I think if we choose a random 10 numbers and bet the other 27 with +1 -1 we would have OK results
That´s what I tried to say in KTF. In a large number of trials where almost all possible combinations are present, you will find that for every bet there is a maximum number of successive wins and successive losses. For any say 24 numbers bet, such as two dozens at a time, in 100000 spins you can expect to fail up to about 21 times in a row. It can be sooner or later. For a 27 numbers bet the expected losses in a row come down to around 10. So waiting for 10 unique numbers before betting is like a virtual loss afterwhich you can bet the opposite 27 numbers knowing the worst case scenario has passed and the next spin a sure win.