I think you're off by a factor of 2 there.  The easiest bet to win you have is 2%.  Players win that on average 49 of 50 plays.  On average they would win 49 times in a row before losing.  So they'd win 49 times 5 = 245 BTC, not 490.  And then you'd win 490 BTC when they eventually lose, twice what they've won so far on average.

Notice that although the player loses 1% of everything they risk, you profit by 98% of everything you risk (on the >2% game).  In a typical 50 plays, where the player bets 1 BTC each time, the player wins 49 times, winning 1/98 BTC each time for a total payout of 49/98 BTC = 0.5 BTC, and the lose once, losing 1 BTC.  So the player risked 50 BTC, made a net loss of 0.5 BTC.  That's a net loss of (loss/risk) = 0.5/50 = 0.01 = 1% of everything he risked.  But you (the site) only risk 1/98 BTC each time he bets 1 BTC, so your total risk was 50/98 BTC, and your net profit was 0.5 BTC.  That's a profit of (gain/risk) = 0.5/(50/98) = 49/50 = 0.98 = 98% of everything you risked!

The Kelly criterion (for bankroll management) suggests that to maximise the growth of your bankroll you should be risking a fixed percentage of your bankroll per bet regardless of the multiplier, if the house edge percentage is the same for each bet.

If you can afford to risk 1% of your entire bankroll when offering a 50/50 bet with a 1% house edge, then you can also afford to risk 1% of your bankroll when offering any other bet with a 1% house edge.  Conversely, if you don't have 500 BTC to your name then you shouldn't be offering 5 BTC bets at 50/50 because your risk of ruin is too high then.

In summary, the maths says that you should set the max bet such that the player profits by 1% of your available bankroll.  As your bankroll grows or shrinks, you adjust the maximum accordingly.