Different machine, different numbers? We tested draw machines and ball sets
One idea keeps coming back on lottery forums: with a new machine, a different ball set or after maintenance, different numbers will come up. We checked it against real data. In short: there is no measurable difference.
The myth
The reasoning sounds sensible: no ball is perfect, every machine has its quirks. Maybe one ball is slightly lighter, maybe one machine mixes a little differently. In the UK, some players record which machine and which ball set were used at every draw, looking for "their" machine. It's a fair question. But what do the numbers say?
UK: every machine and ball set is on record
For the UK Lotto, we know for each draw which machine (Arthur, Guinevere, Lancelot, Merlin, Amethyst, Topaz, Sapphire …) and which ball set were used. We analysed all 3,262 draws since November 1994, splitting the two formats: 6 from 49 until October 2015, and 6 from 59 since.
For every machine and ball set with enough draws, we checked whether all numbers come up equally often (chi-squared test, corrected for drawing without replacement), and whether machines or sets differ from each other.
| Comparison | Draws | Result |
|---|---|---|
| 7 machines, 6 from 49 | 1,795 | no difference (p = 0.88) |
| 4 machines, 6 from 59 | 1,026 | no difference (p = 0.54) |
| 8 ball sets, 6 from 49 | 1,994 | no difference (p = 0.13) |
| 10 ball sets, 6 from 59 | 1,139 | no difference (p = 0.87) |
The practical test: picking by machine or ball set
The decisive test is the one forums suggest: you know which machine and ball set will be used next, and you play the 6 numbers that have come up most often with that machine or set so far. We replayed this for every draw, always using only the data known beforehand.
Across all draws, the strategy found 1,337 and 1,366 correct numbers where chance predicts 1,329 and 1,348 (6 from 49). For 6 from 59: 598 and 567 instead of 633 and 596. Knowing the machine and ball set doesn't help.
Germany: new machines and new balls since 2023
The biggest possible "change" happened with Germany's LOTTO 6aus49: since 18 January 2023, the draw uses new machines. The old ones were over 60 years old. The table-tennis balls (40 mm, 3.09 g) were replaced by plastic balls measuring 54 mm and weighing about 25 g, roughly eight times heavier. One ball set is used per draw with two more on standby; the balls are checked before every draw and the machines are serviced once a year.
We compared the 1,013 draws from May 2013 to January 2023 with the 390 draws since the switch. The frequencies of the 49 numbers don't differ significantly (p ≈ 0.12), and since 2023 all numbers have come up evenly. The average sum of the 6 numbers was 148.6 before and 147.7 after, against 150 expected. Balls eight times heavier, and still no visible effect.
What about maintenance, new balls, "fresh" machines?
Lotteries don't publish when a machine is serviced or a ball set replaced. But if such interventions shifted the numbers, machines and ball sets would differ from each other, and the complete replacement in 2023 would show up. Neither is the case. That is exactly what the checks are for: balls weighed and measured, test runs, and every draw recorded by a supervisor.
The machine, the ball set or maintenance change nothing about your chances. Every combination has exactly the same chance in every draw. Picking by machine is really picking at random, just with more effort.
Randomness test in the LabUK Lotto number frequenciesCan you audit a lottery?
Our calculations: chi-squared homogeneity tests corrected for drawing without replacement; walk-forward strategy test (prior data only); LOTTO 6aus49 4 May 2013 to 31 Oct 2026, UK Lotto 19 Nov 1994 to 31 Oct 2026.