2022 Worlds Seeding — Rollout the Analysis
Using what we know about Riot’s 2-year international performance metric to check on Worlds Seeding
Off the back of a Twitter discussion earlier today (where I was a little too frustrated at no one knowing about how World seeds were determined — sorry folks, especially Ashley Kang!) I decided to map it out for myself.
Upon a few quick google searches, I realized that there was very little communication from Riot outside of when the seeds changed in 2020.
Although the communication is sparse and buried in the overall 2020 Worlds Format announcement, we do get a glimpse at how the seeds get determined.
“…global power rankings. These rankings look back at the last two years of international tournaments”
While in my tenure at Riot, this system was something the Stats Team(s) helped work on in order to figure out a system to see what Regions got the extra seeds. Rather than taking pure community sentiment and records at face value, it used a series of calculations based on how the Regions performed and took an average of all teams representing that Region to determine a “Power Ranking” for the Regions.
I decided to re-create a similar, although much quicker formula in order to calculate the “Power Ranking” of each Region based on the results of MSI and Worlds in order to show why discussions about losing and/or gaining seeds are often misconstrued in Region bias/fandom more so than actual analysis (and how Results-based analysis is not the best to determine things like this).
The Math & Clarifications
The way I decided to calculate things isn’t that complex — I assigned a value equal to the placing of a Team at MSI or Worlds, and then noted which Region those teams represented at the event.
Things that are not accounted for in this formula include:
- MSI vs Worlds
I decided to keep the placement scaling the same for both events as to not downplay MSI vs Worlds, and it gives MSI victory more weight.
- Weights based on seeding or event recency
Although I do not believe the official method used weighting for seeds from each Region, a placing by the 1st Seed of a Region is just as impactful as the same placing from a 3rd or 4th Seed in our metrics. Likewise, placing in 2021 or 2022 MSI will have similar values (to an extent, IE — 2021 MSI had 10–11th the same value at last place. 2022 MSI had 11th as last, with 9–10th being the same value).
- No jump in points for making it out of Play-Ins
As you will see below, I scaled points higher for making Knockout Stage at International Events. I did not do this for Play-Ins, so there is no larger jump in points for finishing 9th-16th.
I scaled these values with a few criteria:
- More points for making Knockout Stage compared to losing in Group Stage
- More points for making Semifinals compared to Quarterfinals
- More points for making Finals compared to Semifinals
- Most points for winning the entire event
- Values start at 1 for last place, and scale based off placement (IE — 13th gets 1 point, 9–11th gets 2 points, etc)
If the values are low, I can go back and scale things more aggressively later without having to change much. For example, in my first pass winning gets 2 more points than 2nd, which is 2 more points than 3–4th. I can scale it higher so that winning means more.
The Results — Region Power Rankings
Counting 2020 to Present
If we go back to the start of 2020, there are 2 MSI events accounted for compared to the 3 it would take into account normally. Because of this, the general concern from social media is that the VCS may not have their 2 seed slot. However, because the “Power Rankings” are an average…
The LCK and LPL are clearly ahead of the pack and deserving of their 4 seeds. Interestingly here, the LCK is a higher average ranking than the LPL is. If we add some qualitative analysis (IE — expert opinion) to our model, we can infer that on average the LPL is a stronger region than the LCK in the past 2 years. This shows us that our model could use some adjusting, especially towards the top-end of the formula. Similarly, the LCS and LEC seem very close using this model — which can be judged heavily even by just watching how EG vs G2 played out at MSI this year.
Even with just one team performing at 2022 MSI, the VCS hangs on to it’s 2 seed slot vs a surging LCO region (mostly, thanks to a Pentanet performance at 2021 MSI). PCS is also still comfortable with their 2 seed slot, although a lot closer to both 4th and 6th than you would think.
Either way, while this might need tweaking to show more precise rankings, it does show a clear divide between 1–2 vs 3–4, and a smaller but still visible rift between 3–4 vs the rest of the Regions.
Counting 2019 to Present
Let’s assume that because of the 2020 MSI situation that Riot goes back to the start of 2019 to determine “Power Rankings” for the upcoming 2022 Worlds event. How different does that look with another year of competition under our belts?
Again, the LCK and LPL take the top 2 slots of the rankings, and again our qualitative analysis tells us our formula could use some adjusting. The gap between LEC and LCS however, looks much more as to what it should be based on previous event performances. PCS and VCS are also much closer to one another, although both still keep the same placing and their same number of seeds.
The winner here is the LCL, as they bump down the LCO in the rankings and are looking for (although, still further away) a 2nd seed at Worlds events.
Adjusting our values
As alluded to above — sometimes analysis isn’t just numbers and nothing else. Qualitative analysis is when you take opinions from experts in the field and apply them into your statistical data analysis methods.
Looking at our above results and knowing our formula is quite basic, we can adjust the values of just the Winners at MSI and Worlds and we can see that it changes a bunch for our 2019 to Present rankings.
By adding only 2 points to an event win we can see that the LPL is the higher ranked Region (as most experts will currently tell you). It also shows us that the LEC separates itself more from the LCS in the rankings (because of their 2019 MSI victory.)
Overall, this was a fun exercise to try and vocalize what I was feeling this morning when I flip through social media and see discussions about what people think is accurate analysis — but it ends up veiled in bias. I used the term “Region Hate” to describe a bias one has for/against a Region — not as a meaning of hating a specific Region for one reason or another (which is just silly, and you shouldn’t do that).
Overall, there is a clear cut between the Top 4 Regions when it comes to international results and the other Regions, which is what I was trying to vocalize earlier (albeit, poorly). And although the method Riot uses is more precise and can account for some more factors, it more-or-less should sync up to what we’ve uncovered in this analysis article.
We’ve still got a few months of games ahead of us before 2022 Worlds kicks off — what teams and regions will you be rooting for?
If you enjoyed this analysis, consider following me here on Medium to be updated on my latest posts! You can also tune in to my Twitter account or my Ko-Fi account for the most up-to-date on what I’m doing, and check out my Tableau Public Profile to enjoy consistently updated LoL Esports visualizations for free!