Boards›Esports

How an engineering student predicts LCS winners

All Access 400·4/14/2015, 5:53:46 AM·3 votes·1,179 views

Hey guys. I'm an Korean engineering student who loves e-sports, and I wanna share an idea with u guys

When I watch matches, the broadcasters say that good teams are good at snowballing their advantages. They show gold/min for teams and players, but I don't think the gold/min does not directly represent ones' or teams' snowballing ability.

Then, I came up with this idea: what if I make scatter plots on time vs. gold for each teams' matches? What if I derive equations from the plots? What if I take double derivatives of the equations just like disposition -> velocity -> acceleration learned in physics?

So, I actually tested the idea with NA & EU LCS playoffs semi-finals took last week.

  1. I made scatter plots (time vs. team gold) for each sets of a series.
  2. I drew trendlines (thanks to Excel) in quadratic equation to get constants as results of double derivatives rather than different functions. Interestingly, the R^2 values are very close to 1.00; the equations and plots are highly correlated.
  3. After deriving the constants(I will call these constants "gold acceleration"), I calculated the averages for each teams' performance in semi-finals.

I don't major in math or stats, so there may be some mathematicians who have better ideas, but I'm pretty confident that these numbers are meaningful in judging teams' performance.

According the numbers I got, Fnatic and C9 will win this LCS playoffs. (check the attached files)

Let's see how it goes next week.

Please share your thoughts with me. (twitter.com/kaist_Q)

7 Comments

Ixionas4/14/2015, 3:46:54 PM1 votes

It seems to me that the assumptions of independent observations are violated, so can you say that this data has any significance?

FarRockBF4/14/2015, 3:54:08 PM1 votes
  1. Not enough sample data
  2. Each player plays significantly differently with different champions
  3. Gold income is very strongly correlated to which team they are facing

These all make over 50% of your underlying data meaningless.

~ A financial engineering graduate

xJLx MCHammer4/14/2015, 8:58:38 PM1 votes

Can I ask OP, does stating you are an engineer or a Korean engineer make a difference to this thread? As far as your prediction goes, no.

GundayMonday4/14/2015, 9:08:07 PM1 votes

Cool idea. You may want to incorporate some power/gold ratios at the champion level. A hyper carry with 6 items is a lot different than a tanky support with 6 items. Would be interesting to account for that.

Makes sense to me though. Generally the teams that are earning gold faster are doing more things correctly, working towards winning the game. There's still value here even with the small sample size.

npbeisbol4/15/2015, 2:54:40 PM1 votes

The flaw i see is: do teams win because of their high gold acceleration or rather, and this is what i think, they are winning and snowball their advantages like the broadcasters say, and consequently end up with more gold aka a higher gold acceleration average.

And is this acceleration really telling you something different than gold per minute?

just some thoughts...its cool to try and look at it in a different way though.