3. Now, if all you do is calculate what he WOULD have won had the dice result been shifted by 2, then I calculate +420,784 BTC!
4. If you shift it by 1, I calculate -427,822 BTC!
If what I said in point 3 and 4 are correct, then you wouldn't find this fishy? Almost like he was purposely leaving a trial?
I can very confidently say that both 3 and 4 are wrong. Somewhere you messed up. Note how 3 and 4 are VERY similar sizes but opposite sign? Note also that he was nealy always betting high. So most of the time shifting a win would stay a win. The only way you'd get that result for real was if he exactly alternated big/small bets and the RNG generated in high/low alternating sequence as well. And if he did that, shifting 2 would be roughly same results as he actually got.
Your program is broken.
Following on from my last post. If he waits for 2 small bets where the first wins and the 2nd loses, then bets big, then shifting by 1 will result in the big bet losing, while shifting by 2 will result in it winning.
This would give exactly the results Peter is seeing.
Yeah - very good point. If he's betting based on certain past actions then shifting WOULD give ridiculous looking results. And that would in no way be evidence of chearing - in fact the opposite if anything.
If you can break down his behaviour into rules then in theory at least you could check to see whether it was just luck or he'd somehow spotted a real pattern (meaning the seeding system is broken). That would mean running loads of tests where you ran the betting rules against a long series of bets from multiple sets of seeds.