FREE $20 RP
Download the Blitz Esports App.
Learn More
Feature

How best-of-ones will change the coaching landscape of the LCS: Less stats, more analysis

By Tim Sevenhuysen
Oct 18, 2017
What’s old is new again.

The North American LCS recently announced that in 2018 their regular season schedule will revert to a best-of-one (Bo1), double round robin format. Rumors suggest that the European LCS will follow suit. It’s a significant change, and loud voices have been raised both in favor of and in opposition to the move.

Of course, if you’re an LCS team, it doesn’t really matter whether you are for or against the change; it’s happening regardless. What matters is how you react. On the competitive front, that starts with understanding how the shift to best-of-one play affects the optimal approach to preparing your team each week.

The first topic many will raise is that with fewer games being played, scrim time becomes more important. In their announcement of the change, Riot directly acknowledged that idea, commenting that they are “working with the Players Association and teams to make practice and scrims more efficient.” That’s a nice sentiment, but it barely scratches the surface. It’s hard to believe that any team wasn’t already constantly seeking to maximize the value of their scrim time.

In truth, the implications of the Bo1 structure for coaching and analysis go much deeper. Bo1 play will reduce the value of certain kinds of analysis, while rebalancing how teams ask their analysts to spend their time.

Stats mean less

The analytical consequences of the Bo1 format start with a topic that is especially dear to my heart: statistics. Simply put, in a Bo1 regular season, stats become less valuable for coaches and analysts.

The dirty little secret of LoL statistics is that most of them don’t communicate much information that is truly important. The value of most statistics comes from testing and validating assumptions. If you have the impression that a team gives away too many dragons, you can look at their dragon control stats to test your idea. Or if you’re looking through the stats and notice that a player has an especially high First Blood rate, you can go watch that player’s games and specifically look for how he is earning those early kills.

Simply put, in a Bo1 regular season, stats become less valuable for coaches and analysts.

This core truth about the value of stats in LoL hasn’t changed, but sample sizes are shrinking, and that means it’s more important than ever to go beyond simple numbers, measurements, and models.

If a best-of-three series averages 2.5 games played, then Bo1s will lead to about 60% fewer games being played by a given player or team, or played on a given patch (unless Riot forgoes their 2-week patch cycle for the 1-month cycle they’re considering for next year). When sample size decreases, each data point within a given piece of analysis has a larger effect on the overall result, which makes it more likely that outside factors—which we can call “noise”—will bias the findings.

Champion picks and matchups are one example of noise in LoL stats. Let’s say an analyst wants to form an opinion on Søren “Bjergsen” Bjerg’s skill at 1v1 laning by calculating his average CS difference at 10 minutes (CSD10). If there are 25 games to analyze, each individual game counts for 1/25th (4%) of the end result. If Bjergsen was drafted into a really difficult matchup twice, and ends up with -15 CSD10 in those games, we could say that 8% of the data has meaningful bias. That might be a problem.

Now imagine there are only 10 games of data. Suddenly those two games represent 20% of the sample, and that's a real problem. If the analyst remembers that those two games are part of the sample, they can factor that into the interpretation. But if they’re like me and can’t recall every champion matchup played in every game across an entire season, you can see why reduced sample size becomes more likely to create problems, forcing the analyst to be that much more careful to check their assumptions and do more research—in this case, looking up Bjergsen’s champion pool and laning matchups.

This is how stats lose some of their value in a Bo1 environment. When analysts can assume that most statistics available to them are suffering from noise, they might as well relegate the stats to secondary importance by default and dive straight into the deeper research.

Less scouting; more self-improvement

Statistics are only one tool analysts use, but the Bo1 format doesn’t only affect analysts’ tools. It also affects their philosophies.

Analysts are responsible for preparing their teams in two main ways. First, they review their own team’s performances and try to recognize strengths and weaknesses, using that information to guide the team’s practice time. Second, they scout upcoming opponents, once again identifying strengths to defend against and weaknesses to exploit, and using those findings to help create specific game plans.

Both self-improvement and opponent scouting are useful, and they always exist in a balance. But in a Bo1 league, the potential value of scouting is greatly diminished, and self-improvement must become a much higher priority.

With the game count diminishing, scouting-heavy analysts will lose a lot of their ammunition—at least as far as the regular season is concerned. Since there are just 40% as many games to draw on, it is much easier for the opponent to throw a curveball that could invalidate your research. In a Bo3 league, it might be useful to point out that the enemy support player often roams into the river at a specific time to place a ward, or that a certain enemy mid laner often stands too far forward at the start of team fights. In a Bo1 league, the opportunities to spot those trends are heavily reduced, and the implied likelihood of the trends being repeated is lower.

By necessity, teams will need to reallocate their analytical resources. Scouting work in the regular season is unlikely to go much further than tracking opponents’ champion pools in solo queue or identifying any especially unusual setups in their level 1 movements or jungle routes. And even these tasks should become as fluid, automated, and tool-based as possible, to allow the analysts to focus on refining their own team’s strategies and execution.

Both self-improvement and opponent scouting are useful, and they always exist in a balance. But in a Bo1 league, the potential value of scouting is greatly diminished, and self-improvement must become a much higher priority.

The way forward

With the decreasing value of statistical analysis and opponent scouting, teams’ weekly preparation should instead be focused primarily on defeating the enemy within. Professional analysts should mostly spend their time on deep dives and rich breakdowns of games, dissecting key moments, understanding the decisions players are making, and reverse-engineering the thought processes that led them there.

In a Bo1 environment, teams don’t need dedicated statisticians—as much as it pains me to say it! What they need are well-rounded analysts who can use video, data visualizations, and, yes, statistics, to learn useful things—like how the world’s best teams set up their vision control or execute Baron dances—and clearly communicate that insight to others, supported by the evidence they gathered. And teams need to provide their analysts with access to the tools, resources, and environment that enable them to find and deliver that value.

It's back to Bo1s, and back to basics, for the LCS in 2018, and the teams that recognize the implications of the format change and adopt the right analytical priorities will come out ahead.

-----

Tim “Magic” Sevenhuysen is the founder of OraclesElixir.com and the Product Lead of Shadow LoL, a professional analytics platform.

Feature
NA LCS
Tim Sevenhuysen profile
Tim Sevenhuysen- Contributor
Tim “Magic” Sevenhuysen is the founder of OraclesElixir.com and the Product Lead of Shadow LoL, a professional analytics platform.
Subscribe to our YouTube channel!