The thing is that the Martingale strategy with the 50% bets is basically identical to repeatedly playing the 85-97% winning odds bets with the amount of Bitcoin that you would risk at maximum with your Martingale strategy.

That's interesting.

I looked into this, and found a surprising result:

If I bet on "lessthan 8000", and keep increasing my bet each time I lose such that my overall profit when I win is constant, and commit to making 19 bets if necessary, I am effectively placing a large bet on "lessthan 60012.5":

>>> (1-(1-8000/65536.0)**19)*65536

60012.49314412828

That's a little easier to win than "lessthan 60000", which pays out 1.071x. Since I'm placing multiple bets, and getting a slightly easier-to-win bet than "lessthan 60000", you would expect my payout in the event of a win to be a little less than 1.071x, but it's actually a little higher:

win on turn 1 : bet = 18.34444122, spent = 18.34494122, win = 146.76222078, lose = 0.09122220, back = 0.00000000, profit = 128.41727956 (87.793% to lose all)

win on turn 2 : bet = 20.95186542, spent = 39.29730664, win = 167.62263649, lose = 0.10425932, back = 0.09122220, profit = 128.41655205 (77.076% to lose all)

win on turn 3 : bet = 23.92990113, spent = 63.22770777, win = 191.44808952, lose = 0.11914950, back = 0.19548152, profit = 128.41586327 (67.667% to lose all)

win on turn 4 : bet = 27.33122595, spent = 90.55943371, win = 218.66002143, lose = 0.13615612, back = 0.31463102, profit = 128.41521874 (59.407% to lose all)

win on turn 5 : bet = 31.21600495, spent = 121.77593866, win = 249.73977624, lose = 0.15558002, back = 0.45078714, profit = 128.41462472 (52.155% to lose all)

win on turn 6 : bet = 35.65295485, spent = 157.42939351, win = 285.23711477, lose = 0.17776477, back = 0.60636716, profit = 128.41408842 (45.789% to lose all)

win on turn 7 : bet = 40.72055959, spent = 198.15045311, win = 325.77993920, lose = 0.20310279, back = 0.78413193, profit = 128.41361802 (40.199% to lose all)

win on turn 8 : bet = 46.50845857, spent = 244.65941167, win = 372.08539985, lose = 0.23204229, back = 0.98723472, profit = 128.41322290 (35.292% to lose all)

win on turn 9 : bet = 53.11903225, spent = 297.77894393, win = 424.97258068, lose = 0.26509516, back = 1.21927701, profit = 128.41291376 (30.984% to lose all)

win on turn 10 : bet = 60.66921318, spent = 358.44865710, win = 485.37698775, lose = 0.30284606, back = 1.48437217, profit = 128.41270282 (27.202% to lose all)

win on turn 11 : bet = 69.29255432, spent = 427.74171142, win = 554.36709722, lose = 0.34596277, back = 1.78721823, profit = 128.41260403 (23.881% to lose all)

win on turn 12 : bet = 79.14159147, spent = 506.88380289, win = 633.16325523, lose = 0.39520795, back = 2.13318100, profit = 128.41263334 (20.966% to lose all)

win on turn 13 : bet = 90.39054140, spent = 597.27484429, win = 723.15926428, lose = 0.45145270, back = 2.52838895, profit = 128.41280894 (18.407% to lose all)

win on turn 14 : bet = 103.23838355, spent = 700.51372784, win = 825.94703784, lose = 0.51569191, back = 2.97984165, profit = 128.41315165 (16.160% to lose all)

win on turn 15 : bet = 117.91237969, spent = 818.42660753, win = 943.34475917, lose = 0.58906189, back = 3.49553356, profit = 128.41368520 (14.187% to lose all)

win on turn 16 : bet = 134.67209391, spent = 953.09920144, win = 1077.42904272, lose = 0.67286046, back = 4.08459545, profit = 128.41443673 (12.455% to lose all)

win on turn 17 : bet = 153.81398395, spent = 1106.91368539, win = 1230.57166670, lose = 0.76856991, back = 4.75745591, profit = 128.41543722 (10.935% to lose all)

win on turn 18 : bet = 175.67664520, spent = 1282.59083059, win = 1405.48152682, lose = 0.87788322, back = 5.52602582, profit = 128.41672205 ( 9.600% to lose all)

win on turn 19 : bet = 200.64679995, spent = 1483.23813054, win = 1605.25255315, lose = 1.00273399, back = 6.40390904, profit = 128.41833165 ( 8.428% to lose all)

I'm risking 1483.23813054 BTC in total, to win 128.41 BTC. That's a payout of 1.0866x:

>>> 128.41833165 / 1483.23813054 + 1

1.0865797129981056

How can it be that committing to 19 "lessthan 8000" bets gets me a bigger chance of winning than placing a single "lessthan 60000" bet, but at the same time gets me a higher payout multiplier, as well as a bigger effective "max bet" (it's not possible to bet 1483 BTC on "lessthan 60000" directly, but using Martingale betting I can effectively do just that).

tldr: Using Martingale betting it seems to be possible to effectively bet far more than the maximum bet, increase your chances of winning, and increase the payout multipler too - Win/Win/Win. Is there a flaw in my reasoning here?