Hard to explain and if I did then everyone would know what I am doing, so no advantage to me.
Essentially it comes down to popularity and behaviour, but not of the actual numbers, a derive a dimensionless variable that I then apply various dynamic and statistical measures against. I map this out like a topographical map and then trace the dimensionless variables movement on these maps.
People need to move away from statistical analysis and apply dynamic measures. Any statistical measurement gives you a static snapshot of the data, if applied from game to game it does not provide a dynamic shift in the statistics that match the dynamics of the shifts in the numbers.
If you constantly pick the hot running numbers you will build a list of the hot numbers that does not change much from game to game, likewise the cold or past due numbers. Meanwhile the game is merrily throwing numbers all over the place in defiance of your hot and cold lists and any other statistical measures you apply.
This is why statisticians say that it is completely random, it's because they are static in their thinking and application of analysis. BTW it is completely random from a superficial analysis, but if you look at thing like the law of averages then you will see a loose correlation in that if you measure the variance over time the highest and lowest are never further apart than a certain amount. So most lotto games carry a loose interaction to this as well as some other very vague connections.
Over time I have teased out each one of these and then carried out very detailed analysis of them, I got sweet stuff all in return. Why? because I was using statistical analysis and not looking at the dynamics of what was happening.
It's all about the dynamics, and to run the calcs on this takes a lot of computer HP and time as well as a lot of lateral thinking and tenacity.
It's all very large.....