It’s been a long, weird season, but it’s almost over. Only six more Overwatch League games remain in 2020, and the team that comes out on top will be crowned the 2020 OWL champion! And the team with the best odds of doing that is …

Well, it’s obviously Shanghai, right? Owners of a “real” 19-2 regular season record, the Dragons dominated the tougher APAC region. According to advanced metrics, the Dragons had an amazing 2.69 SRS and 0.854 SAMP, which were light-years ahead of the #2 teams in those categories: Guangzhou and San Francisco.

Wait … Guangzhou?

That’s the dilemma this heavily altered season has provided us. Before COVID-19 prompted a rearrangement of the schedule, the two teams that weren’t really Asian but wound up in the APAC region — New York and London — did very well against their North American counterparts. New York was 4-1 and London 3-1 — half of the Spitfire’s 2020 wins — and that appears to have skewed the strength of those two teams immensely, as compared to their true North American counterparts. As a result, SRS and SAMP rankings have very heavily favored the APAC teams. That wasn’t too much of a factor when they were only playing against each other, but now that the two regions are about to cross over, it’s a problem.

As they exist now, here are the SRSes and SAMPs of the final four teams:

Shanghai: 2.69, 0.854
San Francisco: 1.16, 0.654
Seoul: 1.06, 0.644
Philadelphia: 1.03, 0.635

Does anyone believe that Seoul and Philadelphia are roughly the same? Or that Shanghai is that far ahead of the rest of the field? If I went with these rankings as they are above, that would give Shanghai a 91.7% chance of winning against Philadelphia while San Francisco would have just a 52.1% chance against Seoul. That just feels off, and seems like it overvalues the APAC strength of schedule.

That can easily be seen if you sort by SOS in the 2020 standings chart. The top seven entries are all APAC teams, and they range from 0.69 to 1.11. The 13 NA teams’ strength of schedule ranges from -0.91 to -0.48. Yeah, that’s screwy.

I’ve always upheld SRS, and by extension, SAMP, as “true” measures of a team’s value, in that they represent actual values, notably how much a team is better than average or another team, rather than an arbitrary measure like Elo. However, I think something needs to be done to normalize these values a bit, even if that means varying from the actual results of the matches.

I examined a few possibilities before settling on the following. I took the averages of the SOS for both the APAC (0.93) and NA (-0.67) teams. I then subtracted that value from their initial SOS to come up with a new SOS, which I then applied as usual to get SRS (= MOV + SOS) and used that value to compute SAMP. The key thing I wanted to preserve with any measure was that, in a normal season, these modifications should have no effect. That would be the case here, since normally SOS would average out to zero, and so subtracting that from a team’s “real” SOS would have no effect.

Here are the results of those computations, for all league teams:

[table “129” not found /]

I included Map Win % because SAMP should be reasonably close to that, and you can see that’s the case here. Overall, SOSes via this new method range from -0.25 to 0.19. That’s reasonably close to what it was in previous, “normal” seasons (-0.30 to 0.27 in 2019 and -0.19 to 0.27 in 2018), which is another good sign.

Are these ratings perfect? No, definitely not. I probably should have pulled those early London/New York games from the computations and maybe found a way to incorporate “Old Vancouver” into the rankings somehow. And maybe this system actually overvalues NA teams a bit, because New York and Guangzhou slide below Florida and Paris.

But I think these are at least a little closer to reality. As you can see, San Francisco, Shanghai, and Philadelphia are much closer to each other, while Seoul is probably about where it should be, only slightly above average.

I’ll use these modified values to come up with my final predictions for the season, which I’ll post tomorrow. And hey, maybe after we’ve had a few more matches between NA and APAC teams, the “real” ratings will look a little more acceptable. We’ll know that for sure after this weekend.