Jump to Main ContentJump to Primary Navigation
ESPORTS

Introducing Major Leads and Deficits: A new way to look at gold leads

A new metric that looks at the time teams spend with meaningful leads or deficits.

Gold is the simplest, most widely used number to evaluate how a team is doing, whether it is to see how a squad is performing in an ongoing game or how they stacked up versus other teams historically. A team that spends a lot of time ahead should likely be favored over one that spends most of its time with a deficit. More important than if a team is ahead though is the size of the lead - there is a big difference between having a 1 gold or a 5000 gold advantage. Today we are introducing a new stat that looks specifically at the time a team has spent with a major lead.

While I was working for League of Analytics, I introduced a metric called "Significantly Ahead/Behind" aimed at measuring how much time a team spent with a relevant gold lead or deficit. After some optimization with the Riot stats team (and a healthy name change), "Time with Major Lead/Deficit" instead looks at the percent of game time a team has spent with at least 51.5% (Major Lead) or 48.5% or less (Major Deficit) of the total gold in the game.

This stat is intended to:

  • Give a more meaningful measure to evaluate how often a team is really ahead/behind

  • Tell us how a team is performing overall, beyond just the win-rate. This way we can identify which teams might be winning convincingly or which are competitive despite losing

Below you will find an explanation of the exact calculations, what these relative (%) leads mean in terms of actual gold leads in the game, why we chose the exact threshold and more. If you don’t want all the specifics and are eager to find out how your team performed in this metric in the 2017 Summer Regular Season, you can jump directly to the section where we present that data for the NA and EU LCS (LCK and LMS can be found at the very end of the article).

 

How is it calculated?

Time with Major Lead/Deficit (M-LEAD/M-DEFICIT) measures the time a team spent with 51.5% or more/48.5% or less of the total gold as a percent of the total game time from 0 to 40 minutes. The formula for the stat is as follows:

1
2

The stat is calculated for each game and then averaged over all games played. This gives equal weight to games of different lengths. If a team has a Major Lead of 45.0% in 35 minutes of game time in one game and a Major Lead of 35.0% in 25 minute of game time in another game, the average will be 40.0%. The metric is cut off at 40 minutes as gold differences become less meaningful after this point with players beginning to reach 6 item builds.
 


A closer look at Major Leads and Deficits

Now that you know the math behind the stats, let's examine how it works in practice. What does a Major Lead/Deficit look like? Below you can see the what 51.5% of the gold would have equated to on average in the LCS this split:

M-LEAD Gold Difference Threshold

3

2017 EU LCS + NA LCS Regular Season combined

The threshold for a Major Lead, by definition, rises steadily. If a game has a lot of kills and objective trades, that threshold will be higher (but not by that much, the maximum at 20 minutes was 2.3k); if nothing happens, it will be lower. Since this metric is mirrored (if one team has a Major Lead the other one automatically has a Major Deficit), the threshold also tells you how much a team generally had to fall behind to have a Major Deficit.

4

2017 EU LCS + NA LCS Summer Regular Season combined

 

How frequent are Major Leads and Major Deficits?

Evaluating the frequency of Major Leads/Deficits shows some clear trends that are consistent across EU and NA. Before two minutes, a Major Lead was rarely achieved. This is not surprising, given that only First Blood would have triggered it. After that the chance that a team has a Major Lead steadily rose until it hit a peak at 76.9% for NA (30 minutes) and 68.8% for EU (32 minutes). That means that even the relative gold difference (not just the actual gold difference) increased, on average, until about 30 minutes. The reasons for this include more opportunities to set yourself apart the longer the game goes and snowball effects from early leads. After 30 minutes it slowly declined as games with large leads are more likely to end early than close games.

Frequency of M-LEADs over time

5

2017 Summer Regular Season


Does this metric have analytical value?

Owning 51.5% of the total gold at 20 minutes might seem valuable, but does it actually help identify who's likely to win? Looking at win-rates, It becomes clear that having a Major Lead means being in a very good position. Teams that managed to put themselves in this position for large parts of their games were doing well. The ones that won most of their games after having a Major Lead were good at closing out games, while teams that came back from a Major Deficit showed resilience in digging their way back from a substantial early hole.
 

Win% with M-LEAD over time

6

2017 Summer Regular Season

The small dip in the win-rate when having a Major Lead in the first few minutes can be explained by a very low sample size due to the difficulty of triggering a Major Lead this early. The reason why we are not starting at five minutes is that it would require further qualification when explaining the stat without significant improvements. After those first few minutes, we see the win-rate when having a Major Lead rise until the mid-game for both EU and NA. At 15 minutes, it was 74% in EU and 82% in NA this split, at 30 minutes it was 89% in EU and 90% in NA. We included data past the 40-minute mark to show how the win-rate wildly varies in both regions and loses value as a predictive tool at that point - hence the data’s exclusion from this metric. Towards 50 minutes and beyond, the sample size gets very low (until there is only one game left), which is why you see these extreme win-rates at the end.

 

Why the 51.5% threshold?

In determining the 51.5% threshold, we evaluated average win-rates for teams across different gold differentials with the intent of identifying the percentage that fulfilled two major objectives:

  1. The threshold cannot be so low that it includes leads that are not meaningful and have a low predictive power for the final outcome of the game. The goal of the metric is to isolate meaningful leads, if we set the bar too low, we will fail at this.

  2. It cannot be so high that it only includes huge gold leads that have too much predictive power in that it is basically impossible to come back from. In this case the metric would be rather boring and not suited to tell stories about the teams. We don’t want to sit there and say “this team has a Major Lead, it’s over, let’s go home”. Seeing which teams manage to close out the game from a strong position, which ones falter and which ones rise from a Major Deficit is very valuable in evaluating them.

After using 52.0% as a threshold for a while, it became clear that this was too high of a number and failed at objective (2). Looking at win-rates between 51.0% and 51.499% on the other hand revealed that gold leads in this interval don’t necessarily constitute a meaningful lead, therefore failing at objective (1). With 51.5% we feel we have found the threshold that holds up to both of these objectives, while still being a fairly clean number.

 

Evaluating NA and EU LCS Teams

After taking a deeper look at the makeup of the new stat and why we think it helps us evaluate teams, it is time for the fun stuff: looking at the individual teams and how they are performing in this metric.

Here are the Major Lead and Major Deficit numbers for the 2017 EU and NA LCS Summer Split Regular Season (LCK and LMS numbers at the end of the article):

NA LCS - 2017 Summer Regular Season

7

EU LCS - 2017 Summer Regular Season

8

In NA, IMT topped the league with 36.8% of the time spent with 51.5% or more of the total gold between 0-40 minutes (M-LEAD). At the same time they posted the lowest (and therefore best) Major Deficit with 19.0%. TSM, the highest ranked Regular Season team in NA, reveals some interesting facts. They were only 6th in Major Lead, but 2nd in Major Deficit. Without watching their games, you could quickly surmise that they were likely getting their wins by not falling far behind and often times winning games that were fairly close (58.1% of the time they were neither majorly ahead or behind, that is 11.1 percentage points higher than for any other team in NA).

Looking across the pond, Fnatic (FNC) dominated EU, finishing the split with a 42.0% Major Lead (1st) and 5.7% Major Deficit (1st). Overall the EU table aligns a lot more closely with the actual standings, the Unicorns of Love (UOL) being the only real outlier. The difference between the NA and EU table is enormous. Not only is Major Lead a better indicator of rank in EU than in NA, comparing the distance from 1st to 10th place also reveals just how big the discrepancy between top and bottom was in EU compared to NA. IMT’s 36.8% Major Lead was “only” 19.4 percentage points higher than FlyQuest’s (FLY). In EU the difference between FNC and Mysterious Monkeys (MM) was 35.9 percentage points. The discrepancies for Major Deficit follow a similar trend.

 

An example: Unicorns of Love

Let’s take a closer look at UOL to see how these numbers can help us understand more about the team. If you have not watched many of their games this split, their low Major Lead and high Major Deficit numbers might be surprising for a team high in the standings. We can already see that their wins did not result from consistently being ahead and closing out games from an early lead - they spent way too much time with a Major Deficit. One cool feature of our new metric is that it is easily split into different time intervals to see differences in early vs mid/late game performance:

 

UOL M-LEAD and M-DEFICIT over time

9

2017 Summer Regular Season

In the first 10 minutes of the game, UOL performed poorly, having Major Leads less than 10% and Major Deficits more than 15% of the time. This trend continued between 10 and 20 minutes. During this period, UOL spent more than a third of their game time with a Major Deficit, usually a sign of a team that does not win many games. This makes the shift that happens after that so impressive. Through forcing fights on their terms and coming out ahead more often than not, they managed to even out the time spent with a Major Lead and Deficit from 20 to 30 minutes. After 30 minutes they were able to continue that trend and post a good 41.2% Major Lead. Combined with a still fairly high Major Deficit of 29.3% during that time, a clear picture emerges. UOL is a team that had weak early games, but often time managed to turn the game around in the mid-game. On the other hand, if they lost, they often times got beaten convincingly, highlighted by the fact that their Major Deficit remained high for a high win-rate team.

Comparing this analysis to an approach that only looks at how much time a team has spent with any lead or deficit would in this case miss part of the transition from early to mid game, because it would count the instances where the team had already climbed back into a decent position but were still slightly behind. It would give this small and mostly meaningless deficit the same weight as one where the team is thousands of gold behind. Looking at average gold differences on the other hand (GD@10, GD@20, etc.) would show that UOL did not have an average lead higher than +130 until after 35 minutes. The problem with this analysis is that the team has had a few games with deficits of 8k and more after 30 minutes. Those values heavily skew the average, even though they actually spent most of that time ahead.

To summarize the disadvantages of (1) time spent with any kind of lead and (2) the average gold differences compared to Major Lead/Deficit:

  1. If a team is ahead at 20 minutes 75% of the time, we don’t know if their leads were actually relevant since there is no distinction between a 1 and a 5000 gold lead.

  2. A team that has a +500 average gold difference at 20 minutes (GD@20) might have spent most of their games slightly behind, but a few games with very large leads put them at a positive average number.  

On the other hand, If a team has a high Major Lead and a low Major Deficit in the first 20 minutes, we can be sure that the team is performing well in the early game overall. The same holds true for evaluating other time intervals. The metric does a good job of combining the advantages of the two approaches above without inheriting their main weaknesses.

 

Conclusion

As with any metric, putting the information into context is important. A team like TSM - who doesn’t rank very high in Major Lead but are 2nd best in Major Deficit - might be drafting for the mid/late game and therefore staying even early is reasonable. That being said, while drafting/team composition may lead  to a slight deficit early, falling behind far enough to trigger a Major Deficit is almost always a concern, no matter how much the draft favored the other team’s early game.

To summarize, we can say that spending a large amount with a Major Lead is a sign of a team that is able to generate meaningful advantages consistently, while posting a high Major Deficit number reveals consistent struggles. A team that manages to place high despite performing poorly in those metrics might have very specific strengths (like UOL’s mid/late game), but have also shown consistent weaknesses. The metric therefore allows us to dig deeper than the win-rate and can highlight high ranked teams that we should be worried about or low ranked teams that might have performed better than their standings suggest. Beyond just evaluating teams, Major Lead & Deficit spreads across a league can also tell us more about the parity within that region.

We're excited to introduce this new stat to you guys and would love to hear your thoughts on it. Let us know what you think in the comments below, or hit us up on Twitter @EULCSSTATS. Below you can find the Major Lead and Major Deficit numbers for LCK and LMS teams in the 2017 Summer Regular Season.

LCK - 2017 Summer Regular Season

10

LMS - 2017 Summer Regular Season

11