How the numbers work

Every figure on this site is a claim, and a claim is only worth as much as the method behind it. This page sets out where the coverage is thick and where it is thin, how one driver stays one driver across 78 years of timing sheets, what the ELO rating is actually counting, and which columns are measured rather than reconstructed.

What the archive covers

Momentum covers circuit racing in New Zealand: races run at permanent circuits, and at the airfield and street circuits used earlier in the period: Ardmore, Wigram, Ohakea, Renwick, Dunedin, Waimate and Wellington among them. It does not attempt to cover rally, speedway, drag racing or off-road.

As at the date at the foot of this page the archive holds 2,007 meetings across 31,133 sessions, 13,307 drivers and 3,987,306 recorded laps, spanning 1949 to 2026 and 42 circuits and layout configurations.

Coverage is not evenly distributed across those 78 years, and this matters for how you read anything cross-era. From roughly 2000 onward the record is consistently detailed: meetings arrive from transponder timing with lap-by-lap detail for every car, individual lap times, positions, and sector splits where they were published. Earlier meetings survive as a classification (who entered, what they drove, where they finished, often a best lap), and for the earliest decades a meeting may be a single result sheet. Neither is more or less accurate than the other; there is simply far more of it per meeting as you come forward in time.

The practical consequence: statistics that count laps, positions gained or lap-by-lap behaviour are effectively modern-era statistics, because the underlying rows only exist where the timing captured them. Counts of races, wins and finishing positions go back the whole way.

Identity: how one driver stays one driver

A single competitor appears across the source data as many different strings. The timing system appends a driver code, so "Blair McDonald" arrives as "Blair McDonald MCD". One meeting has them in capitals, the next in title case, a third with a double space or a middle initial. Names are misspelled and later fixed. Classes are renamed every few seasons; the same car is entered under three descriptions.

So each driver, car and class has a canonical record, and the raw strings are resolved onto it by exact normalised matching plus a curated list of known aliases. Two rules govern this:

• Matching is exact, never fuzzy. Similar names are not assumed to be the same person. There are several distinct drivers on this site whose names differ by one character, and merging them would destroy two careers to tidy up a list. Where a mapping is uncertain it is queued for a human decision instead of guessed.
• The original strings are never overwritten, and every merge or rename is reversible. An archive that rewrites its own source data cannot show its workings.

Two-driver entries in enduros are held as a pairing: the entry belongs to both drivers, with a seat recorded for each. Because a pairing is not a single person, some per-driver statistics deliberately exclude co-driven entries rather than credit a shared result to one driver; where that happens the page says so.

ELO driver ratings

A finishing position only tells you who was quickest on the day, not who you beat. Winning a ten-car club race and finishing sixth in a national championship field are not comparable as positions, but they are comparable as results against the drivers who were there. That is the problem an ELO rating solves, and it is the same mechanism used for chess ratings.

Every race a driver finishes moves their rating: up for finishing ahead of drivers the rating did not expect them to beat, down for the reverse, and barely at all for finishing where the rating already predicted. Because the calculation is made against the drivers who were actually on the grid, a result that was genuinely hard to achieve is worth far more than a win over a field the driver was always going to beat.

The model also takes account of the strength of the field, of retirements, of grid position (reverse grids, ballots and handicap starts are common in New Zealand club racing, so starting at the back is not treated as a performance the driver chose), and of how few races a driver has to their name. The individual weightings are not published. They are recalibrated as the archive grows, and every rating is then recomputed from the full history rather than carried forward, so a rating is always the current model's reading of a whole career rather than a running total.

A rating for a driver with only a handful of races is treated as provisional and pulled toward the average, because a two-race sample is not evidence of a level.

Two things the rating is not. It is not a measure of outright speed; it counts results against opposition, so a quick driver in a weak field and a mid-pack driver in a strong one can rate similarly, which is the intended behaviour. And it is not a cross-era ranking: ratings connect drivers through the fields they shared, so drivers who never raced anyone in common are only loosely related, and comparing a 1960s rating with a 2020s one is not something this data supports. The deep dives go into what the rating predicts well and where it is weakest.

Strength of field

Each session carries a strength of field figure: the average rating of the opposition, taken before the race is scored so a driver's own result cannot inflate it. It is the number that makes "won a race" interpretable, and it drives the competition-level readings on driver pages. A high strength of field means the drivers present were rated highly, which is a statement about the entry list, not about lap times.

Lap records

• Records come from races, not from practice or qualifying. Qualifying laps are shown where they exist and are labelled as fastest laps, never as records: a driver on a single flying lap with a clear track is not doing the same thing as a driver in a race.
• Records are scoped to the exact circuit configuration, not to the venue. Several circuits here run more than one layout, and the layouts differ by more than a kilometre in places. A lap time does not transfer between them, so a record is always stated against the configuration it was set on.
• Times are shown as they were timed. The original string is what the timing system published, and it is never rewritten; comparisons and sorting use a numeric copy alongside it.
• Implausible times are held back rather than published as records. Timing sheets occasionally contain a value that cannot be a lap of the circuit it is filed under. Where a candidate record fails a plausibility check against the circuit it is not quoted as a record.

Measured versus derived

This is the distinction most worth understanding, and the site flags it rather than blurring it. Some columns are recorded by the timing system. Others are reconstructed for the meetings whose source never carried them, calculated from the lap times we do have:

• Lap-by-lap position, gaps and intervals are reconstructed for the older sessions whose timing output recorded only lap times. The method is validated against sessions that do carry those fields, but a reconstructed position is still a calculation and is flagged as one in the data.
• Pit stops and safety-car laps are inferred from the shape of the lap times. For older meetings this is the only way those events can be identified at all, so each one carries a confidence value rather than being stated flatly.
• Estimated distance uses one representative lap length per venue, so any kilometre figure is an estimate and is labelled as one wherever it appears.
• Weather is the recorded conditions for that place and date from a historical weather archive, accurate for the location and day but not a trackside measurement.

Nothing derived is presented as a timed measurement, and no derived value is ever promoted into a lap record.

Live timing

Live timing during a meeting is a provisional feed straight from the timing system. It is useful while the racing is happening and it is not the official result: it carries no penalties, classifications can change after the flag, and the driver names are the raw strings the feed publishes rather than resolved identities. Personal-best and record comparisons shown live are scoped to the circuit configuration in the same way as everywhere else, and are suppressed where the configuration cannot be determined. Once the meeting is imported, the official results supersede it.