I'm doing some calculations for online casino slot promotions and am curious if anyone has advice. In a given game of slots, what factors would determine how confident you can be that a certain sample size is sufficient to do a realistic ROI calculation?
For example, say you'd like to be sure your data is within 1% of the real house ROI. The main factor I can think of is jackpot size.
-If I have 10,000 spins of data, betting 1 unit per spin, and the jackpot is 150 units, then a single spin could change the house edge calculation by 1.5%. Therefore 10,000 spins is not enough.
This makes me think it should be done by a rule-of-thumb "safety factor" of jackpots that fit into your variance amount.
In my example, say if the safety factor is 5 jackpots and 1% is the variance:
5 jackpots* 150 units per jackpot /.01 = 75000 spins.
Are there any other things I should be considering? If my example makes sense, what "safety factor" would you feel comfortable with? Let's assume there are no posted odds available for the slots and data gathering is all we can do.
Thanks for looking. If this needs to be moved to the probability forum, go for it
Estimating each payout would take WAY too much effort for anyone to even consider doing it. Maybe on a super simple old machine, but not on any modern multi-line.
There must be an amount of spins you'd be confident at. As an extreme example, a billion spins on a low jackpot machine would be plenty. But even a billion powerball tickets might not be enough
Just throw out the jackpots and average the rest, and adjust the final figure by the published jackpot expectation. Sample size needed for a given confidence interval around the result is easily calculated.
But is there not a published hold percent or payout percent for the game?