The Verification Venue · grip, not distance
The Apex You Give Away
The fastest way through a corner is not the shortest way through it. The quick line is deliberately longer than the straight-through geometric line, because a car's corner speed is capped by grip, and grip speed rises with the square root of the radius you can carve. So you give away the apex, tuck a tighter, slower entry, and buy a wider, faster exit that pays you back down the whole straight that follows.
Every driving-school diagram says "take the racing line," and every gamer feels the lap time drop when they stop cutting corners. Here is the physics under it, live. One equation caps how fast any grip-limited car can hold a curve of radius r: the tyres can only supply so much sideways force, and that force is what bends the path. Set the friction and the radius below and read the ceiling.
The speed ceiling v = √(μ · g · r)
~0.5 wet road, ~0.85 warm road tyre, 1.1 to 1.4 racing slicks.
The radius the car actually drives, not the painted corner.
The derivation is one line, and the punchline is in it. The sideways (centripetal) force needed to hold speed v on radius r is m v² / r. The most the tyres can push sideways is friction, μ m g. At the limit they are equal: μ m g = m v² / r. The mass m sits on both sides, so it cancels, leaving v = √(μ g r). A loaded truck and a go-kart share the same cornering-speed ceiling on the same surface. And because v grows as √r, opening the radius by 20% (say 50 m to 60 m) lifts the ceiling by about 9.5% (the square root of 1.2). That square root is the whole game.
So why is the shortest line slow?
The shortest path clips the inside of the corner: it is one tight, constant-radius arc. The tighter that radius, the lower the grip ceiling, and worse, the car is still turning as the corner opens onto the straight, so it cannot get back to full throttle. The late-apex line does the opposite. It sacrifices a slower, tighter entry to straighten the exit: a bigger exit radius means a higher speed ceiling and the wheel is already unwinding, so the driver is flat on the throttle earlier. That higher exit speed is not a one-time bonus. It sits on top of every metre of the straight that follows. Drag the line and watch it happen.
Apex speed (slowest point)
–
geometric: –
Exit speed (onto the straight)
–
geometric: –
Lap-time delta (vs geometric)
–
over corner + straight
Where you leave the outside edge and start to steer. Later = tighter entry.
Where the line kisses the inside kerb. Later apex opens the exit.
The lever. Slide it to 0 and the late-apex advantage vanishes; open it up and the faster exit compounds. Watch the delta cross zero.
What you are watching: the tool samples your line, reads the local radius r at each point, caps the speed at √(μ g r), then runs the car through on a friction budget (it can spend grip on turning or on accelerating, never more than the tyres have), and integrates the time. Then it does the same for the geometric line and subtracts. The late apex starts behind: its apex speed is lower. But it reaches full throttle sooner and exits faster, and once there is a straight to run down, that exit speed wins the second back and keeps winning.
| following straight | geometric time | your-line time | delta | winner |
|---|
Same corner, same car, same friction: only the length of the straight that follows changes. Short straight, the geometric line's higher mid-corner speed wins. Long straight, the late apex's higher exit speed wins. The break-even is a real, computed number.
The honest twist: "later apex" is not a law
The reversal is real, but it is the answer to one specific question: a single corner followed by a straight. Change what comes next and the verdict can flip back. The late apex only pays if there is a straight for the exit speed to compound down. Take that straight away (drag it to zero) and the late apex is simply the slower line, because you gave up mid-corner speed and got nothing for it. The people who actually do this for a living have a rule for exactly this:
"The faster the corner, the closer to the geometric line you should drive; the slower the corner, the more you need to alter your line with a later apex." Ross Bentley's rule, as summarised in Wikipedia's "Racing line"
You can watch that rule appear in the sandbox. Drag the corner angle down toward a long, fast sweeper (a gentle bend at big radius): the geometric line already carries huge speed, and the late apex barely helps or hurts. Now tighten it into a slow hairpin: the exit-opening trade becomes worth much more. Same with sequence. If the corner is followed not by a straight but by another corner, the constant-radius geometric line that preserves mid-corner speed can be the faster choice, because there is no long straight to cash the exit speed on.
What actually flips the verdict
Underneath both lines is the same reversal: you are not minimising distance, you are maximising the speed you can carry where it pays off most. Which frame you are in decides which line that is:
Slow corner → long straight
Give away the apex. A later, tighter entry opens the exit radius, lifts the exit-speed ceiling, and that speed compounds down the straight. Late apex wins.
Fast sweeper, or corner into corner
Nothing to cash a faster exit on, and mid-corner speed is precious. The constant-radius geometric line keeps the highest minimum speed. Drive closer to geometric.
And the true fastest line over a whole lap is neither heuristic applied corner by corner: it is a global optimisation across the entire sequence, where the exit you set up here is the entry you inherit into the next corner. The late apex is the right answer to "one corner, then a straight," not a universal command to apex late.
Not the brachistochrone
This looks like a cousin of the "fastest slide down a ramp" puzzle, but it is a different problem with a different constraint. There, gravity does the work and the limit is how the descent trades height for speed. Here there is no falling: the car is on the flat, the engine and brakes set the pace, and the hard ceiling is tyre friction. The curve you want is not a curve of fastest descent; it is the line that keeps the most grip-limited speed where a straight can use it.
The check: every number recomputed in front of you
Nothing here is stored. For the current settings, the speed ceiling and the lap-time delta are recomputed live from the two equations:
The offline gate reproduces all of these two independent ways: node research/fastest-line-through-a-corner/verify-fastest-line-through-a-corner.mjs.
Free choices & what is idealised (name them or the page overclaims):
- Flat corner. v = √(μ g r) assumes no banking and no aerodynamic downforce (downforce raises the grip ceiling with speed, which changes the numbers on a real racecar), a point mass, and one single grip value. The friction "circle" here is the equal-grip special case of the friction ellipse.
- The win needs a following straight. Stated as Ross Bentley's rule above; the sandbox shows the advantage vanish at zero straight and for fast sweepers.
- Model constants. Top speed is capped at 90 m/s and forward acceleration out of the corner at 3.5 m/s² (a power/engine limit; braking uses the full grip). These set the exact milliseconds, not the direction: a stronger engine or higher top speed shifts the break-even, it does not remove it. Track width is 11 m.
- Two-arc idealisation. The line is an entry arc joined to an exit arc; a real driver blends them with a transition. The radii and speeds are exact for this idealised geometry.
- Illustrative, not a constant. The often-quoted "the geometric line's entry advantage lasts ~100 m, the ideal line's exit advantage lasts ~500 m" (Driver61) is a teaching example of the exit lever, not a measured universal. This page computes its own break-even instead.
- Single-corner heuristic. The track-wide fastest line is a global optimisation over the whole sequence of corners; "apex late" is the answer for one corner into a straight.
What is exactly true here, and what is a model
Exactly true (mechanics). A grip-limited car on a flat curve of radius r cannot exceed v = √(μ g r), because the only thing bending its path is friction μ m g and it must equal the centripetal demand m v² / r. The mass cancels. Speed therefore scales with the square root of radius, so a wider radius is a higher speed ceiling. A larger exit radius lets the car both hold more speed and return to full throttle earlier. These are standard circular-motion and racing-line facts.
The lap-time model. The speed profile is a quasi-steady-state (QSS) simulation, the workhorse of lap-time engineering: cap speed at the grip ceiling along the sampled path, then a forward acceleration pass and a backward braking pass on the friction budget a_long = √((μ g)² − a_lat²) (capped at the engine limit for acceleration), then integrate ∫ ds / v. QSS omits transient weight transfer and yaw dynamics; it captures the steady balance that decides this question. The geometric line is defined here as the symmetric, single-radius member of the family (its entry and exit radii are equal); the late apex trades entry radius for exit radius.
Why the delta is fair. Both lines are timed from a common point far enough up the approach that each has already reached top speed, so the comparison does not depend on where the measurement starts. The only differences are in braking, cornering, and the straight.