# Setups and Your Sessions

> How Braking Lab links a setup to the session you drove, how to confirm, correct or associate it yourself, and how your verdict becomes evidence.

A setup is only worth what it does on track. When you drive with the desktop
client running, Braking Lab records the setup that was on the car alongside the
telemetry — so a run stops being "a session" and becomes "this version, tested".

> Setups is in **Public Beta** and requires a personal **Plus or Ultra** plan. A Team seat alone does not grant access.

## Automatic Detection

Capture stores a snapshot of the car's setup with each session. For cars with a
**reviewed identity adapter**, Braking Lab compares that snapshot with your
vault. No Le Mans Ultimate car has a reviewed adapter yet. For every other car,
[associate the setup by hand](#associating-by-hand).

When exactly one version in your vault has the same values, Braking Lab proposes
it. The session detail then shows **Setup used in this run**:

| Field | What it tells you |
|-------|-------------------|
| **Source** | Detected, Added manually, or Declared by Race Engineer |
| **Match** | **Exact values** for a detection; **Declared by you** for an association you made |
| **Confidence** | A percentage with a Low, Medium or High band |

A detection is only a proposal. It does not count until you confirm it. When no
version matches, or several versions share identical values, the card says so and
nothing is proposed.

The match is on **values, not names**. Renaming a setup in the garage does not
break it, and two setups that happen to share a name do not merge.

The setup's [type and personal tags](/en/docs/ai-setups/types-and-tags) are not
part of that identity either. They help you organize and find it, but they never
confirm which version you drove or make a session count as evidence.

### The dials you turn in the car do not rename your setup

Brake bias, traction control, engine braking, fuel — the things you move
mid-stint — are read and shown, but they are deliberately **not** part of the
identity Braking Lab matches on. They have to be: they change while nobody edits
the setup, so counting them would split one unchanged setup into a new identity
every time you touched a dial, and the exact match could never fire.

Measured on a real corpus of eight synced snapshots of one iRacing **fixed**
setup — a setup that cannot even be edited in the garage — the brake-bias dial
alone (55 / 54 / 53) split seven captures of that one setup into three
identities, and the garage export matched none of them.

### Uploading after you drove

Detection also works backwards. When you upload a setup, Braking Lab re-checks
the sessions you already synced for the same car. A session whose snapshot
matches exactly one version, and that has no association yet, gets a proposal
in the **Needs confirmation** state. This only happens for cars with a reviewed
identity adapter. It never confirms anything for you.

## Confirming, Correcting, Dismissing

Under **Setup used in this run** you can:

- **Yes, this version** — opens the association review for the detected version. When you confirm it, the run counts as a test.
- **Choose another** — pick the version you actually ran. Only your versions for the same simulator can be chosen. Compatible means the same simulator and car identity; the track does not need to match.
- **Not in my library** — the setup was never uploaded.
- **I changed setup mid-session** — the run used more than one setup.
- **Dismiss match** — the detection is wrong.

The last three all reject the detected association. Braking Lab does not store a
separate reason for each. A rejected match means exactly that: *this session will
not count as a confirmed test of that setup.*

### Mixed sessions

If you change setup mid-session, the captured snapshot describes the **last**
state, not every stint. When Capture sees more than one setup state in a session,
the proposal carries a **Mixed session** label and a warning that the final
snapshot may not describe every stint. Check the version before you confirm it.

If you know which setup was on the car for which laps, reject the detected match
with **I changed setup mid-session** and associate each part as a lap range by
hand.

## Associating by Hand

Not every run gets detected: a car without a reviewed adapter, an ambiguous match,
a mixed session. You can state it yourself from two places:

- In the session detail, click **Associate setup**.
- In the setup detail, click **Associate with session** on a version.

Choose the setup, the immutable version and the session, then where you used it:

- **Whole session** — the setup stayed on the car for the full captured run.
- **Inclusive lap range** — the first and last captured lap where you used it.
  A lap range needs telemetry with a verified lap order, which requires
  **Capture 1.3.8 or later**. A session with no laps, or with too many laps to
  list safely, can only be associated as a whole.

The lists show up to your 50 most recent sessions, and up to your 50 most
recently updated setups and versions, in that simulator.

Click **Review association**. Braking Lab re-checks ownership and compatibility
before saving, and it is honest about what it cannot verify:

| What it says | What it means |
|--------------|---------------|
| **Compatible** | Setup and session agree on the car |
| **Identity needs confirmation** | The car identity is incomplete or ambiguous — you confirm you know it was on the car |
| **Incompatible** | Two known car identities that do not match. This one cannot be saved |

A track mismatch is shown but never blocks: testing a setup at another circuit is
a legitimate thing to do.

If the selection overlaps active associations, the review lists them. To save,
confirm that all of them may be replaced. Each overlapping association is
replaced completely and kept in the history; a partial overlap is not split into
smaller ranges. If the exact same association already exists, **Use existing
association** keeps it and creates no duplicate. Then click **Confirm
association**.

### Retiring an association

When you no longer know which setup you used, click **I don't know which
setup** on the active association. This retires it. The association and any
evaluation stay in the audit history; nothing is deleted.

The session shows every association with its state: **Needs confirmation**,
**Active**, **Retired** or **Replaced**.

### Through your Race Engineer

Your engineer can also prepare an association: it shows you a preview with every
warning, and saves only after you confirm again. Braking Lab collects that
confirmation directly from you through your MCP client, never from the engineer.
This needs an AI client that supports MCP form elicitation. If yours does not,
nothing is written — make the association in the app instead. Retiring an
association through the engineer works the same way.

## From a Run to Evidence

Once a run is confirmed you can rate it, and the rating is about the **setup on
track** — not about the Race Engineer's recommendation that may have produced it.

Start with the quick verdict — *Worked for me* / *Needs another change* — and add
detail when you have it: how this run compares with an earlier one (**Better**,
**About the same**, **Worse**, **Inconclusive**), what influenced that call (pace,
consistency, entry stability, rotation balance, exit traction, tire behavior,
race-stint suitability… fifteen in all, up to twelve per verdict), and a private
note.

The note is saved with your evaluation and is never sent to analytics. Your Race
Engineer can read it when it reviews your experiment history, so it passes
through your AI client's provider. Write it with that in mind.

Until you decide, the run reads *No verdict yet — this setup is still being
tested*, which is a truer state than a blank.

### The objective half

Confirm the same setup in **two** runs and Braking Lab puts their telemetry side
by side — baseline versus test, computed from your own lap data. It reports
**test minus baseline, and stops there**: no winner, and no claim that the setup
caused the difference. Requirements, limits and how to save a comparison are in
[Measuring a setup change](/en/docs/ai-setups/measuring-changes).

## What the Race Engineer Can Do With This

Ask it, and it reads the same evidence: which versions you have tested, what you
said about each, and your saved comparisons. That is how a remix stops being a
guess from a blank page and becomes the next step in a line of experiments — see
[Remixing setups](/en/docs/ai-setups/remix).

## Related

- [Measuring a setup change](/en/docs/ai-setups/measuring-changes) — the objective comparison between two confirmed runs
- [Comparing setups](/en/docs/ai-setups/comparing) — two setups side by side, parameter by parameter
- [Capture](/en/docs/capture/introduction) — the desktop client that records the snapshot in the first place