Double Down: Upcomer's Worlds 2021 Knockout Stage Simulation
Close Menu

Hit enter to search or ESC to close

As we approach the knockout stage of the League of Legends World Championship, it’s time for everybody’s favorite game: Wildly speculating about what will happen. Upcomer’s staff have given their view of which team they think will win, and their years of esports writing experience have led them to eight different answers.

May as well add a ninth, from the point of view of the model I’ve built at After generating predictions for each possible matchup, I simulated the knockout bracket 50000 times to see how likely each team was to win.

The world champion

You’d scroll down for this first anyway, so here’s the result:

Team Probability of winning
Royal Never Give Up 28.64%
DWG KIA 17.29%
Edward Gaming 16.54%
Gen.G 11.25%
T1 9.95%
Hanwha Life Esports 5.96%
Cloud9 5.72%
MAD Lions 4.65%

Most people believe that DWG KIA are the strongest team here, but my model’s giving Royal Never Give Up an edge here. Before Worlds had started, I had used a similar simulation-based approach to generate power rankings, and RNG were my overall top team going into the tournament. DK’s 6-0 group stage run makes RNG’s 4-2 look pretty mediocre, but six games against each team’s long history of professional play did not shift the model by much. Public perception seems overly fixated on these last six games, and there’s no doubt that DK looked absolutely dominant in their matches, but best-of-ones just don’t let teams truly show their potential.

Edward Gaming take the next spot. I find them to be just slightly weaker than RNG in each matchup, but if they can get past the first round they have what it takes to win it all.

The other three Korean teams fill in the next three spots, and the Western teams sit at the bottom of the board. They’ve got a combined 10% chance to take home the Summoner’s Cup, so don’t let up on your hopium supply quite yet, but don’t hold your breath either.

The betting market offers odds on this, so we can compare my model’s predictions to that of the market (and, by extension, general public perception):

Graphic of odds for the final eight Worlds 2021 teams’s offered odds on the winner of Worlds

From these odds, we can back out the market’s view of the probability of each team winning, and normalize those probabilities to remove the vig:

Team Odds Odds-Implied Probability Normalized Probability
DWG KIA 1.95 51.28% 35.07%
Edward Gaming 2.7 37.04% 25.33%
Royal Never Give Up 4.5 22.22% 15.20%
T1 5.5 18.18% 12.43%
Gen.G 14 7.14% 4.88%
Cloud9 27 3.70% 2.50%
Hanwha Life Esports 30 3.33% 2.28%
MAD Lions 30 3.33% 2.28%

The odds are pricing in a 35% chance of a DK win. They’re also giving the bottom teams almost no chance – the bottom four teams have a combined 12% chance, which makes this a pretty lopsided final eight. Conversely, my model is giving the underdogs a 26% chance of flipping the script, so there’s always hope.

What’s also interesting is just how much stronger the market believes T1 is compared to Gen.G. My model gives GEN a slight edge, but places the two teams pretty close in probability, while the odds market jumps from a 12% chance for T1 to a 5% chance for GEN. T1 did take a 3-1 victory the last time these two teams met in the LCK Summer Playoffs, and their group stage play at Worlds was much more convincing, but this still feels like an overreaction to recent events.

Similarly, Edward Gaming have almost twice the chance to win as RNG do. Sure, EDG were China’s top seed, and have done slightly better in the LPL recently, but my model still has its mechanical, unbeating heart set on RNG in that head-to-head (and therefore in the overall bracket as well).

I wouldn’t necessarily suggest betting on these discrepancies, however. If you’ve followed along with my guides, you’ll recall that those unnormalized probabilities add up to more than 100% because of the vig, which is how oddsmakers make their money. For a single match between two teams, the vig is typical around 6%, which means that if you win the bet you’re getting paid out 6% less than what would be fair. Here, the vig is 31.6%, which makes pretty much any bet not worth it from a value perspective. Pick’ems might help scratch that same itch!


That same simulation gives me full bracket predictions, so we can pull out each team’s probability of making it to the final:

Team Probability of Final
Royal Never Give Up 41.85%
DWG KIA 39.17%
T1 28.20%
Edward Gaming 25.59%
Gen.G 20.99%
Hanwha Life Esports 18.53%
MAD Lions 14.10%
Cloud9 11.57%

There are some changes in order here, compared to my first predictions, due to how the bracket is set up. T1 are third most likely to make it to the final, even though they’re only fifth most likely to win overall, because of their potentially easier path to get there. MAD have an easier path to the final than C9 do, but have a tougher opponent if they do get there, so those two teams have swapped spots for this prediction.

RNG and DK are almost equally likely to reach the final, and that’s the most likely outcome. However, there’s still only an 16.29% chance that the final is exactly those two teams.

Luckily for us, bookmakers offer odds on this as well:

odds for the final 8 Worlds 2021 teams to make the finals’s offered odds on the Worlds finalists

Once again, we can convert this into the general perception of each team’s chance to be a finalist:

Team Odds Odds-Implied Probability Normalized Probability
DWG KIA 1.28 78.12% 68.93%
Edward Gaming 2.2 45.45% 40.11%
Royal Never Give Up 2.5 40% 35.29%
T1 4 25% 22.06%
Gen.G 5 20% 17.65%
Cloud9 15 6.67% 5.88%
MAD Lions 15 6.67% 5.88%
Hanwha Life Esports 21 4.76% 4.20%

Almost a 70% chance of seeing DK in the finals! That seems… absurdly high? I mean sure, they’re pretty heavily favored against MAD Lions, and favored over T1 as well, but I wouldn’t have even put just the DK/T1 matchup at 70/30 on its own. A team would have to have two 83/17 matchups in a row to have a 68.93% chance of winning both of those, and while that maybe sounds plausible against MAD (though still quite aggressive), there’s just no universe in which T1 have only a 17% chance against DK.

EDG and RNG feel a bit more reasonable, though these odds are still effectively pricing in a 75% edge for either team against GEN in the semifinal. Also, the relative closeness of the probabilities for these two teams to make it to the final imply a relatively even matchup between the two – the biggest hurdle for either team is the fact that they face each other in round one. This seems at odds (no pun intended) with their market-implied probabilities of winning outright. They have similar chances of making it to the final, but EDG have a significantly better chance of winning (presumably against DK) from there than RNG do? It’s possible since teams can have different strengths against each other, and so on, but it definitely feels out of sync.

These probabilities are ultimately a reflection of bettors’ aggregate sentiment, so perfect rationality isn’t really a realistic expectation. 70% on DWG KIA still feels insane with any amount of caveats.

The vig on this set of odds is surprisingly tame: 11.8%. It’s still about twice as high as it should be for a typical match, but the lower uncertainty of having to predict just one of the two teams in the final compared to picking the winner means you don’t get raked quite as hard by the oddsmaker. Same disclaimer applies though, finding value might be tough.

Placement distributions

For good measure, here is each team’s probability of finishing in any particular place, according to my model:

graphs for the eight teams in Worlds 2021 knockout stage
Alacrity’s Worlds Knockout Placement Distributions

Each team is more likely to take 5th-8th than any other specific place. This sounds weird, but actually makes sense. No team is guaranteed to make it past that round, and even in a favored matchup the probability of just losing in round one is higher than any other specific sequence of wins and losses that would be required for another place. For example, to get second, a team needs to win, then win, then lose, in exactly that order.

Vanya is the founder of, where he leverages seven years of quantitative finance experience to build esports prediction models. He'll happily debate you on anything from math to video games to the impending subjugation of the human race by artificial intelligence.