For more explanation as to how I derived these rankings, see my explanation here. For more information about these Elo rankings or why I am confident that they are generally predictive, see my most recent Metagame Fortnight. To thank NoxvilleZA for honing and curating this Elo, or to ask him questions, follow him on twitter. While you're at it, follow me, too.
- Gorgon @theWonderCow
Note:
These are odds, not predictions. I did not change any figures based on non-data factors (see my previous post for a description of the limitations of this model). We say the odds were predictive if the outcome with the best odds happens. If the best-odds scenario does not happen for all matches but does happen for most matches, this is to be expected.
As I said in the first article, statistics are not a crystal ball, but they are a powerful tool for informing predictions.Because these regions are so close, the odds of any projected outcome for either of these regions are less than the odds of all other possible outcomes. Below are simply the most likely of very close possibilities. Expect a lot of upsets and contention in this region!
Unfortunately, SEA has some tournaments for which data is not automatically tracked at Datdota, my primary resource, so the SEA odds are the best with available data. There may be a couple of dozen games played which are not represented by these models. Models can only be built with the best data available.
NOTE: where I used "less than 39" I meant "less than 40." My apologies for the error.
Postscript: Note that because each match has three potential outcomes (2-0, 1-1, 0-2), if the most likely outcome is below 39% likely, it must have less than a 9% projective advantage over all other options. That is a small margin of success, and we'd expect a lot of "upsets" in these matches which are really too close to call with confidence. That being said, nearly 66% of these low-odds games were correctly predicted in the previous groupstages.
Comments