Within the genius’s playground that was Bell Labs within the Nineteen Fifties, Claude Shannon, the daddy of knowledge principle, constructed an outguessing machine. The machine’s activity was easy: a human would select one among two levers, left or proper, and the machine would guess which lever the human would select. Shannon was impressed by a colleague, whose outguessing machine had a slim however particular success charge of 53 per cent. Shannon’s personal model was easier and even higher, guessing accurately 65 per cent of the time.
The legend associated in William Poundstone’s guide Find out how to Predict the Unpredictable (2014) (within the US, Rock Breaks Scissors) is that no person ever beat the machine over an prolonged run, with one exception: Shannon himself. Because the machine’s algorithm was easy and chic, he was in a position to mentally emulate it and play the other of no matter he knew the machine would predict.
How did the machines work? The logic was easy: each took notes of how their opponent behaved. For instance, in the event you’d gained a few occasions in a row, did you normally trip your luck by sticking with the identical alternative or did you have a tendency to modify? The presumption was that no human performs randomly. All of us have sure habits or instincts. No supercomputing was required: Shannon’s machine had simply 16 bits of reminiscence. We people should not arduous to learn.
An excellent easier exploit is the double-tailed coin. Any scoundrel with such a coin can resolve any disagreement by providing to toss it, then calling “tails”. However even when he provides his mark the chance to name the toss, he’ll in all probability win, as a result of individuals will name “heads” most of the time.
One of many first lecturers to determine this out was Louis Goodfellow, who within the late Thirties recruited volunteers to write down down a sequence that mirrored an imagined set of coin tosses. Seventy-eight per cent of his individuals started their sequence with “heads”.
In a bigger dataset — initially gathered in an effort to check telepathy (a narrative for one more day) — Goodfellow discovered that greater than 20 per cent of individuals both selected the sequence HHTHT or HHTTH. Different random-seeming sequences starting with heads had been additionally widespread, similar to HTTHH and HTHHT. Few individuals favoured something starting with tails, and the least widespread sequence was TTTTT, chosen by about one particular person in 600. Any mathematician can let you know, nevertheless, that there are 32 doable outcomes of tossing a coin 5 occasions, and every of the 32 is equally doubtless.
Fifteen years later, Alphonse Chapanis, a pioneer of ergonomics, requested volunteers to take an hour or so to fill a big grid with 1000’s of random digits. Analysing the outcomes produced some placing observations. The ten least widespread adjoining combos of digits, so as, had been: 66 99 00 11 33 44 88 22 77 55. And the ten hottest? 32 43 21 76 65 10 31 87 86 54 — one thing concerning the human mind likes descending couplets, it appears. In a really random sequence, 66 is simply as more likely to seem as 32, however what appears random as we scribble it down is something however. Such regularities may be exploited by an outguessing machine, a conman, even a password guesser.
They will also be exploited by lazy college students. Writer William Poundstone provides just a few tricks to brute-force your means by way of multiple-choice assessments. If in case you have no concept what the reply is on a true-false query, guess “true”. And if you already know what the reply to the earlier query was, guess that this time it would change. Supplied 4 choices, go for “B”. All the time select “the entire above” or “not one of the above”, if out there — they’re disproportionately more likely to be the right reply. None of those strategies ensures success, however they do enhance your odds as a result of most setters of multiple-choice questions are extra predictable than they realise.
If you really want a random sequence, then don’t ask a human. One different is to ask a machine to generate the randomness — however watch out. In 1980, the Pennsylvania “Decide 3” lottery was rigged by workers on the TV studio the place the draw was televised. Each quantity besides the 4 and 6 balls was weighted with a considered addition of latex paint, that means that when the fan was switched on to blow the light-weight balls round, all of the 4 and 6 balls floated to the highest. Alas for the conspirators, the heavy bets on 4-6 combos coupled with the portentous successful quantity 666 had been sufficient to lift suspicions.
Nearly the identical story performed out within the digital realm extra just lately, when an IT professional on the Multi-State Lottery Affiliation admitted introducing a backdoor to the software program that generated lottery numbers: whereas it was normally random, on sure dates it might produce random-seeming numbers that he might predict.
Pc cryptography depends on having a big quantity that can’t be guessed, and the easiest way to have an unguessable quantity is to make it random. Sadly, that is tougher to do than one may think, and several other pc techniques — most notoriously Sony’s PlayStation 3 — have been hacked as a result of the attackers found out how the “random” quantity was being produced.
If lottery machines — bodily or digital — may be rigged, and people are incapable of inventing actually random sequences of numbers, it’s tempting to resort to the great old school coin toss. Not so quick. Persi Diaconis, celebrated each as a mathematician and magician, has discovered {that a} coin toss is, usually, a 51/49 proposition in favour of ending the identical means up that it began. As an instance the thought additional, Diaconis constructed a coin-tossing machine. Insert a coin heads-up and the machine will flip it properly within the air — to land heads-up each time.
That is no mere theoretical downside. On the Worldwide Congress of Mathematicians this summer time, maths creator and YouTuber Matt Parker and pals ran a stall through which maths-curious members of the general public might strive tossing a coin. Parker promised to present $1,000 to anybody who managed to toss 10 heads in a row — a enjoyable problem to immediate individuals to discover what an unlikely consequence actually felt like.
Alas, it wasn’t as unlikely as he had imagined: having been pressured handy over the $1,000 prize twice on the primary day, Parker seemed on the accumulating information and realised that the complete distribution of coin flips was unexpectedly biased in the direction of strings of tails or strings of heads. The impact was stronger than Diaconis had predicted, maybe as a result of a lot of Parker’s coin tossers had been quite younger and unable to reliably flip cash excessive sufficient to rotate a lot. Parker’s answer: on the second day of the convention, he switched to cube. The sample of outcomes reverted to one thing a lot nearer to pure randomness . . . no matter that’s.
Written for and first printed within the Monetary Instances on 28 August 2026.
I’m operating the Oxford Half Marathon in October in help of an excellent trigger and in reminiscence of my mom. If you happen to felt in a position to contribute one thing, I’d be extraordinarily grateful.
