Methodology · How EVStreak knows

Fair Score

Your Fair Score is a weighted average of four factors. It is ready-only and re-normalized: a factor without enough data is left out of both the top and bottom of the average entirely — it is never scored 0 and never quietly drags your number down.

Drive detail completeness: Presence ratio · How many recent completed drives carry distance, duration, efficiency, and pace. A complete trip picture, not a guess.
Energy rhythm: Variation (capped) · How steadily the car uses energy across recent drives — a coefficient-of-variation score, capped so one outlier trip can't dominate.
Trip rhythm: Variation (capped) · How repeatable your average pace is across recent drives — the same capped variation transform, separating a pattern from a one-off.
Energy detail: Presence ratio · Whether EVStreak has both energy-used and regen totals, so it can explain the number instead of inventing one.

The bars above are illustrative of the recipe weights, not an owner stat — they visualize how the four factors are blended, never a personal score.

The score only goes live once you have at least 3 completed drives, and each individual factor needs at least 3 drives of its own data before it contributes. Until then you see an honest "not enough data yet," not a placeholder number.

How the average is computed

EVStreak takes each ready factor's 0–100 value, multiplies by its weight above, sums them, and divides by the total weight of the ready factors only. Presence factors (completeness, energy detail) measure how much of the required data is present. Variation factors (energy rhythm, trip rhythm) score steadiness using a coefficient of variation, capped at 0.5 so a single unusual trip cannot swing the result. No factor is ever assumed, defaulted, or back-filled.

Source · weights and floors read directly from the live Fair Score deriver in EVStreak (@evstreak/shared) when this page renders. Telemetry-derived from your authorized drives.

Vehicle Health · six axes

Each Vehicle-Health axis has a real refuse-below threshold. We would rather show "not enough data" than a number we can't stand behind. Here is every axis and exactly when it speaks.

Mileage & age

Speaks now

Odometer now; rate waits for a baseline

Odometer and connection age are shown as soon as we have them. EVStreak does not annualize lifetime odometer against account age because that can mix pre-EVStreak miles into a short connection window. A future mi/year rate needs a baseline odometer at connection plus later telemetry; 14 days is only the stability floor once that baseline exists.

Charging hygiene

Waits for data

Real once there are enough sessions

"% charged to ≤80%" and supercharge frequency are derived from your real charge sessions — but only once there are at least 5. Below that floor the readout would be noise, so the axis refuses instead of guessing.

Battery health

Waits for data

Tesla-blocked — honest owner-paste path

The in-car Tesla Battery Health Test is NOT on the Tesla Fleet API, so EVStreak will not invent a state-of-health number from range or charge level. Paste your in-car test result and we surface it as Tesla-verified; otherwise we say so plainly. A capacity trend needs at least two readings.

Maintenance forecast

Speaks now

Derived from real intervals

Service reminders are projected from your real mileage trend and published Tesla service intervals — never a scare tactic, never an upsell.

CO₂ impact

Waits for data

Disclosed estimate above the kWh floor

A CO₂ comparison needs an honest energy basis, so it stays "not enough data" until lifetime charging passes 50 kWh. Above that we publish a disclosed estimate using a stated grid intensity and gasoline baseline — the assumptions are shown inline, not buried.

Charging cost

Speaks now

Real receipts; disclosed default otherwise

Cost uses your real charge energy. When a session has no price attached we apply a single disclosed default rate and label it as such — we never silently fabricate a dollar figure.

Source · the numeric floors (14-day stability floor for future baseline-backed mileage rate, 5 sessions, 50 kWh) are the same constants the Vehicle-Health methodology uses — one shared source, so this page and the app always agree.

How we label every number

No competitor publishes its data-trust vocabulary. Ours is fixed and visible: every number EVStreak shows carries one of these five badges, so you always know how much weight it can hold.

Tesla-verified
Straight from a Tesla first-party source — a Battery Health Test, the Tesla app, an official receipt.
Telemetry-derived estimate
Computed by EVStreak from the authorized telemetry frames your car already streams.
Range-derived estimate
Inferred from range or energy fall-off patterns — lower confidence than telemetry, and labeled that way.
Cohort comparison
Relative to other EVStreak vehicles in your segment, only when the group is large enough to be fair.
Not available for this vehicle
The source does not exist for this vehicle. We refuse to render a number — never a 0, never a mystery dash.

The freshness and provenance axes

Alongside the badge above, every owner-facing reading also carries a freshness state (fresh, stale, asleep, never seen, missing permission, unsupported, provider error, private, or pending live proof) and a provenance tag (live telemetry, imported, derived, or mixed). EVStreak determines these on the server from the real data — the interface only renders the verdict, it never guesses it.

The premium move

What we refuse to claim

The most honest thing a product can publish is its limits. These are ours — stated as promises, not apologies.

We don't fake battery state-of-health. The in-car Tesla Battery Health Test isn't exposed on the Fleet API, so we will not reverse-engineer one from range or charge level. Paste your real in-car result and we show it as Tesla-verified; otherwise we tell you it's unavailable.
We don't overstate autonomy. EVStreak reports what the telemetry counters actually say about supervised driving features. We do not editorialize them into capability claims the car never made.
We don't invent dollars. Charging cost comes from your real energy and your real rates. When a price is missing we apply one clearly-disclosed default and label it — we never present a made-up figure as fact.
We don't fill blanks with zeros. When a source doesn't exist for your vehicle, the value is marked unavailable. You'll never see a confident-looking number that's secretly a placeholder.

This page is maintained alongside the EVStreak codebase. Every number above is read from the live product when the page renders, so the methodology and the app cannot disagree. EVStreak is an independent product and is not affiliated with Tesla.