I decide to play a game of heads or tails with an absolute "fair" coin $1,000,000$ times. Suppose (however unlikely) I get $999,999$ heads in a row!

If I get heads on the 1 millionth toss, I win a prize! Should the past coin tosses influence my present decision?

Logically, the last toss is still $50$% to get heads... but can I allow my mind to bias the probability and favor heads over tails?

Thanks for any insights!

  • 3
    $\begingroup$ en.m.wikipedia.org/wiki/Gambler%27s_fallacy $\endgroup$ Sep 4 '20 at 19:09
  • 3
    $\begingroup$ It’s tricky. Practically speaking the issue is that if you ever found yourself in this situation then it’s extremely likely that the coin isn’t actually fair. To do a complete analysis involves an analysis of why you think you know the coin is fair which is underspecified by the problem. $\endgroup$ Sep 4 '20 at 19:10
  • $\begingroup$ In addition to my answer - indeed as @QiaochuYuan has mentioned, if this happened in real life we would obviously start to question the fairness of the coin. But if we (theoretically) knew the coin was absolutely fair - as my answer states - the probability remains exactly the same on the last go (${50\%}$) $\endgroup$ Sep 4 '20 at 19:19
  • 2
    $\begingroup$ Say you believe that there is a one in a million chance that someone slipped you an unfair coin. Then that possibility will loom very large given the preposterously improbable event you observe. $\endgroup$
    – lulu
    Sep 4 '20 at 19:19
  • $\begingroup$ Call heads. If the coin is fair, it doesn't matter what you do, and if the coin is biassed, it's overwhelmingly likely that the next toss will be heads. $\endgroup$
    – saulspatz
    Sep 4 '20 at 19:29

The answer since you stated the coin was known to be fair:

If you get a huge number of heads in a row - the chance you get another heads on the next toss is still ${50\%}$. The thing is - those ${999,999}$ heads in a row have already happened, and the events are independent. [Basic conditional probability]: for independent events ${A,B}$ we know that


In this case - ${B}$ is "I get a heads next toss" and $A$ is "I just got ${999,999}$ heads in a row".

If you ask "what's the chance I get ${1,000,000}$ heads in a row?" without having done any tosses already, then the answer is a small number. Nothing has happened yet. It may seem odd since your intuition tells you "come on, the last one must be a tails!" but it's still the exact same probability, since those tosses have now already occurred.

For a real world coin - if you throw it a lot of times in a row - it technically I guess could become slightly "weighted" overtime and have a slight bias as a result - but in a theoretically perfect coin - the probability remains the same. Also if you did get that many heads in a row with a real world coin - we would obviously start to question whether that coin really was fair.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.