Why Past Results Don’t Predict the Next Roll
On 18 August 1913, a roulette wheel in Monte Carlo landed on black 26 times in succession. As a complete sequence, that run carried a probability of about one in 68.4 million. Every single spin inside it, though, held steady at 18 chances in 37, roughly 48.65%, and not one of them had any idea what the spin before it had done.
The wheel wasn’t building towards anything.
That night gave the gambler’s fallacy its second name, the Monte Carlo fallacy, and the belief behind it has stayed with us for over a century because it feels so reasonable while you’re watching the run unfold, as IFLScience noted in revisiting the episode. If the dice can’t remember a rough run, they can’t hold one against you either. Nothing is owed in either direction.
The same logic follows you onto your phone. Whether the wheel sits in a Monte Carlo salon or you’re loading a spinbet casino lobby between other things, each result is generated on its own, with no reference to the one you just watched.
What follows is what independence means for your next roll, why fair results arrive in clumps rather than tidy alternations, what Australian spending figures suggest about choosing your number before you start, and the one habit that published research supports.
Memory Is the One Thing Dice Don’t Have
Independence sounds technical until you see what it removes from your thinking. There’s no correction coming, no balancing mechanism, no tally being kept anywhere.
What we do instead is fascinating, and it’s been measured. Psychologists Peter Ayton and Ilan Fischer ran experiments showing that people expect a run of roulette outcomes to break, while at the same time expecting their own run of correct predictions to keep going. Their work documented negative recency for the wheel and positive recency for personal performance, held by the same people simultaneously. Two opposite beliefs about randomness, side by side.
This isn’t a laboratory curiosity either. Rachel Croson and James Sundali analysed 18 hours of security footage from a single roulette table in Reno, and their field study found small but statistically significant effects that lined up with what the lab had already shown.
Here’s what I think is going on underneath. We read a sequence of results the way we read a plot, expecting tension to resolve and loose ends to be tied off. Dice don’t have plots. Each roll is a separate sentence with no knowledge of the one before it, which is why a run of six sixes tells you precisely as much about the seventh as a run of nothing at all.
You’ll meet that principle everywhere you play. On a site like SpinBet, where the lobby carries over 5,000 games from more than 20 software providers, the titles differ wildly in theme and pace while working identically at the level that counts: each outcome is generated fresh.
So if nothing is owed, why does a long streak feel so heavily loaded?
Fair Doesn’t Mean Tidy
Because our expectations were never calibrated for what randomness looks like up close.
Research into how we judge randomness found that people rate binary sequences alternating more often than chance as the most random of all, and that the bias appears specifically once a cause is named, such as a coin or a shooter. That study tells us something useful about ourselves: a perfectly fair run of identical results reads as rigged, when clustering is exactly what fairness produces.
Presentation plays its part too. Work on the role of experience in the fallacy found the bias showed up strongly when outcomes were experienced one at a time, and weakened when the same history was displayed all at once. The drip feed builds the sense of a pattern. The numbers themselves don’t.
Setting the feeling against the arithmetic makes the gap clear:
| What it feels like | What the evidence says |
| A long run of one colour means a correction is due | The 1913 run was unlikely as a whole sequence, near one in 68.4 million, while each spin held at 48.65% |
| A number that hasn’t appeared lately is overdue | Danish lotto players demonstrably bet less on numbers drawn the previous week, with no advantage to show for it |
| A personal hot streak means you’ve found your rhythm | Positive recency for one’s own predictions is a documented bias, recorded by Ayton and Fischer |
| Alternating results look properly random | Rating high-alternation sequences as most random is the biased expectation, not the accurate one |
This is also why the low-stakes tables SpinBet steers newer players towards make good sense. You get to watch variance behave normally, in clusters and all, without any single sequence carrying much weight on your session.
The alternation finding is the part that reframed all of this for me, if I’m honest. Our mental picture of random is far neater than random has ever been.
The Number You Pick Before You Play
Once you accept there’s no correction arriving, the decision in front of you changes shape. It stops being about timing and becomes a straightforward question of budget.
Australians are well practised at this sort of arithmetic already. IBISWorld puts per capita gambling expenditure at $1,572.23 for 2025-26, up 1.3% on the year, which represents average net spend per person across a full financial year. For scale, Finder’s analysis of household figures shows the average Australian household spending $148,493 across 2025, about $2,856 a week. And the squeeze is real: Finder’s Cost of Living Pressure Gauge read 80% in July 2026 after 81% in June, sitting in its extreme band and not far off the April 2023 peak of 84%.
You’re already fixing amounts in advance for insurance, groceries and fuel. It’s the same discipline behind moving past budgeting apps to genuine financial literacy, where the value sits in the decision you make up front rather than the tracking you do afterwards. The same move works here, and there’s evidence behind it. In an experiment by Kim and colleagues, every participant shown a prompt set themselves a time limit, against just 4% of those given no prompt, and the prompted group went on to play for less time. Related limit-setting research found the people most likely to stay inside a limit were the ones who set it before starting rather than partway through.
Understanding the maths is less daunting than it sounds. As Stephen Woodcock, Associate Professor of Mathematical Sciences at the University of Technology Sydney, put it while explaining an improbable lottery result in The Conversation: ‘A basic understanding of probability and human psychology helps explain why this outcome isn’t as implausible as you might think’. His companion point lands nicely, too. The numbers 9, 18, 27, 36, 45 and 54 are exactly as likely to be drawn as 1, 18, 19, 28, 30 and 46, even though most of us instinctively rank the second set higher.
Deposits work best treated the same way, settled first and then left alone. SpinBet’s rewards section fits that rhythm, with bonuses activated deliberately before you fund an account rather than reached for once a session is underway.
If the next roll can’t be read from the last one, what is a chasing plan being based on?
Play the Session and Not the Sequence
Nothing, is the short answer. Which is liberating, because it means there’s no ledger to maintain.
Plenty of people are keeping one anyway. A 2024 study in the Journal of Risk and Uncertainty, using player-level administrative data from natural experiments in Haiti and Denmark, found the average lottery player avoids numbers that recently won, a small group chase them instead, and there’s no sign of players switching between those beliefs as streaks lengthen. A 2023 analysis in Judgment and Decision Making, covering hundreds of millions of entries across 118 consecutive semiweekly draws, tracked that avoidance fading over time as a number stopped feeling recent. Worth being straight with you here: that behavioural evidence comes from Denmark, Haiti, the US and the UK, since no Australian-specific study of this bias turned up. The money figures above are ours; the psychology is borrowed, and consistent across every population tested.
What the research supports comes down to five things:
- Decide your amount before you open the lobby, because limits set in advance are the ones that hold
- Treat a long run as ordinary rather than informative, since fair sequences cluster
- Ignore hot and cold number tables, as recent frequency carries no forward information
- Judge a session on whether it was fun for the price, not on whether the sequence turned
- Review outcome history in one go rather than watching it tick over, which weakens the pull of the pattern
There’s a small pleasure in giving up the tracking altogether. SpinBet’s ‘I’m Feeling Lucky’ feature, which picks a game for you at random, makes the point neatly; handing over the choice costs you nothing when no sequence carries an edge.
Nothing Is Owed, So Nothing Gets Lost Track Of
A memoryless game has one quality no other spending decision can offer you. Because the outcomes can’t respond to each other, the figure you settled on beforehand can never be undermined by what follows. Your grocery budget gets hit by price rises and your fuel budget by a longer commute. This one only moves if you move it.
That’s becoming more useful over time. Australian National University research cited by The Australia Institute records online participation at 33% of adults in 2025, up from 8% in 2017, with online play overtaking in-venue play for the first time in 2024. Closing time and the drive home used to do some of the deciding for us. On a phone, the number you chose is doing that work instead.
Advisory Notice: Play for fun, not for profit. Decide how much you’re willing to lose before you begin and hold yourself to it. If you catch yourself trying to win back losses or feeling unable to stop, take that as your cue to step back. Free, confidential help is always available through Gambling Help Online whenever the stress outweighs the enjoyment.
The post Gambler’s Fallacy: Why Past Rolls Predict Nothing appeared first on Daily Emerald.