In my previous article, On Rating HEMA Tournaments, I proposed the PB25 algorithm for getting an overall rating of the quality of competition of a given HEMA tournament. This metric is overall not especially interesting to people who aren’t stats nerds. But what is interesting is that once there is a usable metric we can start doing tournament comparisons.
The first thing everyone is going to want to know is “which tournament is the highest rated?”. While fun, this is also not a particularly helpful question to gain insight into HEMA tournaments. It’s well known that HEMA Ratings have island effects, both high and low*. So even if you have a number from two different tournaments you have to think about other factors going into the ratings of the fighters attending each. (Also you should know without me having to calculate anything that the highest rated tournament is the 2019 European Games.)
*While most of us are familiar with island effects that over-inflate rating in relation to skill, you can also get islands that are under-rated in relation to skills. However no one ever gets mad that someone else is rated worse than they are, so people don’t tend to talk about that.
What is much more interesting is to compare trends over time, as this gives us a tool to see how the face of competitive HEMA is changing.
North America, 2017-2020: Fragmentation
- As of 2017, Longpoint 2017 was the largest and highest rated event in North America.
- As of 2018, Longpoint 2017 was the largest and highest rated event in North America.
- As of 2019, Longpoint 2017 was the largest and highest rated event in North America.
- As of 2020, Longpoint 2017 was the largest and highest rated event in North America.
The scope of Longpoint in 2017 compared to everything else in pre-Covid North American HEMA is staggering. Longpoint 2017 is shown in red, and other Longpoint events are in orange.
Clearly we can see how cornerstone an event Longpoint was to high level North American HEMA, if you lop the red and orange out from the graph above it becomes a very different story.
And that is why I call this section “fragmentation”. For those of you in the audience who weren’t around in 2017, this is when the Longpoint team called a hiatus on the event due to such deadly opponents such as burnout and life changes. The event ran once more in 2019 with a reduced scope, but more or less nothing in the intervening years filled its place as “the” HEMA event for North America.
The plot above probably matches the intuition of most who have been in the North American HEMA scene since the 2010s, but it’s good to see born out by evidence. (protip: many things you feel confident about without proof are not actually true.)
Europe Post Covid: Death of Best-On-Best Tournaments
If you’ve ever wondered if the Euro vs NA talk is justified, let’s have a look at the comparison.
Only 3 tournaments in the top 16. But this is also not exactly an interesting finding, people who travel between scenes were pretty well aware of the discrepancy.
What is more interesting is the huge decline in the rating of the top European longsword tournaments. Below are the top European Longsword tournaments from the pre and post Covid eras.
This is a staggering drop of almost 100 points across the board.
Which brings up an interesting question: what is driving this? The only possible explanations are that the overall number of high level fencers has taken a calamitous drop, or that we don’t see concentrations of all the best fighters at European events like we used to.
A cursory glance at the top fencers in HEMA Ratings will show you they are just as full on the top end as they ever are. Which leads us to the only inevitable conclusion: we just don’t see best-on-best events in Europe like we did pre-covid.
Post-Covid North America: Growth Of Regionals
After taking a look at how the pre and post Covid trends have shaken out in Europe, have we seen a similar effect in North America? Already heading into Covid the North American scene was increasingly regional, without anything new to attract the best-on-best in a way that Longpoint used to.
What we actually see is a reverse of the European trend, with the overall HEMA scene coming out of Covid stronger than it went in. Is this because we are seeing more people competing cross-regionally, picking up the gap left by Longpoint? The funny thing is, if you look at tournament medalists you see that if anything North America is more regional than it ever has been. Meaning that in general the regions are stronger than they have ever been.
Europe vs North America: Inflection Point?
So, with the overall ratings of European tournaments trending down, and North American tournaments trending up, where does that put us?
In a surprisingly balanced place. While before the top was overwhelmingly Euro biased, now the distribution is 50:50 split. Looking at the plot from earlier, we see the distribution of events.
This data shows that the post-covid tournament scenes in North America and Europe have actually started to converge. This is being driven by:
- New growth in North America, with regional tournaments now as high level as continental tournaments were pre-covid.
- No new tournaments picking up a true continental mantle in Europe, leading to an overall decline in level of competition at the highest level events.
What conclusions can I draw from this? Though data shows the competitive scene in North America has “caught up” to Europe, this is largely due to a change in attendance patterns at European events. And as such I would caution anyone from North America who hasn’t competed overseas against getting too excited.
At the same time, there is extremely strong regional growth in North America, and if there were true best-on-best North American competitions (as there were pre-Covid European ones) the overall picture may be quite different.
For the time being it’s a pretty safe conclusion that the gap between North American and European competitive scenes is narrowing, and one would expect this to contribute to a decreasing skill gap between communities.
Note: the best way to do analysis is to assume all trends will continue forever, and nothing changes. So use this to make some strong claims about what the future holds for each.
Stuff For Nerds
How did I get these numbers? Maybe you should read On Rating HEMA Tournaments. 😛
The only thing not included in that article was how to back-derive deviation from Weighted Rating. The “Rating” that HEMA Rating shows us on the page is “Weighted Rating”, which is Rating – 2 * Deviation. If you’ve looked into how the Glicko-2 algorithm works you’ll know everyone has a Rating, meant to be an estimation of skill and Deviation, meant to measure the uncertainty in that estimation. And to use the Glicko-2 formula to calculate win percentage we need Rating and Deviation for both fencers.
I wanted the PB25 algorithm to work solely on Weighted Rating, as that is what people will see when they open up the page and look at fighters ratings. Which means that rather than use actual deviation scores for individuals, a function to derive it from rating must be determined.
I used the simplest approach: get a data dump of all Weighted Ratings and Deviations, and then do a linear regression. This is only marginally accurate if we want to consider an individual, but that’s ok. What we need it for is as a consistent measure we can apply to every event.
So in the end we arrive at: