Law of thirds
How many numbers to expect in a cycle
The law of thirds observes that, over a number of draws equal to the numbers in play, on average about two thirds of the numbers appear at least once and one third lags behind. It's a statistical property of sampling, not a rule about which numbers will appear: it measures how far the recent archive departs from the theoretical expectation.
- Expected vs observed: theoretical against actual counts per cycle
- Normalised deviation index per game and per wheel
- A descriptive reading of the archive, not a predictive one
Cyclometry
Numbers on the 1-to-90 circle
Cyclometry places the 90 numbers on a circle and studies their geometric relationships: counter-figures (opposite positions), complements to 90, symmetric triplets at 120°, circular distances and cyclometric tens. These are deterministic transforms that group numbers into the recurring figures of Lotto tradition.
- Counter-figures, complements and symmetric triplets computed on the fly
- Circular distance between numbers and cyclometric tens
- Deterministic figures: same input, same verifiable output
Delays & frequencies
Reading the archive: delays, frequencies and relationships
On the official ADM archive we compute delays (draws of absence), frequencies per period with a monthly heatmap, the most recurring pairs and triplets, moving tens, plus tools such as an order-1 Markov chain and the chi-square goodness-of-fit test. To measure how far a number departs from the norm we use the Z-score: they are different lenses on the same data, useful to describe patterns — not to guess them, because draws remain i.i.d..
- Delays and frequencies per game, wheel and period
- Pairs, triplets and moving tens with historical counts
- Order-1 Markov and chi-square to measure regularity and anomalies
The math detail, in brief
The order-1 Markov chain estimates the transition probability from one state to the next using only the previous draw: handy to quantify whether short-range dependencies exist, which on a truly random process should not emerge.
The chi-square goodness-of-fit test compares observed frequencies with those expected under uniformity: a high value flags a departure, but on its own it is not proof — it must be corrected for multiple testing.
When you test many combinations, some "anomaly" shows up by pure chance. We apply corrections (e.g. Bonferroni / FDR) before deeming a pattern significant: after correction, no exploitable predictive edge survives.
What these methods do NOT tell you
These analyses describe the archive: they do not indicate which numbers will appear nor make the game profitable. Draws are independent and, after correcting for multiple testing, no exploitable predictive edge emerges. The game's expected value stays negative.