Luck in the short run, meaning your bad plays won't be necessarily punished immediately and your good plays won't be rewarded.

Definitely skill in the long run, meaning your bad plays will get punished and good plays rewarded.

Imagine a bag, where there is one black and three white balls. They're identical and you can't see which one you are taking out. What's the probability of taking out a white ball? Spoiler alert: 3:1 or 75%.

If you do it only once and maybe take out the black ball, what's the actual outcome? 100% Black Ball, 0% White Ball.

If you do the same thing 10 times, let's say you take the Black Ball out 3 times and the White Ball 7 times. Outcome: Black Ball 30%, White Ball 70%. Although unlikely, if you do it only 10 times, it might happen, that you take the Black Ball out 1 time or 4 times or maybe even 6 times.

Now, let's say you do the same thing 10,000 or 100,000 or 1,000,000 times. I can guarantee you, that the outcome will be ALMOST 75% White Ball and 25% Black Ball.

Deduction? You need a large sample size for the probabilities to work properly.

If you have 500$, you don't want to bet the whole amount, that the next ball taken out will be white, because 25% is still quite a big probability in the short run. You might get unlucky and lose all of your money. But, if you have 500$ you can quite comfortably keep betting 1-5$ and you will show profit in the long run.

Deduction: You need a good

bankroll management to "survive" until the probabilities start actually affecting your results.

EDIT: Also, I think the sufficient sample size depends on the probability. Following the previous example, I think you're able to get your actual outcome matching to the expected outcome quite quickly, given that it's 75-25. But 60-40 or 55-45 will and should take longer time to get there.