Fuji Endurance Case Study: From P12 to P3 in GT3
Updated: Aug 20
Fuji Endurance Case Study: the driver's Run from P12 to P3 in Class
This is the kind of Driver Result Valor should publish: the outcome first, the telemetry behind it second, and the caveats beside the numbers instead of buried underneath them.
the driver drove the Ferrari 296 GT3 for the endurance team at Fuji. The team started P20 overall and P12 in GT3 and finished P10 overall and P3 in class. the driver completed 161 race laps, recorded a 1:35.850 best lap, and finished the event with 11 incidents attributed to his driver result.
The result at a glance
GT3 result: P12 start → P3 finish
Overall result: P20 start → P10 finish
Positions gained: +9 in class, +10 overall
the driver's race laps: 161
the driver's best lap: 1:35.850, set on lap 392
Gap to the team's best lap: about 0.117 seconds
First 20 vs last 20 race-lap median: 1:37.455 → 1:36.630, an improvement of 0.825 seconds
Modeled turn loss: 1.933s → 1.347s, down 30.3%
Corners improved in the race comparison: 12 of 13
Best five-lap average: 1:36.138
Those numbers are why this is a useful case study. The story is not simply that one heroic lap appeared. The stronger signal is that the later clean-lap sample was faster while the corner-level loss model also moved in the same direction.
Before the race: the development work was already visible
The Fuji program did not begin on race day. Valor's archive contained 129 valid clean laps from the preparation work for this exact Fuji No Chicane / Ferrari 296 GT3 combination.
The first recorded clean lap in that archive was 1:39.106. The early personal best reached 1:35.347. Comparing the first five useful laps with the last five produced a more conservative measure of the progression: 1:36.201 → 1:35.824, or 0.377 seconds faster.
More importantly, the modeled turn loss moved from 0.869 seconds to 0.596 seconds — a 31.4% reduction. That matters because it tells us the stopwatch was not moving by itself. The corner model was seeing less execution loss at the same time.
Two examples make that concrete:
Turn 4 exit speed improved from roughly 118.9 mph to 123.9 mph.
Turn 13 minimum speed improved from roughly 55.1 mph to 57.0 mph.
Those are not generic coaching statements. They are measurable changes in how the car was being carried through specific parts of Fuji.
Race day was not a clean laboratory test
the driver's race consisted of four fuel stints, and the conditions were not equivalent from one stint to the next.
Stints 1 through 3 started on fresh tires. Stint 4 was completed on used tires, and the driver reported a clear loss of front-end response that required him to change how he loaded the front of the car to achieve turn-in. Stints 3 and 4 were run in darkness. Multiclass GTP traffic and active battles could cost substantial time when they arrived at the wrong part of the circuit. An operational problem around the beginning of Stint 3 also created an abnormal segment that should not be treated as representative pace.
That context matters. If we simply averaged every lap and announced that the last stint was faster or slower, we would be mixing driver execution with tire age, traffic, visibility and race events.
What changed during the race
The clean-lap comparison gives us a much better answer.
Across the driver's race sample, the median of the first 20 representative laps was 1:37.455. The median of the last 20 was 1:36.630. That is 0.825 seconds of observed pace improvement from the early sample to the late sample.
At the same time, modeled turn loss dropped from 1.933 seconds to 1.347 seconds. That is a 30.3% reduction, with 12 of 13 turns improving in the comparison.
That second number is critical. A lap can become faster because of draft, traffic timing, fuel load or a single unusually good sector. When the aggregate lap pace improves and the corner-loss model improves across nearly the entire circuit, the case for genuine execution improvement becomes much stronger.
the driver's best five-lap average reached 1:36.138. His individual best of 1:35.850 came within roughly 0.117 seconds of the team's best lap, which was approximately 1:35.733.
How the driver compared with the GT3 field
In the official split analysis, the driver ranked 13th of 72 GT3 drivers for average pace and 19th of 72 for fastest lap. His incident rate ranked 7th of 72 in the cleanliness comparison.
Among the 25 Ferrari 296 GT3 drivers in the same analysis, his pace ranked 4th and his cleanliness ranked 3rd.
That gives the result useful scale. This was not simply a driver getting faster in an empty practice server. The pace sat competitively inside a large endurance field while the team converted its starting position into a GT3 podium.
What Valor was actually doing
This distinction is important because it keeps the case study credible.
the driver raced with Valor overlays active and chose to keep spoken corner coaching off during the event. He still used Valor's race-engineering information, particularly strategy and relative-gap updates, while the preparation work had already produced the baseline-backed telemetry and corner comparisons used to understand the circuit.
So the claim is not that an automated voice told the driver how to drive every corner and created a podium.
The useful claim is that Valor contributed to a continuous engineering loop:
Reference: establish measured Fuji/Ferrari baseline behavior.
Observe: show where the driver's lap and inputs differ.
Diagnose: isolate corner and phase loss rather than relying only on lap time.
Test: change the driving and run another sample.
Race: keep the useful visual and race-engineering information available without forcing coaching chatter on the driver.
Review: compare later performance with the earlier evidence.
That is much closer to how an actual development program works.
The used-tire stint matters to the story
Stint 4 is especially useful because the driver's own feedback explains why a reference cannot be treated as a script.
On the used tire, he felt the front-end drop-off and had to deliberately load the fronts differently to get rotation. A good telemetry system should not look at that and say, “Your trace no longer matches the fresh-tire lap, therefore it is wrong.”
The engineering question becomes: Did the altered input sequence produce the best available execution for the grip that remained?
That is why Valor uses a baseline as a measured reference rather than as a command to copy every input perfectly. Tire state, fuel, traffic and track conditions change. The reference helps expose the difference; the driver and the analysis still need to understand why the difference exists.
What the data supports — and what it does not
The data supports several strong statements:
the driver's later representative race laps were 0.825s faster in median pace than the early sample.
Modeled turn loss fell 30.3% across that comparison.
12 of 13 turns improved.
His best five-lap run reached 1:36.138.
His best lap was 1:35.850, within about 0.117s of the team's best.
His team advanced from P12 to P3 in GT3 and from P20 to P10 overall.
The preparation archive had already shown reduced turn loss and specific speed gains at Fuji before the event.
What the data does not prove is that Valor alone caused every tenth. There was no controlled A/B race with identical tires, fuel, traffic, weather and driver state. the driver drove the car. The team executed the race. Conditions changed.
That limitation does not weaken the case study. It makes the conclusion more useful: Valor gave the driver and team a measured way to see whether the driving itself was changing inside a messy real race environment.
Why this result matters for Valor
The most encouraging part of this program is not the podium by itself.
It is that the different layers of evidence agree with each other:
practice pace progressed,
practice turn loss declined,
specific corner speeds improved,
later race pace improved,
race turn loss declined across almost the whole circuit,
the driver produced competitive sustained pace in the official GT3 field,
and the team converted the event into P3 in class.
That is what a real Driver Result should look like. Not “the software said he improved.” The stopwatch, the corner model, the telemetry and the race result all give us something concrete to inspect.
Want to analyze your own driving against a measured reference? Explore Valor or download Valor for iRacing.
Comments