I might be wrong, but this is what I think. The last 3 draws entered are perhaps sort of like a fingerprint of the "randomness" of the game at that particular time and maybe LottoSync produces a random stream output of numbers and compares the sequence of the numbers of the last 3 draws against it's random output and tries to find the same sequence or as much the same as possible in it's output as it was on the 3 draws entered and when it finds it it follows that sequence and records that numbers that follow it and it does these many times and compares all the numbers that follow the particular sequence against themselves and or records the times that the particular digits come out and maybe the digits that follow after the sequence of the last 3 draws the most often or the most times are the most likely numbers to come out next, provided that the output of LottoSync is "random" in exactly the same way as the randomness that produced those 3 draws that were entered.
LottoSync tries to reproduce many many times the "randomness" or section(s) of randomness that produced those last 3 draws and then "looks" at what numbers are produced next after that particular section of randomness, does this many times and maybe the numbers that follow it that come out the most often or times are the ones that have the most chance of coming out the next time or on the next draw.
But since they say that you can't predict "randomness or random" it can only come close.
It's impossible to exactly emulate the conditions that produced those 3 draws or numbers and to , make things worse, every new draw the conditions are not exactly the same, so the conditions that produced the first set of draw numbers was not exactly the sames as the conditions that produced the second set of numbers and also the 3rd set of numbers and so the conditions won't be the same for the next set of numbers to come out next either.
So if all of these is true, then that is why is so hard for LottoSync to come out with winning numbers, it's trying to do the impossible or what they say that it's impossible, it's trying to reproduce a past random dvent many times and to predict a "random" dvent that is yet to happen, based on reproductions of the past random dvent and the sequences of numbers that follow it (the reproductions).
But of course, I might be wrong.