Arcade handling is the part of a racing game players judge in the first ten seconds, and it is also the part small teams tend to tune last, usually by feel, usually the week before a demo. The result is a car that one designer loves and everyone else spins out of.

This is the checklist we would want a small team to work through before calling handling "done". It assumes an arcade or sim-cade feel rather than a full simulation, and it works whether you are on Unity, Unreal or a custom engine.

1. Write down what the car should feel like, in words

Before touching a single number, agree on three or four sentences: how quickly it reaches top speed, whether it drifts on its own or only on a button, how forgiving a wall tap is, and whether the camera or the car "sells" the speed. Without this, tuning sessions turn into arguments about taste.

2. Separate input shaping from physics

Most handling problems are really input problems. Keep these as distinct, tweakable layers:

  • Dead zone and response curve on the stick or triggers.
  • Steering speed (how fast the wheels reach full lock) and steering return speed.
  • Speed-sensitive steering, so full lock at 20 km/h is not full lock at 200 km/h.

If these are baked into tyre friction values, you will never be able to tune keyboard, pad and touch separately.

3. Test on every input device you ship on

A pad with analogue triggers, a keyboard with digital steering and a phone with tilt or touch buttons all need their own profile. Keyboard players in particular need steering that ramps in rather than snapping to full lock, or every corner becomes a series of taps.

4. Build a fixed handling test track

Not a real level. A flat, cheap greybox with:

  • a long straight (top speed and stability),
  • a hairpin (low-speed rotation),
  • a fast sweeper (high-speed grip),
  • a chicane (weight transfer, left-right-left),
  • a wall to scrape (collision response).

Log lap times and a short note for each tuning change. When someone says "it felt better yesterday", you can actually check.

5. Decide how forgiving collisions are

In arcade racers, a glancing wall hit that kills all your speed feels like a punishment for the track, not the player. Decide up front how much speed a scrape costs, whether the car is nudged back parallel to the wall, and how quickly control returns after a crash or a spin.

6. Tune the camera with the car, not after it

Field of view, camera lag, a little roll in corners and a speed-based FOV kick do as much for perceived speed as the actual velocity. Tune them in the same session as the handling, on the same test track, or one will keep fighting the other.

7. Watch new players, not the team

The team has hundreds of laps of muscle memory. Hand the pad to someone who has never played it, say nothing, and watch the first two minutes. Note where they brake, where they hit walls and whether they ever find the drift. Two or three of these sessions are worth more than another week of internal tuning.

8. Keep the HUD readable at speed

Speed, position and the next corner need to be readable in a glance while the eyes are on the road. We wrote up a short HUD readability checklist that applies just as well to racing UI.

9. Lock a handling version before content production

Once tracks are being built, every handling change can break a corner that a level designer has already balanced. Version your handling profiles, freeze one before track production ramps up and treat later changes like any other risky change: reviewed and replayed on the test track.

Wrapping up

Good arcade handling is rarely one magic number. It is clear intent, separate input and physics layers, a repeatable test track and honest playtests with people who have never seen the game. If you are planning a racer and want a team that has worked across genres, our racing game development page outlines how we approach it.

What does your team put on its handling test track? I would be curious to hear what we have missed.