Can anyone remember a more insane week of Overwatch? When doors opened on Thursday, the New York Excelsior and Vancouver Titans were on the brink of locking up the Atlantic and Pacific Division titles, and Shanghai had bounced back from their loss to the Titans. The first match of the week resulted in the Spitfire defeating the Charge in typical Spitfire fashion.

Then all hell broke loose. The Toronto Defiant upset the Dragons. Then the Florida Mayhem defeated the Spitfire. By the time the week had ended, the Titans and NYXL had each been swept for the first time in team history by the Washington Justice and Chengdu Hunters, respectively. Chaos, thy name is Stage 4.

But did any of it matter in the end?

It’s a harder question to answer than you’d think.

The Overwatch League was a very top- and bottom-heavy league through the first three stages of the season. This means there are teams that have already locked-in playoff spots (the Titans, NYXL, and San Francisco Shock have already clinched top six); meanwhile, the Defiant, Justice, and Mayhem can play spoiler to those in the playoff hunt, but not much more. For both sets of teams, the answer to the question of “Did this week matter?” is largely… no.

For the rest of the league, the answer is much different. The three remaining guaranteed playoff spots are still wide open, and every non-eliminated, non-clinched team still has a shot—however miniscule—of reaching the play-in gauntlet.

The next question is, just how likely or unlikely are those chances?

To answer this question, I turned to Overwatch League Elo to provide a model with which to simulate the outcomes of the rest of the season and derive likelihoods of playoff placements from the results of that simulation. OLE provides not only a starting point for current team strength in a given meta, but also the framework for how a team might progress through the rest of their matches. The simulation shouldn’t have expected that the Justice would upset the Titans (and it didn’t—the Titans were favored to win in 89% of predicted outcomes), but it should know that in the event that an upset did occur, it should re-evaluate the strength of the Justice in its hypothetical outcomes (and it did: the Justice stole nearly 134 Elo from the Titans in their win).

The example above serves as reason to use OLE as the model for evaluating future match outcomes, rather than flipping a virtual coin. While it is possible that the Boston Uprising could win the rest of their matches and sneak into the top 12, OLE is less optimistic about those chances compared to an un-weighted coin flip at every turn. OLE also has the ability to predict match outcomes using a Monte Carlo simulation, which is essentially a way to brute-force likelihoods by way of simulating thousands (or in this case, tens of thousands) of outcomes and rolling them all up into a percentage of a massive total aggregate.

Applying a Monte Carlo simulation to the entire remaining schedule as well as evaluating tiebreaker scenarios is much more involved than simulating a single match, but this stuff is cool, so I did it just for you all. Introducing: Simulation of Overwatch League Outcome Quantities—or SOLOQ. For this system, 100,000 SOLOQ. outputs will be used to evaluate the likelihood of each team’s season placements at the end of Stage 4.

I secretly turned on SOLOQ at the beginning of this week, which means we can compare how the insanity of Week 2 changed the landscape of the league. Let’s first compare each team’s projected average rank at the start of the week versus where they ended up after one of the most upset-filled weeks the league has ever seen:

Team Projected avg. rank: end of Week 3 Projected avg rank: beginning of Week 3 Gain/loss (lower is better)
Titans 1.24 1.21 0.03
NXYL 2.15 2.14 0.01
Shock 2.6 2.84 -0.24
Gladiators 5.45 5.8 -0.35
Spitfire 5.91 6.24 -0.33
Spark 6.51 5.97 0.54
Dynasty 7.74 8.61 -0.87
Fusion 7.99 9.56 -1.57
Dragons 8.52 7.8 0.72
Reign 9.54 11.33 -1.79
Charge 11.16 10.95 0.21
Hunters 11.29 11.5 -0.21
Valiant 13.1 13.03 0.07
Fuel 13.13 12.4 0.73
Outlaws 15.61 14.31 1.3
Eternal 16.11 14.55 1.56
Uprising 16.03 16.15 -0.12
Defiant 17.13 17.14 -0.01
Justice 18.89 19.22 -0.33
Mayhem 19.87 19.6 0.27

Right away, two things are clear: the top and bottom teams didn’t really move, and two pairs of Atlantic teams may have vastly changed their destinies during the week of chaos. First, the good news: Philadelphia, after losing two close tiebreaker matches in Week 1, recovered to defeat Houston 3-1 and scrape by Toronto 3-2. These two wins bumped their projected average rank (PAR) to 7.99, an improvement of over a rank and a half. Beating Houston also gave them a definitive head-to-head tiebreaker advantage, should the two teams end the season with the same match and map differential. If the pair had split their matchups (Philadelphia previously beat Houston 4-0 in stage 2), they would have had to play a tiebreaker match to determine their final placements!*

*Shanghai and Guangzhou could find themselves in this situation after splitting their two matchups with the same map score.

Speaking of beating Houston, the Reign did that, too, and obliviated any chance Houston had in a potential head-to-head tiebreaker. The Reign are now on a five-match win streak dating back to their homestand and improved their PAR by nearly two ranks in a single week. The final team with big gains was Seoul, who only played a single match against the Valiant and improved by 0.87 PAR—also setting themselves ahead of the Valiant in their season head-to-head record. The Valiant’s PAR did not change drastically because they also played and defeated the Fuel (whose PAR did suffer).

Now the bad news. Improvements in the season outlooks for the Reign and Fusion came at the cost of the Outlaws and Eternal, whose PAR both increased by more than an entire rank. I mentioned before that by virtue of losing to their division rivals, Houston essentially punted two end-of-season potential tiebreakers in a single week. Paris also lost one of their tiebreakers after being swept by the Reign.

Despite the doom and gloom, the Eternal and Outlaws still have a chance. PAR, after all, only tells one small part of the story. Let’s zoom out a bit more and take a top-down view of where each team is most likely to place, given their SOLOQ output:

img1.png
Relative likelihood of each team reaching each unique playoff rank. Bubble size corresponds to percentage probability of reaching rank, and labels indicate average playoff ranking.

Fans of Houston and Paris, after squinting a bit, are probably thinking, “So you’re telling me there’s a chance!” This visualization shows just how uncertain ranks 4-12 still are. While the Gladiators, Spitfire, and Spark are still favored to claim the final top-six spots, the Dynasty, Fusion, Dragons, and surging Reign still have realistic chances of evading the play-in tournament.

Their chances also extend in the other direction, as we approach the play-in line. It’s incredibly likely that the likes of Spark, Seoul, Fusion, etc., will at least make top 12, but there are still non-zero outcomes where they do not. It is in these rare situations where teams like Houston, Boston, and Paris still could find their way into the play-ins, ready to make one last stand for their spot in the season playoffs.

You should now be wondering: how rare are these situations? Below, I’ve visualized the SOLOQ likelihoods for each team to reach the top six and top 12 by the end of the regular season:

img2.png
Likelihood of each team reaching top 12 and top six placement by the end of the regular season.

Your eyes do not deceive you—not only do Boston, Houston, and Paris still have chances to make top 12, Dallas, Chengdu, and Guangzhou actually all can still make the top six! The least likely of the trio, Dallas, reached the top six in 83 out of 100,000 SOLOQ outcomes, good for a 0.1% chance (rounded up). To do so, they needed a hilarious list of things to go their way, not limited to but including:

1. The Spitfire or Gladiators losing all of their remaining matches.

2. The Fuel winning all of their remaining matches and also increasing their -14 map differential above the Spitfire’s +9 or the Gladiators’ +17.

3. The Fusion winning no more than one more match.

4. The Dynasty winning no more than one more match.

5. Either the Spark or the Dragons winning three or more matches (but not both).

Like Dallas and other top-six hopefuls, Paris has a teensy chance to still reach the play-in tournament. Out of 100,000 SOLOQ simulations, the Eternal reached the top 12 in 263 outcomes. Here’s their own partial laundry list of hilarious things that need to go their way to get there:

1. The Eternal winning all of their remaining matches.

2. The Uprising not winning any of their remaining matches.

3. The Hunters not winning any of their remaining matches.

4. The Fuel and Valiant losing three of their remaining matches.

There are alternative paths that involve the Charge losing all of their remaining matches. This the power of SOLOQ: I don’t have to think about every single potential thread the rest of the season could produce. Instead, I just ask SOLOQ politely what it thinks—100,000 times—and by doing so, even the rarest of outcomes are likely to happen at least once. The middle of the league is still up for grabs, and I’m excited to see how the race tightens up as we enter the endgame of the 2019 regular season.

Week 3 of Stage 4 kicks off on Thursday, August 8, when the Spitfire (15-10) take on the Titans (22-2) at 4 p.m. PDT. Watch all 2019 season matches live and on demand on overwatchleague.com, the Overwatch League app, our Twitch channel, MLG.com, and the MLG app.