Curse or Coincidence? Analyzing the 90% Fan Vote in LCS
It’s Day 2 of the first LCS week. Dignitas faces off against Cloud9 in the second game. After a cautious start, a double lane-swap occurs with LemonNation on Thresh and Sneaky on Ashe, pushing the top-lane against KiWiKiD’s Kalista and CoreJJ’s Nautilus. Cloud9 plays aggressively, with LemonNation already flashing to avoid an early gank from Azingy on Zac. KiWiKiD hooks Sneaky, CoreJJ finishes the job, and Azingy secures First Blood. Dignitas takes an early lead.
Cloud9 ultimately loses the game, but not because of First Blood. Popular belief suggests that teams with over 90% of the Twitter vote, conducted live by Riot among the fans, are cursed and will lose.
So, I looked at fan votes from the 2015 Spring LCS in NA and EU. In 180 games, only 24 (13%) had a team with over 90% of the vote. Out of those, the 90% curse struck four times.
However, it’s important to note that the 90% curse doesn’t guarantee a loss. The curse is supposed to make the team do worse, but it’s hard to define what “worse” means without comparing it to regular fan votes. For example, does a team with 89% of the vote do better than one with 91%?
Looking at the charts comparing different vote percentages, teams that received over 90% of the vote performed better than those with over 80%. However, in Europe, teams with 90% support didn’t perform significantly better than those with 80%, and no team in either region won over 95% of games. Perhaps there is a hint of a curse at work there.
The Accuracy of Fan Votes as Predictions
How accurate are fan votes in predicting game outcomes? To answer this, I used a metric called a Brier Score, which compares predictions (expressed as percentages) and results to measure accuracy.
The fan vote was right 62% of the time in EU and 66% in NA, with Brier scores of about .23 and .22 in each region. To put that into context, a perfect score is .25, assuming every game is a 50/50 chance. Therefore, fan votes are somewhat informative but not extremely accurate.
In comparison, predictions based on teams’ accumulated points were almost as accurate as the fan vote.
NA was more predictable than EU in the 2015 Spring LCS. However, there were also differences in predictability among the teams. For example, Coast was the most predictable team, while the Unicorns were the least predictable. Fan predictions about the Unicorns would have been just as accurate with a coin toss.
Some teams consistently surprised fans with their performance, such as Impulse and H2K. On the other hand, fans held onto their memories of Alliance when thinking about Elements.