Here's what I think about it after making probably over a million bets on various dice sites. Indeed the left picture can hardly be the result of a random distribution because the dots are evenly distributed across the surface, and this is not what happens when the process is truly random. If that was the case the martingale would work perfectly, but the thing is that in a really random process you can get 20 wins in a row, winning 2,000 sats with that, and you can get 20 losses in row, losing over 1 BTC with the same settings.

But, although there is always a deviation from the even distribution, it's very important to realize that there is no way of predicting where the deviation is going to occur.