How every derived number on this site is calculated, where the data comes from, and what each metric can and cannot claim.
Every number on this site that isn't a plain historical fact is calculated, and every calculation involves choices. This page documents those choices: the formulas, the edge cases, the things deliberately excluded, and the questions each metric can and cannot answer. If a number here can't be explained, it shouldn't be on the site.
Historical results come from F1DB, an open Formula 1 database released under CC BY 4.0. The version currently in use is v2026.10.0. Nothing on this site is hand-entered from memory or estimated — every lap count, grid position, and finishing order comes from that source.
Everything beyond the raw results — the f1δ Score, the rates, head-to-head records — is derived from that data by scripts run at build time. A cron job checks for a new F1DB release every three hours; when one is found, it pulls the update and re-derives all computed data automatically. Historical seasons don't change, so those figures are stable; current-season figures update as the season runs.
The derive scripts are: derive-f1.mjs (career stats and H2H), derive-f1delta.mjs and derive-f1delta-lenses.mjs (f1δ scores and their lenses), plus scripts for records, teams, circuits, grands prix, standings, people, and races. All are pure, network-free, and idempotent — they can be re-run any time the source data changes.
Every season is scored as a share of that year's maximum possible points, on a 0–100 scale, then summed across a whole career. Formula: season f1δ = (points scored ÷ maximum points winnable that season) × 100; career f1δ = the sum of every season. Worked example: Verstappen's 2023 = 575 ÷ 620 × 100 = 92.7. Why: dividing each season by its own era's maximum makes eras comparable — dominating a 7-race 1950 season and a 24-race 2024 season both come out near 100%, even though a win was worth 8 points then and 25 now. Summing rewards longevity. Indy 500 excluded; a DNF is simply a zero.
A driver's highest single-season f1δ — their most dominant year, on its own. Formula: the maximum of a driver's season scores. Worked example: Verstappen's peak is his 2023 season — 92.7. Why: the career total rewards longevity, so a short, ferocious peak can hide inside a lower lifetime sum. Peak isolates the single best year, so a driver whose brilliance was intense but brief stands where they belong.
Rank every driver by f1δ each season. Finishing top-3 that year is a seasonal podium; finishing 1st is a seasonal win. A dominant stretch is the longest run of consecutive seasons at top-3. Formula: for each season, rank all drivers by f1δ; seasonal podium = f1δ rank ≤ 3; dominant stretch = longest run of consecutive seasonal podiums. Worked example: Hamilton holds the record — 8 straight seasonal podiums (2014–2021). Schumacher ran 7 in a row (2000–2006); Verstappen's run stands at 7 (2019–2025). Why: dominance is where you placed among the field, not your raw score. Hamilton in 2016 lost the title by five points but still scored a top-3 f1δ — something a binary "champion / not champion" would miss.
A win was worth 8 points in 1950, 10 through the 1990s, and 25 today — and seasons have grown from 7 races to 24 plus sprints. The total points available in a season have gone from roughly 150 to over 600. Career points totals therefore say more about when a driver raced than how well.
f1δ scores each season as a share of what was actually winnable that year, then adds those seasons up. A dominant 15-race season and a dominant 24-race season both land near 100, because each is measured against its own ceiling.
Verstappen's 2023: 575 points from a possible 620 → 92.7. Schumacher's 2004: 148 from 180 → 82.2. Farina's 1950: 30 from 54 → 55.6. These figures are drawn from the same data pipeline that drives every driver page — they update automatically when the source changes.
Raw counts (29 wins, 91 wins) are shaped by how many races a driver entered. A driver who raced 250 times had far more chances to win than one who raced 100 times, even in the same era. Dividing by starts normalises for career length: a 30% win rate is a 30% win rate whether it came from 10 wins in 33 starts or 50 wins in 167.
A win in 1958 (9 entrants, no DRS, no tyre compounds) and a win in 2024 (20 entrants, every team a billion-dollar operation) are both wins in the record books. These rates treat them equally. The f1δ Score — which divides by the maximum points winnable that season — is the better cross-era comparator; rates are most useful within a single era or for comparing two drivers who raced at the same time.
Pole position is counted where the driver started from the front row as a result of qualifying — not from a pit lane start, not from a grid penalty reversal. Formulas: win % = wins ÷ race starts × 100; podium % = podium finishes ÷ race starts × 100; pole % = pole positions ÷ qualifying entries × 100.
Win %, podium %, and pole % normalise for career length — a driver with 250 starts had more chances than one with 50 — but they do not normalise across eras. The f1δ Score is the better cross-era comparator; the rates are most meaningful within a single era or when comparing drivers who shared a grid.
The share of race starts that ended without a classified finish. Formula: races not finished ÷ race starts × 100. Races the driver did not start are excluded from both sides. A disqualification counts as a finish that scored nothing, not a DNF. Why it sits apart from the era-fair rates: unlike a win or a podium, finishing meant something very different in different eras. Cars in the 1950s–70s broke constantly; modern cars almost always reach the flag. A high DNF rate in 1955 says far more about the machinery than the driver. Comparing two drivers within the same season is meaningful — they faced the same conditions. Across eras it's context, not a verdict.
The share of race starts that ended in a points-paying position. Formula: races finished in a scoring position ÷ race starts × 100, using that season's real points system. Why it's era-dependent: the number of positions that pay points has changed repeatedly — the top five scored in 1950, the top ten score today. A modern driver has roughly twice as many scoring positions available for the same relative performance, so a higher percentage doesn't necessarily mean a better driver. Like DNF rate, it's shown as context.
These two rates are deliberately shown apart from the era-fair rates, because unlike a win or a podium, their meaning has changed dramatically over time. The gradient is stark: Fangio failed to finish 24.1% of his starts; Verstappen, 12.8%; Hamilton, 7.7%. That difference is mostly machinery, not driver.
This is why, on the compare tool, the reliability rows only show a winner in season-vs-season mode — when both drivers raced under the same conditions.
The career-shape chart on the compare tool can align two careers three ways. Each answers a different question:
| Alignment | Question it answers | X-axis |
|---|---|---|
| Career season (default) | "Whose trajectory was better?" | 1st season, 2nd, 3rd… |
| Age | "Who was better at 27?" | Driver's age that season |
| Calendar year | "What was happening that year?" | Real timeline |
Lining two careers up by season number seems neutral, but it isn't. Fangio's first F1 season came at age 39; Hamilton's at 22. Aligned by season number, a rookie year at 39 sits directly against a rookie year at 22 as though they were the same thing. Age alignment corrects that. In calendar alignment, careers that never overlapped appear as two non-overlapping series — that's accurate, not a bug.
Age is computed as the calendar year of the season minus the driver's birth year — the same convention used by Baseball-Reference, the North Star for this site's analytical approach.
A teammate H2H page exists for every pairing where a meaningful race comparison is possible. "Meaningful" has a precise definition:
A pairing gets a page only when the combined total of default plus DNF-comparable races is at least one. Pairings where the two drivers shared a team but no qualifying race comparison exists have no page.
f1δ measures how much of the available success a driver actually took, in their own era. It does not claim to say who would win in equal machinery, or who was more talented.
A dominant season reflects a driver and a car — Formula 1 has never been a single-variable sport, and a great driver in a midfield car will always score lower than an equal driver in the best car of the year. No points-based metric can separate those, and this one doesn't pretend to.
What f1δ does remove are the distortions that make raw points meaningless across eras: different scoring systems, different calendar lengths, different numbers of points-paying positions. What's left is an honest measure of how completely a driver beat the field in front of them.
If a number here looks wrong, it may be. The data pipeline covers 75 years of records and is automated — errors are possible. Reporting one is welcome. See the About page for contact details.
f1δ Career Leaderboard · Peak f1δ · Dominant Stretches · Compare Drivers · About