Roulette is a mechanical device. Whatever is mechanical is not capable to give you real randomness of events. It will always have some kind of order. Well, not always, but most times. Or just many times.
Remember, that is what British and Polish intelligence used when cracked German Enigma. When operators used it, they did not use it really randomly, they had some patterns repeated over and over again. Because it is very hard for either human or mechanical device to be always random. And there you had both at work, just like at the roulette wheel in a casino. So they tracked those patterns thoroughly. And they were never able to crack the machine completely, they just cracked lots of info, but not all. Just like in roulette, even if u can win, it is impossible to do all the time, each time for thousands of spins. If you try u will always lose.
Computers are different story though. Computer can give u anything u want if programmed properly for that goal. So here goes an answer to your question why RNG and a physical wheel are different.
Iggiv
i like the above, i like the part ( they had some patterns repeated over and over again.)(So they tracked those patterns thoroughly. And they were never able to crack the machine completely, they just cracked lots of info, but not all.) This is like the FOBT its patterns, patternes of what? Answer 0x,1x and >1x there you have it.
Now i collect games on FOBT around 879, but have neglected up dating, but this has inspired me to up date the FOBT games, the results in the FOBT data, is close to the data in jackpot247. The info from them is knowledge,it gives avg to hit,max to hit and much more, Gut,LOTT,X uniques in X spins, will all help in your decissions.
That small collected data is so valuable.
Imagine this then. Roulette has been played for what over 200 years, imagine if someone had started collecting games at the start,we'd have 200 years of game data showing what,average forwhat a non-hit takes to hit, max a non-hit could take, after 200 years of game data,one using a rolling 100 games, another for 1,000 games and why not a million, each time a new games data is added,would it change that dramatically. If all 3 sets of data said the 10th non-hit takes a max of 6 spins, its avg 2 spins,after 200 years if you was playing and see it missed for 3 spins would you not now bet