Winning Lottery Combinations: Pairs, Triples and the Math of Coincidence
When players look for "winning combinations," they imagine a secret formula — a set of numbers that somehow wins more often. The reality works differently: the archive shows which combinations have already won, and that is not the same question as "which combinations will win tomorrow." Yet the archive is not useless. It provides structure, ranking and context for a meaningful choice. Let's look at which tools on the site help you work with this structure, which combinations really do stand out from random chance, and where self-deception begins.
"Winning" in the archive and "winning" tomorrow are different questions
In the France Loto draw (5 from 49) thousands of draws have been held over the decades. Some six-number combinations appeared there more often, others less often. That is a fact about history — and nothing more. Tomorrow's draw is statistically unrelated to yesterday's draw: the drum does not remember what came up, and any combination of the five main numbers has a chance of about 1 in 1,906,884 regardless of previous outcomes.
But that does not mean analysis is pointless. The archive is useful for two reasons. First, it helps you tell a plausible pattern from a statistical illusion — our brains see patterns in random data, and without checking them against the math it is easy to believe your own inventions. Second, the archive sets the context of choice: which combinations are historically "normal" and which are marginal. For more on the analysis methods themselves, see the pillar article on the 20 methods of lottery analysis.
One important note about bingo-style lotteries: in such games the numbers are pre-printed on the ticket and the player does not choose them. There, analysing winning combinations is meaningless as a selection tool — you cannot "bet on a frequent pair." Everything we discuss below applies to number lotteries where you fill in the field yourself.
Pairs and triples of numbers: frequent links and z-score
The Pairs section shows how often each pair of numbers appeared in a single draw. In a 5-from-49 game there are 1,176 possible pairs (this is C(49, 2)). Each draw contains 10 pairs (C(5, 2)). Across roughly 8,000 draws we have about 80,000 "pair hits" in the history. The expected frequency of any single pair is about 68 times.
In practice the spread around this average is wide, and some pairs appear noticeably more often. This is where the z-score comes in — the normalised deviation from the mean. A pair with a z-score of 2 or higher appeared significantly more often than expected. The larger the archive, the more strongly random spikes are "filtered out."
But there is an important catch. With 1,176 possible pairs and a significance level of p < 0.05, purely by chance about 59 pairs will be flagged as "significant" in any fair lottery. This is called the multiple comparisons problem. So a plain z-score ≥ 2 is a weak filter. Two reliable approaches: the Bonferroni correction (divide the significance threshold by the number of tests — for 1,176 pairs that is p < 0.00004, roughly z-score ≥ 4) and the consensus of several slices of the archive — checking whether a pair stands out in both the first half of the history and the second.
With triples the math is even harsher. In a 5-from-49 game there are C(49, 3) = 18,424 possible triples. Each draw contains C(5, 3) = 10 triples. Over 8,000 draws that gives 80,000 observations — on average about 4.3 times per triple. Some triples appeared 9-10 times, others 1-2. The trap is that with 18,424 triples and a threshold of p < 0.05 we expect about 921 "significant" triples purely by chance. In other words, finding a dozen frequent triples just describes noise rather than uncovering a pattern.
This means triple analysis on its own yields little. It is more useful in combination with pairs: if a triple includes a pair that the frequent pairs view marks as consistently frequent, the signal becomes denser. Another use is comparing different lotteries: in a 5-from-50 game such as South Africa PowerBall there are slightly more triples (C(50, 3) = 19,600), and the overall picture comes out with a different texture than in France Loto.
Top combinations: fully repeated tickets
The Top Combinations section looks not for pairs or triples but for matches of entire draw combinations. In a 5-from-49 game there are 1,906,884 possible combinations of the main numbers, so over 8,000 draws the chance of meeting an exact repeat is not small — the expected number of repeat pairs equals n²/(2 × N), which for n = 8,000 and N = 1.9 million gives roughly 17 cases.
These repeats are more useful as a diagnostic than as a strategy. If a repeat is found, it has already won — twice. But that does not mean it will come up again tomorrow: each individual combination still has those same 1 in 1.9 million odds. On the other hand, repeats show that the drum is working fairly: in a "manipulated" lottery the distribution would be different.
In US Powerball, where the second drum (a Powerball from 1 to 26) brings the total to nearly 292 million combinations, exact repeats almost never occur — the archive is smaller than what is needed for a statistically likely coincidence. This limitation has to be kept in mind: in smaller lotteries the tool is more informative, in larger ones it is almost empty.
"Pretty" combinations and myths about the drum
A common belief: combinations like 1–2–3–4–5 or 10–20–30–40–50 come up "less often than usual" because "the drum avoids them." The Consecutive Numbers section shows the statistics for such cases, and the results are stable from draw to draw. The mathematically boring truth: a consecutive combination has exactly the same probability as any other. The drum cannot tell "pretty" numbers from "ordinary" ones — the physical model contains no notion of beauty.
Psychology creates the confusion. The combination 1–2–3–4–5 looks "special," so its rarity seems like an anomaly. In reality it is exactly as rare as 7–13–22–28–33 — the second one simply is not remembered as "special." This is also confirmed by probability theory: ball 1 has the same chance of coming up as ball 37.
This myth has a practical consequence that is rarely considered. If you bet on 1–2–3–4–5 and win, you will have to split the prize with hundreds of other players who bet on the same "obvious" combination. The same effect applies to "pattern" tickets: diagonals, crosses, birthdays. For the probability of winning itself this makes no difference, but for the size of the win it is substantial. That is exactly why it is worth choosing combinations that are not too similar to the "obvious" ones.
Looking up a specific combination and expanded bets
To test a single hypothesis — "did my combination ever come up?" — two tools work: the archive lookup and its alternative version combination-lookup-2. You enter five numbers, see whether they occurred in the history, and if so — in which draws. Useful both for idle curiosity and for a quick check of a "too convenient" combination.
If a single combination is not enough, there are expanded bets. The idea: you play not one combination but a system — all possible combinations from an enlarged set of numbers. For example, you pick 7 numbers instead of 5 and play 21 combinations of 5 from 7. The chance of the main prize grows proportionally (by 21 times), but so does the cost — the bet becomes 21 times more expensive.
Expanded bets make sense to combine with pair and triple analysis: if your set of seven contains 2-3 "strong" pairs, the proportion of significant combinations inside the system will be higher than random. This is no guarantee of a win, but it is a way not to scatter bets on combinations that are weak from the start.
Practice: building a ticket from analysis
Here is an algorithm that logically ties together everything said above. It does not raise your odds above the mathematical ones, but it removes random noise from the selection process.
Open Pairs and pick 2-3 links with the highest z-score. Check that they keep their position when you look at the last 200 draws — stability matters more than the peak value.
Collect 4-5 numbers from these pairs. If you end up with 4, add one "single" from the top of the ball weight list with a z-score ≥ 1.5.
Check the "prettiness" of the resulting combination. If it is 1–2–3, or all even, or three numbers in a row — redistribute. Not for the chance of winning, but for the size of the prize (less chance of splitting it with a crowd).
Verify in combination-lookup — whether exactly this combination ever came up. If it did, that is not bad (the odds are the same), but it may be interesting for context.
Save it to the notebook and play 10-20 draws. Compare the outcome with your expectations — that is the only way to understand whether your method works or just plays along with your intuition.
Lottery | Total pairs | Possible triples | Exact repeats per 1000 draws (expected) |
|---|---|---|---|
1,176 | 18,424 | ≈ 0.26 (rare) | |
1,225 | 19,600 | ≈ 0.01 (very rare) | |
2,346 | 52,394 | < 0.01 (almost none) |
A frequent pair is not a significant one: p-value and false positives
Let's return to the main catch. Suppose on the frequent pairs view you saw the pair 14 and 23, met 85 times against an expectation of 68. Its z-score is about 2.1. The temptation is to bet on this pair as a "strong" link. But you have to ask yourself: how likely is it to see such a deviation purely by chance?
At a z-score of 2.1 the probability of a random exceedance is about 1.7%. Across 1,176 pairs this means that in any fair lottery we will see about 21 pairs with a z-score ≥ 2.1 — simply because there are many of them. So the mere fact that "a pair is at the top" is not enough to consider it significant.
A practical rule: use a consensus of at least two methods. A pair should be both at the top by frequency and keep its position when the archive is split in half. Or it should match a triple in which both numbers participate. This is not a guarantee, but it seriously reduces false conclusions. For more on the consensus approach, see the article on analysis methods.
Practical takeaways
Winning in the archive and winning tomorrow are different notions. The first is a fact, the second is a probability independent of history.
Pairs with a z-score ≥ 2 are interesting, but not automatically significant: among 1,176 possible pairs about 59 will be "significant" by chance.
Triples on their own are a weak signal. They are useful in combination with pairs: a triple that includes a significant pair is a more reliable guide.
Exact combination repeats in France Loto do happen (≈ 17 over the whole history). This is not a pattern but a statistically expected phenomenon.
"Pretty" combinations (consecutive, all even, dates) have the same probability as any others — but crowds choose them, so the prize is split into smaller shares.
Expanded bets (systems) multiply the odds proportionally to the number of combinations. On their own they are not a "hack," but in combination with pair analysis they give a better bet structure.
Do not write "frequent means significant." With a large number of possible combinations, random spikes are inevitable. Either use the Bonferroni correction, or a consensus of two or three independent methods.
Bingo-style games are not for this analysis: the numbers on the ticket are already printed, there is no choice.
