Lotto Group Sheet

Your sheet splits the numbers into 40 groups of 5. Every group follows one rule: n, n+4, n+15, n+23, n+36. So group 1 is 1, 5, 16, 24, 37. Pick a game below, then choose a tab.

Top 5 predictions

In shortThese are the app's 5 best guesses for the next draw. They use the numbers that come up most, numbers that often come up together, and what a normal draw looks like. Our tests show these guesses don't reliably beat random picks, so treat them as an experiment.

    How did the last predictions do?

    The 5 picks the app made before the latest draw, checked against what came out.

    How the score works

    • Popular numbers (50%): numbers that have come up more often than average score higher.
    • Pairs (30%): numbers that have come out together before score higher.
    • PowerBall, bonus and booster: #1 gets the most-drawn PowerBall (or bonus / booster ball), #2 the next, and so on. In Lotto and UK49s you don't pick a bonus: it counts when the bonus ball matches one of your numbers, so the orange-ringed ball is just a guess at which number it will be. Check: the most-drawn bonus balls from January to June caught about as many later bonus balls as random numbers did.
    • Normal shape (20%): sets that look like a typical draw (total, odd/even and low/high mix) score higher.
    • No set has ever been drawn before, and no two sets share more than 2 numbers.

    Top 5 for every game

    Each card shows the #1 pick. Tap a card to see all 5.

    The draw on your sheet

    In shortEach draw is shown on your paper sheet. Blue circles are the numbers that came out, and a blue group heading means that group caught 2 or more. Use the arrows to move between draws.

    drawn number bonus / booster (not counted) group with 2+ numbers drawn predicted for this draw 77outside this game's numbers

    Every draw this year

    Each column is one draw (oldest on the left) and each row is one group. Darker squares mean more numbers from that group. Click any column to see that draw.

    0123+ numbers from the group

    Group scoreboard

    In shortThis tab checks whether some groups on your sheet come up more than others, and whether following a "hot" or "due" group helps. It doesn't. The one useful tip is to use a group whose numbers can all be drawn in that game.

    Drawn = how many of the group's numbers came out this year. Expected = what an average group would get. Above/below shows the gap: between −2 and +2 is normal, and a group outside that is unusual. Click a heading to sort.

    Which group should you use?

    Test: does the past predict the next draw?

    We replay the year. Before each draw, each rule picks a group using only earlier draws, and we count how many of its numbers came out. Above 0 means the rule did better than average. "Random pick" is the comparison: if a rule can't beat it, the rule doesn't work.

    What the numbers show

    In shortThis tab lists what this year's draws show: the most and least drawn numbers, common pairs, and what a normal draw looks like. The shape of a draw stays the same over time, but the "most drawn" numbers keep changing.

    What a normal draw looks like

    Most draws fit this pattern. Blue bars show the most common results.

    A pick made from these patterns

    5 full tickets

    Built from popular numbers and pairs, shaped like a normal draw, and never drawn before.

    Test: do the patterns last?

    Moon, stars or calendar?

    In shortWe test whether the moon or star signs change the numbers drawn. Made-up groupings "pass" the test just as often as the moon and stars do, so there's no real link.

    How the test works: we sort the draws into groups (for example by moon phase) and compare them. Then we shuffle the labels randomly 1,000 times. If the real groups differ more than almost all the shuffles, the difference is real.

    Result

    All groupings compared

    Beaten by = how many of 1,000 random shuffles did at least as well. Under 50 counts as a notable difference. The decoys can't mean anything, so watch how often they come out notable too.

    GroupingGroupsBeaten byVerdict

    Decoy test: can nonsense pass?

    Press the button. It gives the draws 100 different sets of random 12-group labels and counts how many look "notable". Compare that count with what the moon and stars get.

    Formula lab

    In shortBuild your own prediction formula with the sliders, then press "Run the replay" to see if it would have worked on this year's draws.

    The formula gives every number a score, turns scores into chances, picks a set, and never repeats a past result. The steps below show the maths, with a plain-words version of each.

    The formula, step by step

    1. 1. How often it\'s drawn. fi = (ci + a) / (N + a·M), where ci is how often number i was drawn, N is all numbers drawn, M is the highest number and a is the prior. A large a pulls every number back towards fair when there's little data. In plain words: numbers drawn more often score higher, but we don\'t trust small counts too much.
    2. 2. How recently it came up. ri = e−gi/τ, where gi is the number of draws since i last appeared. A positive weight favours numbers seen recently, and a negative weight favours overdue ones. In plain words: numbers seen lately score higher, or lower if you prefer "overdue" numbers.
    3. 3. Your sheet. hi is the average recent hits of the groups on your sheet that contain i. In plain words: numbers in groups that have done well lately score higher.
    4. 4. Total score. Si = wf·z(fi) + wr·z(ri) + wh·z(hi), where z(·) rescales each part to mean 0 and standard deviation 1 so the weights compare fairly. In plain words: add the three parts together, using your slider weights.
    5. 5. Turn scores into chances. Pi = eSi/T / Σj eSj/T (a softmax). A low temperature T makes the formula stick to its top scores, and a high T makes it close to random. In plain words: higher scores get a bigger chance of being picked.
    6. 6. No repeats. The formula draws numbers from Pi without replacement. It rejects the set if every number in it appeared together in any past draw, or if the set shares more than the allowed amount with the last draw. In plain words: a set that has already been drawn is never used.

    Your formula's pick

    Highest-scoring numbers

    Test: would your formula have worked?

    Press "Run the replay". It picks with these exact settings before every draw since the 30th, using only earlier results, and compares the matches with random picks.

    A school project. Each draw is random, and no pattern changes the odds of the next one. If you play, only use money you can afford to lose. Free help: National Responsible Gambling Programme, 0800 006 008.