They say

A speed figure rates how fast the horse ran, not where it finished — so a horse can win slowly (low figure) or run fast in defeat (high one). The scale is built so runs at different tracks, trips and going are directly comparable.

Par. For each track + distance + going we take the median winning time across years of races — the time a normal winner clocks there. Each runner's time is estimated as the winner's time plus its lengths behind (converted to seconds at that race's pace). The figure is then:
Speed = 100 − 2 × (lengths slower than par). So 100 = par-winning pace, and roughly every 2 points = one length. Above 100 is faster than the standard winner, below is slower.

From this horse's best winning run 05 Oct 2018 · Greyville 1400m (Standard): it won in 84.53s against a par of 83.4s — about 7.8 lengths slower than the standard winner → Speed 84.

Estimated from the winner's time plus beaten lengths (individual clockings aren't published), so it carries some margin of error. It doesn't account for in-race pace or wind. Races with an unreliable winning time (more than ~25 lengths off par) are dropped, beaten lengths are capped at 20, and figures are bounded to 30–140.

Weight-adjusted line. The dashed line normalises each run to a reference of 58 kg carried — the approximate level-weight standard in South African handicap racing. A horse that carried more gets credit (figure rises); one on a lighter load has the advantage removed (figure falls). Conversion: 1 length per 2 kg per 2000 m — so 62 kg in a 1600 m race = +2.5 pts; 55 kg = −1.9 pts. Runs with no weight on record stay at the raw figure.

This is a level-stakes simulation: imagine backing this horse with the same 1-unit bet every time it runs, at the starting price (SP). The line is the running total — rising when it wins at a price that outweighs the losing runs, falling on a losing streak. It asks whether blindly backing the horse would have made money, a tougher test than its win count alone.

Each run settles as:
win → + (decimal odds − 1)  ·  lose → − 1.
So a win at 5/1 returns +5.0 units, a win at 2/1 returns +2.0, and every loser costs −1.0. Fractional SPs (e.g. 9/4) are converted to decimal — 9÷4 + 1 = 3.25 — before settling.

This horse's record 24 runs with a priced SP, staking 24 units in total → a profit of +21.0 units, an ROI of +87.5% (profit ÷ total staked).

Only runs with a recorded SP are counted, so this bet total can be lower than the horse's total runs. It assumes you back it at SP every single time, with no selectivity — a baseline, not a betting tip. It ignores Rule 4 deductions, dead-heats and any each-way element; a flat or rising line is the signal worth noting.