Will AI Change Dirt Trial Setup? โ The Frontier of Data Analysis and Surface Adaptation
Two timed runs of about two minutes each: what can AI learn from so little data? DirtMan asks from the driver's seat
AI has begun penetrating motorsport engineering in earnest. Telemetry analysis, lap time prediction, tyre degradation modelling โ in WRC and Formula 1, these are already reality. But what about Japan's own discipline, dirt trial? Constantly changing surfaces, single-pass time attacks, and limited budgets โ how far can AI function in this particular environment? DirtMan reflects from his own competitive experience.
Data now sits at the heart of top-level motorsport. A Formula 1 car carries hundreds of sensors, and teams pore over what they record across every race weekend. So it is natural to ask whether artificial intelligence could one day help with car setup in a discipline like dirt trial. As a dirt trial competitor, I approach that question with real curiosity โ and with caution, because this sport is unusually hard to read from numbers alone.
Start with how little data there is. In dirt trial, cars run the course one at a time, a run lasts roughly two minutes, and each driver gets just two timed runs, with the faster one counting. At All Japan rounds there is usually an open practice session on the Saturday, then the two heats on the Sunday. Over a whole weekend, that adds up to only a handful of runs per car. Machine learning thrives on large amounts of data; dirt trial hands it very little.
Then there is how quickly things change. With a full field running, a single heat can take two to three hours, and the unpaved surface keeps evolving as car after car passes. At Round 5 at Ebisu in Fukushima this June, all 10 class winners set their best time on the second run, and the PN2 winner went almost nine seconds faster second time out, from 1:36.720 to 1:27.951. A model trained on the first run could easily point the wrong way for the second. And much of what a driver relies on is hard to log at all: how the ground felt underfoot when walking the course, or what the car is saying through the seat and the steering wheel.
That does not make AI useless here. I can imagine it helping in more modest ways. Keeping years of setup sheets, tire pressures, weather and results in one place, so that patterns become visible across seasons instead of living only in memory. Comparing onboard video from one run against the next. Gathering published results โ sheets that already list both heat times, and sometimes even the tire brand on every car โ and asking sharper questions about where a rival found time. None of that replaces judgement, but it could sharpen it.
In the end, the decision is still made at the start line, by a person, with information no model has: the walk, the earlier cars, the feeling in the hands. The real question is not whether a machine can drive the course for us. It is whether it can help us remember, compare and notice a little better โ and on a surface that never stays the same, that alone could make a difference.
This column space is a paid placement for sponsor companies. Exposure across all World Stage Journal articles, distributed in 23 languages worldwide.
Inquire About Placement โShare this article
Comments
0Related Articles

Why I Chose Dirt Trial โ DirtMan
It was my uncle โ an All-Japan Dirt Trial champion โ who first brought me to this sport. I had alwayโฆ
From Airtime to the Feed โ How Social Media Entered Motorsport Sponsorship
For a long time, a logo on a racing car was valued mainly by how often it was seen on television. Soโฆ
From Dirt Trial to Cross-Country Rally โ A Transition Guide and the Real Steps Involved
"If I compete in dirt trial, surely that carries over to rally" โ it is a question that many driversโฆ
FOUNDING READER
This journal exists because 100 people believed in it before the world did.
Seat #001 could be yours.