Consecutive lottery numbers are not as unusual as they can look. In a fair six-from-49 model, the probability that a result contains at least one consecutive pair is about 49.52%. That includes any adjacent pair anywhere in the six-number set, such as 8–9 or 32–33.
This does not mean a particular consecutive selection is more likely than another complete selection. A broad pattern can contain millions of possible results. One exact line is still one result.
The distinction between a pattern and an individual combination is the key to understanding this topic. We will count the pattern directly, explain why adding individual pair probabilities is not enough, and show how to reproduce the calculation.
Define consecutive carefully
For this article, consecutive means numerical neighbors: 1 and 2, 2 and 3, and so on. The labels 49 and 1 are not treated as neighbors in a six-from-49 model. This is a line of integers, not a circular arrangement.
A result such as 4, 5, 17, 28, 39, 46 contains one adjacent pair. A result such as 4, 5, 6, 28, 39, 46 contains two adjacent pairs, 4–5 and 5–6, within a run of three.
The event we want is at least one pair. A result with two or three pairs still counts once toward that event. That is why simply adding the chances of individual pairs can overcount.
We also ignore the order in which balls are drawn. Sort the selected labels before looking for differences of one. The question concerns the set, not whether neighboring labels came out of the machine one after another.
Count the opposite event
It is easier to count results with no consecutive numbers and subtract their probability from one. Suppose k distinct labels are selected from 1 through n, and write them in increasing order as x1 through xk.
If no two selected labels are consecutive, there must be at least one unselected label between each neighboring selected pair. That spacing requirement reduces the number of possible selections.
Define y1 = x1, y2 = x2−1, y3 = x3−2, and continue until yk = xk−(k−1). Removing the required gaps turns the spaced selection into an ordinary k-number selection from 1 through n−k+1.
The transformation can be reversed by adding those offsets back. Because it is one-to-one, the number of selections with no adjacent labels is C(n−k+1,k).
The resulting formula
There are C(n,k) total unordered selections. The probability of no consecutive pair is C(n−k+1,k) divided by C(n,k). Therefore the probability of at least one consecutive pair is 1 − C(n−k+1,k)/C(n,k).
For six labels from 49, the no-pair count is C(44,6) = 7,059,052. The total count is C(49,6) = 13,983,816. Subtracting their ratio from one gives approximately 0.49519845.
As a percentage, that is 49.519845%, usually rounded to 49.52%. The exact count of results containing at least one pair is 6,924,764.
This figure is a property of the stated model, not an observed percentage from a particular historical archive. A finite archive can differ from it through ordinary random variation.
See the transformation on a small example
Take three selected labels from 1 through 7. A no-neighbor set could be 1, 3, 6. Subtract offsets 0, 1 and 2 to obtain 1, 2, 4, an ordinary three-number selection from 1 through 5.
Another no-neighbor set, 2, 4, 7, becomes 2, 3, 5. Reversing the process restores the mandatory gaps. No valid spaced set is lost and no two spaced sets map to the same transformed selection.
There are C(7,3) = 35 total three-number sets and C(5,3) = 10 with no neighboring labels. Therefore 25 of the 35 contain at least one consecutive pair, a probability of 5/7.
The small example is worth listing by hand if the formula feels abstract. It shows why subtracting gaps reduces the available range without changing how many labels are selected.
Why adding pair probabilities gives a different quantity
In a six-from-49 model there are 48 possible adjacent pairs. A particular pair, such as 10–11, appears with probability 6×5 divided by 49×48. Summing that probability over all 48 pairs gives 30/49, about 0.612245.
That number is the expected count of adjacent pairs in a result. It is not the probability that a result contains at least one pair. A result with 10, 11 and 12 contributes two pairs to the expected count but only one occurrence to the at-least-one event.
The difference between 61.22% and 49.52% is therefore not an arithmetic contradiction. One number describes an average count; the other describes a yes-or-no event.
This distinction recurs throughout lottery statistics. Whenever patterns can overlap, count how many times each outcome contributes before interpreting a sum as a probability.
What about five-number main fields?
Using the same formula for five labels from 69 gives an at-least-one-neighbor probability of approximately 26.5037%. For five labels from 70, the result is approximately 26.1595%.
Those calculations concern only the five-number main field. They do not treat a separate special-ball value as another main label, and they do not say anything about a particular prize tier.
The probability is lower than in the six-from-49 model because the selection is smaller relative to the available range. There are fewer opportunities for selected labels to sit next to one another.
For a named game, first confirm its current number format. Our Powerball guide and Mega Millions guide distinguish main and special fields.
A common pattern does not improve a line
Suppose nearly half of all possible results in a model contain a consecutive pair. That category is common because it contains many different complete results. It does not give each result inside the category extra probability.
The set 1, 2, 3, 4, 5, 6 is one exact six-number selection. The set 7, 16, 24, 31, 40, 48 is also one exact selection. Under a fair six-from-49 draw, each has probability 1/C(49,6).
The all-consecutive line belongs to a much narrower subcategory than “contains at least one pair.” Using the broad category’s frequency to describe that particular line is a category error.
Filtering a generator to require or forbid consecutive labels changes which selections it can produce. It does not change the draw probability of any individual valid selection that remains.
How to analyze an archive responsibly
Sort the main numbers in each recorded drawing. Calculate the difference between each neighboring pair. If any difference equals one, mark that drawing as containing a consecutive pair.
Count each drawing once, even if it contains several adjacent pairs. Keep a separate column if you also want the number of pairs. Mixing those two measures will produce misleading percentages.
Exclude special-ball fields from this particular calculation. Also separate historical periods with different pool sizes or pick counts. A theoretical benchmark for one format is not automatically suitable for another.
Report the number of drawings, the date range and the format used. A percentage without a sample size hides how much random variation might be involved. Our sample-size guide explains why short archives can look unstable.
Spreadsheet and calculator checks
For a six-from-49 model, enter 1−COMBIN(44,6)/COMBIN(49,6) in a spreadsheet with a combinations function. Percentage formatting should display about 49.52%.
For a general model, use 1−COMBIN(n−k+1,k)/COMBIN(n,k), replacing n and k with the pool size and selection count. Check that the format permits the combinations you are counting.
An easy boundary check is k = 1. One selected label cannot form a pair, so the answer should be zero. Another is a very crowded selection where avoiding neighbors is impossible; then the no-neighbor count is zero.
These checks help detect swapped parameters or accidental use of permutations. Our jackpot calculator counts complete combinations; this pattern calculation is a separate mathematical analysis.
Questions and a short exercise
Does a consecutive pair mean a drawing is suspicious?
Not by itself. The calculation shows that such pairs can be common under a fair model. Investigating a mechanism requires more than pointing to an ordinary possible pattern.
Does avoiding consecutive numbers improve the odds?
It does not improve the probability of a particular valid line. It restricts your selection to one subset of the possible results.
Are three consecutive numbers the same event?
No. A run of three is narrower than at least one pair. The formula here includes runs of three but also many results containing just a pair.
For practice, examine 2, 3, 4, 18, 29, 40. It contains two adjacent pairs but only one run of three. It counts once in an at-least-one-pair tally. Now examine 2, 3, 18, 19, 29, 40: two pairs again, but two separate runs.
Being precise about the event turns a striking-looking result into a countable question. The useful conclusion is not that consecutive numbers are lucky; it is that broad patterns and individual selections must be evaluated differently.
Do not confuse a run with a pair count
A run of four labels, such as 10, 11, 12, 13, contains three adjacent pairs. It is one run, three pairs, and one drawing with at least one pair. Those are three valid statistics about the same selection.
If you build a historical chart, choose the statistic before counting. A chart of runs should define whether isolated pairs are runs of length two. A chart of pair counts should count each neighboring relationship. A chart of drawings with adjacency should use only a yes-or-no flag per draw.
For practice, compare 1, 2, 3, 20, 21, 40 with 1, 2, 10, 11, 20, 21. Both contain three adjacent pairs. The first has a run of three and a run of two; the second has three separate runs of two.
Neither selection contains more than one occurrence of the event “this drawing has at least one adjacent pair.” That event is simply true for both.
This distinction explains why two honest analyses can report different figures while examining the same archive. They may be counting different objects. Before calling one wrong, check whether its denominator is drawings, pairs or runs, and whether its definition matches the question you actually want answered.
Browse all ticket-checking and probability guides in the Lottery Learning Library.
Sources and further reading
Reviewed October 18, 2026. Examples are illustrative unless identified otherwise. Winning Numbers Today is an independent information website; the selling lottery validates tickets and pays prizes.
