A Couple Graphs on Skill Level vs. Games Played

Hey folks,

I’m brownbear. This is a short piece comparing skill level with total games played in StarCraft II. The data set is ranked 1v1 Legacy of the Void on the North American server.

First, here’s a graph plotting each league’s 90th percentile games-played (current season). In other words, if you’ve played 369 or more ladder games, you’ve played more than 90% of Master’s accounts in North America, but less than 10%.

ArticlesSocialMediaAndMiscContent 2017-11-09 [Games Played and MMR I].png

This graph matches my expectations – players in higher leagues play more. Players in the highest leagues play a lot more. Play more, get better. It’s intuitive, to me at least.

Next, here’s a graph plotting matchmaking rating against total games played across a 3% sample of the entire North American ladder.

ArticlesSocialMediaAndMiscContent 2017-11-09 [Games Played and MMR II].png

This one’s more surprising to me: it’s noisy and the trend line (black) is modest. There’s tons of outliers, from Bronzers with a thousand games to GM players with just a few dozen. There’s enough of these to make the graph look like a ball rather than demonstrate a clear trend.

Intuitively, playing more makes you better: I don’t think anyone would say that’s completely wrong. Also intuitively, a modestly active Grandmaster player will not suddenly drop down to Gold, nor will a Gold player massing games suddenly arrive at Blizzcon.

My guess: how much you play on any given day – or even any given season – features diminishing returns, meaning you need to play consistently over a long period of time to develop higher levels of skill. Furthermore, skill is durable, so taking a long break from the game is not necessarily a disaster.

To prove this, I’d need to graph current matchmaking rating against career total games rather than just the current season. To do that, I’ll need to pull down player profile data, which I don’t currently do. I’d probably also need more sophistication in eliminating smurf accounts, combining barcode accounts to find a player’s “true” total games played, and getting rid of accounts with only one or two games. Ideally, with ladder data spanning a long enough period of time, we could identify the characters improving the most and see if we can find a pattern in how they’re practicing relative to their peers. Anyway, next time!

How about you – what do you think? I’m interested to hear your feedback on how much of a relationship you think playtime has with skill level, in addition to what kinds of data we should use to prove it.

If you’d like to see more content from me, you can follow me on Twitter and Facebook or check out my game-design focused YouTube and Twitch channels. Thanks and all the best!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s