Fortress: Launcher Analysis (The Double-Grind)

Team Strategies go here, if you want to share that is...
Post Reply
User avatar
Graphic Artist
Posts: 6391
Joined: Wed Jan 23, 2008 3:37 am

Fortress: Launcher Analysis (The Double-Grind)

Post by sinewav »


I wanted to take a look at the "launcher" in Fortress. Launchers are areas where players get a speed bonus for driving in-between two trails and are usually self-created. Traditionally, we refer to this as the opening double-grind that positions 2 and 3 perform. It's always suggested that wings should drive 3/4 of the way forward to the edge of the zone before turning back. Allegedly this creates the ideal launcher. I was always skeptical of this claim even though I routinely gave it as advice, so I did a little analysis to find what the ideal launcher length is.

This work was inspired by the continued debate on "bottles." One proposal for improving bottle was to have wings drive to the zone edge before turning back, thus reducing the potential damage of a center attack. Spoiler: this is a bad idea.


This analysis uses a map with a 150 meter wall in place of position 1 (center), stretching from the spawn point at 50m and extending to a "split point" of 200m. I chose 200 meters because this is about where sweepers turn back. CYCLE_ACCEL_RIM is active and GRID_SIZE 5 is chosen for guidelines corresponding to the following test intervals:

60m (basically immediately double-grinding)
90m (the zone edge)

I recorded a series of double-grinds into an AAREC, then captured the recording with OBS studio for frame-by-frame analysis.


The speeds recorded at 200 meters are slower than expected in an actual Fortress grind. I did some tests with a simulated sweeper and found that sweepers add a lot of variance to times and speeds. Final speeds at 200 meters with a sweeper ranged from +3 to +6 mps. (I'll take this moment to emphasize the coordination between wings and sweeps cannot be understated when it comes to a fast, effective opening grind.)

For each of the tests, I performed a simple double-bind backward and double-bind forward. At no point did I intentionally "dig" or grind closer to the center trail. In my preparation for this analysis I tried several variations of grind ranging from rubber usage of 3 to 4.5+ and found that there was little difference in final speeds. The majority of the a wing's speed boost comes from the sweeper, not the launcher or how close the grind to center.

Finally, I did my best to turn precisely at 5 meter intervals, but probably had a +/-2 error range. In the end it didn't make a lot of difference and the results are fairly linear as expected.


The chart in the attached image shows seconds on the Y-axis and speed on the X-axis. This speed is in meters per second at a 200 meter split point. To make sense of this data we need to determine what is the ideal time spent on the grind while maximizing speed for wings? It takes approximately 5.5 seconds for the center and sweepers to reach mid-grid. Ideally, the wings reach the split just as the sweepers turn back in order to maximize their speed contribution. However, as stated earlier, sweepers add a lot of variance to this formula. They might grind late, grind softly, or split early. So it's best to target a launcher length that gives reliable results.

The earliest launcher breaks at 60 and 65m are likely a bad trade off. Wings will over-speed sweepers and be slightly slow on the split. Conversely, the latest breaks at 80 to 90m are likely too slow to reach mid-grid and the speed gains are relatively minor. Turning between 70 and 75m is probably the ideal balance between grind time and speed at split.

As it turns out, the well accepted advice of driving 3/4 the way to your zone's edge before double-grinding really is good advice! However, I feel like this length (turning back at 75m) is actually the upper limit and grinds can benefit from a slightly earlier double-grind and a solid, early outside dig by sweepers.

Further Research

It looks like the community has reached the point where were considering performance changes that affect the game at millisecond resolution. I wonder if one of our great programmers might create a model of the game that allows us to input values and get high resolution data returned to reveal some of the mysteries of the grind?
Post Reply