Stats Theory

Mid/Late-Game Performances: Win Rates From Ahead, Even, and Behind

Note: Scroll down to the data if you want to skip the introductory theory talk!

A team’s early-game performance is relatively easy to measure. We have a wide variety of metrics, like gold different at 15 minutes, first blood/dragon/Tower, and more.

It’s much more difficult to measure a team’s performance in the mid and late game, because those parts of the game are so heavily influenced by what has happened in the early game, but that’s exactly what I’m trying to do here.

To solve the core problem, I developed a set of win percentages when a team is ahead, even, or behind in gold at 15 minutes, using a margin of 750 gold as the benchmark for a lead/deficit. (I explored using 1,000 gold as well, but decided to stick with 750 for now, though the margin is still up for discussion.)

In doing this, I took cues from work Spellsy has done, though I made some tweaks to timing and margins.

The tables below show how well each team played in the mid and late game, from three perspectives. First, how well did they close out games when they had a lead? Second, how well did teams perform in very close games? Third, how well did teams overcome deficits through superior mid- and late-game play?

The Data

Without further ado, here’s the data! (Note that these lists are sorted by Win% from a deficit.)

Click the graphics to see them full size.

Noteworthy Numbers

  • CLG and TSM were the best late-game teams in North America, both making comebacks 50% of the games they were behind.
  • Dignitas was 6-0 when they had a 750+ gold lead at 15 minutes.
  • Team Impulse was just 13-6 when they had a 750+ gold lead at 15 minutes.
  • Elements was an impressive 4-6 when behind at 15 minutes, but only 3-5 when they were even or ahead.
  • Roccat was the third-best team in Europe at playing from behind, but seventh-best at playing with a lead.
  • SK Telecom T1 was the only team in the LCK that won more than a third of their games when playing from a 15-minute deficit.
  • KT Rolster, the consensus second-best team in Korea, was second-best at closing out leads, but was in the bottom half of the league in making comebacks.
  • AHQ dominated the LMS.
  • Hong Kong Esports played very well from ahead, but had a hard time making comebacks, as did most of the LMS.