SCORING SYSTEM

TL;DR

There are 29 leaderboards (Time Trials + Championships).
All submitted runs in a given leaderboard are scored 0–100, with the slowest time receiving 0 and the world record receiving 100. Other scores increase exponentially as times approach the world record.
A player's total score is the sum of all their run scores, plus 35 bonus points for each world record. Missing runs do not contribute to the total score.
Rankings are determined by comparing players’ total scores.

Why this scoring system?

The community used to rank players through discussion and consensus. That captured a lot of nuance but it wasn’t fully transparent or reproducible, and different people could arrive at different rankings. That’s why we moved to a data-driven scoring system.

Ranking players quantitatively turns out to be harder than it looks:

Ranking players by time alone doesn’t work well. Some tracks are much longer than others, so a half-second improvement can mean something very different depending on the course. Adding up time differences would make long tracks count far more than short ones.
Looking only at a run’s place on the leaderboard has its own issues. A 10th-place run a few frames behind the world record isn’t the same achievement as a 10th-place run a full second behind it.
Considering milestones like Benchmarks and Standards has problems too. For one, they don't even exist for Championship courses. And because they use hard cutoffs, very similar runs can fall on opposite sides of a milestone. An exception is made for world records, which receive bonus points toward a player's total score.
There’s also the fact that not every player has a run on every leaderboard. Any ranking system has to balance rewarding consistency with recognizing standout performances.

This scoring system addresses all of these issues. It scores each run based on where it falls between the slowest time and the world record on its own leaderboard. It treats all tracks the same way, avoids hard cutoffs, and is fully data-driven. Anyone can reproduce the rankings and see exactly how every score was calculated.

Example

The easiest way to understand how the scoring system works is to walk through how scores are determined for a real leaderboard. Let's use Drake Lake 3 Lap as an example. Here are the times for all submitted runs:

Figure 1 — All Drake Lake 3 Lap runs, from slowest time to fastest.

Normalizing the Times

For each leaderboard, we re-scale the time axis so that the slowest submitted time maps to 0 and the world record time maps to 1. This doesn’t change the data. It only changes the scale we use to measure it — like how converting miles to kilometers doesn't change the actual distance.

Equation 1 — Normalization equation

Figure 2 shows the same Drake Lake 3 Lap runs after normalization. The runs haven’t moved or changed relative to each other — only the scale on the horizontal axis is different.

Figure 2 — Drake Lake 3 Lap runs after normalization to a 0–1 scale.

Why Normalization?

So far, we’ve only looked at one leaderboard. Normalization matters most when you compare different tracks. Figure 3 shows three very different leaderboards before and after normalization, with original times on the left and normalized times on the same 0–1 scale on the right. See how they become much more comparable?

Figure 3 — Actual times (left) and normalized times (right) for three different courses.

Calculating the Scores

After normalization, every run has a number between 0 and 1. At this point, we could just add up everybody's normalized scores and call it a day. But this doesn't reflect how times closer to the world record are exponentially more difficult to achieve than the times you get when first starting out. To account for that, we need to score the runs exponentially.

Equation 1 — Exponential scoring curve

We use this equation to turn a normalized time into a score between 0 and 100. Scores rise slowly at first and much faster near the top. Plotting the equation gives a scoring curve that looks like this:

Figure 4 — Score curve used for all leaderboards.

Near the bottom of the curve, even large changes barely move the score. Near the top, tiny changes produce big jumps. This same curve is used for every leaderboard. Let’s see where the Drake Lake 3 Lap runs fall on it:

Figure 5 — Drake Lake 3 Lap scores overlaid on the score curve.

That’s the entire scoring system. Every run is scored the same way, using only its position on its own leaderboard. Times are normalized so different tracks are comparable, and scores are weighted so performances near the world record matter much more than those far away. A player’s ranking is simply the sum of their scores across all leaderboards, plus bonus points for every world record. The result is a system that’s consistent, transparent, and easy to reproduce.

Page updated

Google Sites

Report abuse