Wednesday, 14 September 2022

ADLG: V4 Super Armies?

I plan to investigate this thorny question in detail, but not for another 6-9 months.  I'll explain the reason for the delay later.  However, I thought you'd like to see an early analysis.  But first some background.

The answer under V3?
In 2018 and 2020 I used the Art de la Guerre (ADLG) results database to probe the perennial question of super armies.  In early 2021 I used the same data to look at the performance of the Yuan Chinese following their success at the World Championships from 2016 to 2019.

The above chart shows that the more an army is used, the more it's performance (efficacy in ADLG-speak) tends towards 50%.  So in short, the answer is no: there aren't any super armies in ADLG.  Remember, by super armies I mean armies that will have a high degree of success regardless of who is using it.

Will things be different under V4?
In April 2021 version four (V4) of ADLG was released with a few changes (including new troop types), some new army lists, and some list revisions.  This raised a separate question: are there super armies in V4?

To re-examine this question, all I need is sufficient results from tournament games played under V4.   Sadly, the publicly available database doesn't distinguish between games played under different versions, so I couldn't just download some data and repeat my earlier analysis.

In my view, the release of V4 marks a clear break point and the two sets of results shouldn't be mixed; too much has changed.  So I set about writing a spreadsheet that would separate the results under V4 from those under V3.  Having done so, all that remained was to be patient.

Size matters
Why patient?  As of today, there simply aren't enough V4 results for the pattern to be both clear & stable.  Small data sets are all too easily skewed by extreme (one-off) results.  To provide a stable analysis the V4 data set must be larger; ideally as large as the V3 data set.

When I first performed the analysis, in the early summer, there were 7,323 V4 games recorded.  When I checked earlier today there were 8,675 games recorded.  However, the current V4 data set is only a little over a fifth of the V3 data set (40,545 games), and needs to be much bigger.  To me, this makes today's data set only just useable but only with caution.

Early results
The chart below shows the current V4 data.  It's in the same format as the V3 data above, but the horizontal scale is different (smaller).

What does the data say about super armies under V4?  Basically, it looks like the pattern under V4 is going to be the same as under V3 (i.e. there are no super armies).

As expected, the sharp pattern seen in V3 is not present.  Nonetheless, the same pattern looks like it's emerging.  I believe the broader distribution is simply due to the smaller size of the V4 data set, but it could be due to the changes made in V4.  Time, and more data, will tell if I am right about this.

Next steps
The ADLG results database is growing very rapidly.  So, as I said at the start, I will return to this issue in 6-9 months when there should be more than enough V4 data to make a detailed, direct, and meaningful comparison with the situation under V3.

1 comment :

Vexillia said...

Corrected the figure for the total number of games recorded under V3 to 40,545 (was 15,441) and edited text accordingly. This means even more caution is required as V4 games are currently a little over 21% of those recorded under V3.