018 and 2020.

1.1 Super Armies
The ADLG tournament database provides an efficacy rating for each army. It covers 271 armies of the 283 provided in the rules: 12 haven't been used yet.  This represents an increase of 9 from 2018.

To calculate the rating each army is awarded 1 point for every victory and half a point for a draw. Finally, the rating is expressed as a percentage of the total games played.

This chart answers the question posed in the post title:

As before the distribution of the armies' efficacy rating is strongly grouped around 50% however the proportion of armies rated between 41 and 60% has increased from 73% (2018) to 83% (2020).

There are far fewer armies with outlying poor or high efficiency armies and nearly all such outlying armies have been seldom used (see below).

This trend strongly suggests that players have become somewhat more proficient in the intervening two years either through better understanding of the rules; better game play or simply playing more games.

The effect of the number of times an army is used, or its popularity, is shown here:

The pattern is clear: the more an army is used the closer the efficacy gets to 50%.

When an army has only been used a few times its rating is very susceptible to random factors, not least luck, but as usage increases these factors even out and a truer rating emerges.

The bell shaped, or normal, distribution strongly suggests there are no super armies i.e. armies that succeed whoever is in charge.

1.2 Efficacy & Period Choice
This has changed little since 2018.  The major changes (if that) are increases in average efficacy for Ancients (up 1.40%) and Dark Ages (up 0.80%).  Feudal armies decreased slightly (down 0.90%).

Given that the 2020 data set includes the 2018 data set the above changes may be more significant than they first appear and stem from a significant trend in 2018 to 2020. Proving this requires a more detailed analysis of the differences, if any, between the two data sets. Perhaps a step too far.

I did have a quick look at the data set containing just the 12,875 results from 2018 to 2020: the data looked remarkably similar with perhaps a broader spread probably due to the lower number of games in this data set. Even so, I couldn't spot anything that would explain the change above.

1.3 Favourite Periods
Compared to 2018 data set there has been a shift with a slight fall in the popularity of Classical and Roman armies and a corresponding increase in Dark Age, Feudal and Late Medieval armies. The top ten armies are the same with some minor movements.

As in 2018, some care is needed in interpreting this data as themed tournaments could skew player preference.