Session Review
This is the question every gambling forum answers with a made-up number. The honest version is computable, and it uses your own results rather than a rule of thumb: take the spread of your own session outcomes and work out how wide the uncertainty around your average still is. For almost everyone the answer is wider than they expect.
Most people stop counting at the point where counting would have started to work.
What your average could actually be
Standard error, and why even this understates it
Given your logged session results, the calculation is the ordinary one taught in any statistics course. Take the standard deviation of your session outcomes, divide by the square root of the number of sessions, and multiply by 1.96. That is the half-width of a 95% confidence interval around your mean.
The important property is the square root. Halving the width of the interval takes four times as many sessions, not twice. Going from a useless interval to a useful one is not a matter of a few more nights.
The standard error formula assumes results are roughly normally distributed. Slot outcomes are not. They are dominated by rare large payouts, which makes the distribution heavily skewed with a long right tail. On a sample that has not yet contained one of those rare events, the measured standard deviation is too small, so the interval computed from it is too narrow.
Which means the honest reading of your interval is: the truth is at least this uncertain, and probably more. That is the opposite of how confidence intervals are usually presented, and it is the correct direction for this particular distribution.
The forward version of the same problem: why a risk of ruin figure cannot be given without a variance number that no provider publishes.
Frequently asked
It depends on how much your own results vary, which is why this page computes it from your log rather than quoting a number. In practice the spread of session outcomes is large enough that tens of sessions still leave an interval wide enough to contain both "losing badly" and "roughly as expected". The square root relationship is the reason: four times the sessions to halve the uncertainty.
Ten losing sessions is unremarkable on a negative-expectation game where most sessions lose. It is the expected shape of the distribution, not evidence of anything. If you want to check something concrete rather than infer from a streak, read the RTP inside the game's information screen and compare it with the figure the provider publishes, which is a fact you can verify in under a minute.
Because slot session outcomes vary enormously relative to their average. The interval narrows with the square root of the number of sessions, so it shrinks slowly. A wide interval is not a flaw in the calculation, it is the honest description of what a small sample of a high-variance process supports.
Why the spread is as wide as it is
The interval on this page is wide for reasons that are documented elsewhere rather than assumed here.
From 95.45% to 97.02% across 47 titles. That range alone moves the expected result before variance has done anything at all.
A confidence interval is a spread around a centre. This is where the centre comes from, along with an explicit account of the figure it refuses to estimate.
A sample pooled across two RTP settings of the same title is not one sample. Checking which configuration was live is the step before the statistics.
Open a U88 account
Disclosure: this site is run by a U88 affiliate and earns a commission on accounts opened through the link above. A page that helps you see your own losses clearly also carries a link to open an account, and both of those are true at once. The arithmetic on this site runs entirely on your device and does not change because of who published it. U88's minimum deposit is RM30. 18+.