Over the last few weeks, I’ve received a number of enquiries as to why our stats differ to those of TSN or GW Metawatch. So, to try and answer those questions, I thought what better way, than to go in-depth in how we put our stats together and the data that builds those stats.
A Note About Data Analysis
Firstly, let me point out that when analysing data, there is no single correct way of doing it. While methods may differ slightly between analysts, it does not mean one set of data is any less valuable than another. Completely the opposite, in fact. Having another method/analyst that then corroborates your own data is only ever a good thing. Having only one set of data or analysis opens that data set up to even more questions about its validity.
With that’s out the way, let’s jump in.
What’s Included in the Data?
The data we’re looking at is for all ‘singles’ tournaments under the new GHB. Teams tournaments, doubles, and single day events are not included in either our own data or that of TSN. Simple because we both look at the cutting edge, competitive side of Age of Sigmar. If you include one dayers (often referred to as RTT’s), then you may skew your data. This is due to RTT’s being treated by many players as more of a casual tournament or a way to test a list to see if it requires tweaking for a bigger GT event.
GW’s Data Set
GW Metawatch use all the data they are able to scrape from tournament websites such as Best Coast Pairing (BCP), Punpun.nl, and others. They have stated that they include 35,000 individual games recorded in their database and that 200 events are added each month.
The number of events added can only mean that the data they are gathering includes RTT’s and also any other events such as Teams, doubles, and even single matches.
As I said before, there’s no incorrect method. Games Workshop is trying to balance a game system. This balancing should be for everyone and not the minority that attends Grand Tournaments. Therefore, it makes sense for GW to gather as much data as possible.
However, what this does mean is that our data and that of TSN are not really comparable to GW. So, let’s focus on ourselves and TSN.
The TSN Data Set
The latest stats from Ziggy and his team state they have included the following 19 tournaments in February:
And the folllowing in January:
Bear in mind that not all of these events in January are under the new GHB. In fact, on the next slide of the TSN stats, we can see they include 23 tournaments as of the date we are looking at as a cut-off:
Where Ziggy and the team at TSN really shine is giving a totally indepth analysis of matches and battleplans used. Looking at each factions performance against any of the other factions, as well as calculating who would have priority in the first battle round based on the drop counts of the lists. We simply can’t do that due to time constraints, so this is a fantastic source of information if you want a truly deep dive into a factions performance against other factions and on each battle plan.
The Woehammer Data Set
Up until the cut-off supplied about (3rd March), we have gathered information on 26 tournaments. These are:
GT Wargame Garrison Madrid
Østjysk Mesterskab I terningehøjdekast
Rise of the Champions
Quest of Champions – Heat 1
Bloodshed in the Shires – 2023
Columbus Brewhammer ’23
DaBoyz Golden Sprue GT 2023
War of the Ospreys
Realm of Geddon 2023
Gods of War 3
Great South Waaagh 2023
War of the Spider God
Midwest Bash 2 AoS GT
AoS Alliance Open Masters GT
Battle at the Brook
Sydney Salt Smash
The Lone Star Grand Tournament
Battle of Copenhagen
Warhammer Age of Sigmar: Matched Play Event
Small Town Throwdown: Smoke on the Water
AOSFF Solo 2022-2023 – Finale
As such, our dataset is slightly larger than that of TSN. As such, this alone will cause a slight discrepancy in the stats between the two sites. The tournaments that aren’t included in TSN are:
Battle of Copenhagen
Sydney Salt Smash
AOSFF Solo 2022-2023 – Finale
These events involved an additional 94 player entries. However, to truly compare our data set, I will remove these from our data temporarily.
TSN Methodolody and Example
TSN methodology is to count wins as a % of the factions results. (We ourselves include 0.5 of draws/ties as well). This gives them the following win rates for the factions over those 23 tournaments:
Let’s take one faction and test the methodology. In this case, we’ll take Idoneth Deepkin, who TSN has at 53% in the table above. Across those 23 tournaments, there were a total of 25 players, playing Idoneth:
Those players totalled 123 results between them. Of those 123 results, 65 were victories, and 6 were ties. To get to TSN’s 53%, we’ll take those 65 victories and divide them by the total number of results (123).
65 / 123 = 0.5285
To get our percentage, simply multiply this value and stick a percentage sign on the end:
0.5285 x 100 = 52.85%
You’ll notice that they’ve rounded up to 53%. So perfect, our maths match!
Now let’s look at the same results, calculated using the Woehammer method:
Woehammer Methodology and Example
The way we calculate our win rates is to split draws between wins and losses. So a single draw (or Tie if you’re across the pond) would be 0.5 of a win and 0.5 of a loss.
So our win rates for the same period and the same data set are as follows:
So, going back to our Idoneth data set, we have 65 wins and 6 ties. These 6 ties will be split evenly between the wins and losses:
6 / 2 = 3
We’ll add that 3 to the 65 victories giving us a total of 68 positive results for the Idoneth.
We then use the same maths, this time taking 68 as our value and dividing it by the 123 results:
68 / 123 = 0.5528
Again, multiply this by 100 and stick a % sign on the end:
0.5528 x 100 = 55.28%
The effect of adding those draws to the faction result is a total of 2.43% added to the win rate.
Which Method is Correct?
Both. Simples. An argument can be made that draws aren’t victories. But the counterargument to that is that they’re also not losses.
Basically, I (and I’m sure Ziggy and his team would agree) would urge you to look at both data sets and draw your own conclusions as to where yu believe the faction to lie in terms of win rates.
Hopefully, this has gone some way to explaining the variations you can find between different stat providers.
If you want to know more, then please feel free to contact me either on Discord, Twitter, or email (firstname.lastname@example.org), and I’ll be more than happy to discuss statistical analysis all day long.
My future task is to include a statistical discrepancy within our stats. Which, when I do, I will show how this is calculated.
But why not have a play with the data yourself? The file is below. If you have Excel, open it up and have a play. Maybe you’ll create yet another methodology to use!
Either, way I hop you’ve enjoyed this war and peace analysis of datasets! Peace Out!