Gold Shift Events – Capturing Major In-Game Shifts

The purpose of League of Analytics is to provide more than the standard analysis for professional games. We hope that this gives us a better understanding of the game. League of Legends has the advantage that gold captures a lot of the important aspects at once. It is a very good approximation of a team’s current standing in a match. Thus, gaining more gold than your opponent is one, if not the, most important part of winning the game. Whether you achieve that by smart map movements, where you have the other team scrambling to catch up and losing towers and farm in the process, or by constantly winning fights and skirmishes, the outcome is largely reflected in your gold advantage. Keeping this in mind, we thought about ways to quantify relevant events within a game and ended up with gold shift events (GSEs) as the major new team statistic we want to introduce.

Gold shift events

GSEs are events that result in a significant shift in gold in a short period of time. For this, it is not relevant whether the gold lead changes, but only whether the event leads to a significant change in gold difference between the two teams that surpasses a certain threshold within a time-frame of either one or two minutes. Before I jump into a summary and a few examples to illustrate the metric and compare the teams, let me answer two quick questions.

What is the threshold that triggers a GSE? For this, we tried to find a value that constitutes a significant gold shift and does not trigger too easily (one kill at minute 20), but often enough, while being similarly distributed over the time intervals. We came up with a simple formula that scales linearly with time and equals 600 gold at 5 minutes and 2000 gold at 30 minutes. For an exact specification, check out the documentation section at the bottom of this post.

gold shift events threshold

Why a timeframe of either one or two minutes? First of all, the match histories only provide data in minute intervals (1, 2, 3, …,15 ,16, …, etc.). If we only took one minute intervals, we would miss GSEs that overlap two distinct intervals. A team fight that results in a 1000 gold difference change after 10 minutes should trigger a GSE. But if it took place from 9:45 to 10:15, with 500 gold gained on both sides of the minute mark, it would not trigger if we only evaluate single minute intervals. That is why we check the size of the change in gold differences for one and two minute intervals. As a result, one large event could trigger two separate GSEs (think 2k gold on both sides of the minute interval). To account for this, two consecutive GSEs are only registered if they are not in favor of the same team.

GSE statistics and examples

To illustrate gold shift events, I used data from all NA LCS Spring ’16 regular season games. One important aspect to keep in mind is that every GSE registers as a GSE for both teams that participated in it. A team can either “win” or “lose” such an event. Here is a short overview:

  • # of GSEs for whole season: 455
  • # of GSEs per game: 5
  • Average # of GSEs per team: 91
  • Winning teams won ~ 80% of their GSEs
  • Losing teams won ~ 20% of their GSEs
  • Highest / lowest # of GSEs: TiP (105) / FOX (77)
  • Most / least GSE wins: IMT (79) / DIG (23)
  • Highest # of GSEs in a game: 9 (four times)
  • Lowest # of GSEs in a game: 2 (five times)

And here is a graph showing the number of GSEs and what percentage of their GSEs the teams won.

gold shift events win percentage

It comes as no surprise that the dominant team of the split also dominated events that lead to a significant shift in gold. Immortals stand head and shoulders above the rest of the pack at an 85% GSE win percentage. If we look back at the spring split, IMT excelled at aggressive and decisive plays that allowed them to gain big gold leads quickly. Even in the rare situations where they were behind, they won 4 out of 4 GSE that occurred at a time where they only had 48% or less of the total gold just two minutes prior to the event. On the other hand, we have a team like Team Impulse which ended up in 105 GSEs during the regular season and only managed to win 26% of them. This includes only 3 out of 6 GSE wins when being significantly ahead (owning 52% or more of the total gold two minutes before the event).

As expected, winning a large share of your GSEs correlates with high win percentage numbers and vice versa:

gold shift events and wins

For now, we will present the number of GSEs and the GSE win percentage on our stats page. After a few weeks, we will add GSE win percentages when significantly ahead or behind. For those who are interested and in the spirit of full disclosure, the last section of this article will document the creation of the metric in detail. For those who don’t like that dry stuff, thanks for reading and please let us know if you have any feedback. If you like the content, follow us on twitter to get notifications when the stats are updated or new content is released.


Gold difference: Difference in gold between the two teams at any given time (available for each minute)

Gold difference difference (GDD): Absolute difference in gold difference between two time intervals. GDD1 is the difference for one minute and GDD2 for two minute intervals. GDD2 is also evaluated at one minute intervals. So at 5 minutes GDD1 is evaluated from minute 4-5 and GDD2 from minute 3-5. At 6 minutes, GDD1 calculated from 5-6 and GDD2 from 4-6 minutes. This way, it captures every jump from one full minute to the next.

Gold shift event (GSE): Occurs if either GDD1 and/or GDD2 exceeds the gold shift threshold (GST). It is either won (1) or lost (0). It does not trigger if a team has two consecutive GSE wins or losses (but does trigger if a GSE win is followed by a GSE loss or vice versa), to avoid GSEs to be counted twice.

Gold shift threshold (GST): (Minutes + 40/7) * 56 – it reaches 600 gold at the 5-minute mark and 2000 gold at the 30-minute mark.

Check Also

Fatured image LSR

Exploring Our New Lane Swap Metric

We recently introduced lane swap data for the EU LCS, NA LCS, LCK and LMS. Here is an introduction into the metric as well as quick first analysis.