Hi Icewynd, the BIG problem everyone makes is to think that since we can't find a way to do something with our senses, our thinking, our approaches, then it can't be done. The truth is really far from that as soon as we understand that not everything around us is humanly understandable or based on what a human interpret/believes/assumes of it.
So, you say "but surely it impacts all balls equally". You did fall again at the same problem. The answer is "WE DON'T KNOW" because we can't understand it with our own senses. True, there is a strong belief that this should equally impact all balls (I use your example). And I pretty sure that if we do that experiment millions of times, we'll possibly (still we don't know but we assume) that the overall outcome is pretty random. I don't disagree to that. Actually I expect over the very very long run of lotto draws these to behave statistically as pure random events i.e. eventually all the combinations will be drawn approximatelly an equal amount of times. Of course we'll need way more test events than the total possible combinations of our lotto game to verify the pure randomness of this particular experiment but I hope we agree on that aspect, that we expect overall the results to show pure random behavior.
Now, what I say here that has this reduced randomness behavior is that afew concecutive draws do have something in common. If you want to tell you why is this happening, I don't know. I can't tell you reasons for this but it is there most of the time, it can be due to physical factors, it can be due to moon's gravity it can be anything. Really uncountable small and big "forces" affect the result. Based on the results I get, it turns out that a few of those factors play the major part at the produced draw result however, sometimes this relationship is more obvious, sometimes it fades. I call this "locality" behavior reduced randomness. The engine I have designed doesn't look for something specific, it expects from the data to provide what they contain on their own and this relationship changes over time, quite often in major steps. So, the outcome is that locally draws do have connections but over the longer run they do conform to what we would call random. By the way, the effects of those few major factors can't be trapped by traditional methods such as statistics (due to their larger scale application you'll really analyze the random behavior, not the locality), or AI (lotto draws do not have AI), or neural netowrks (because this assumes a given sequence produces always the same result which I believe this will never ever happen). All these approaches are too elementary to even scratch the surface of supposedly random events and of course, we can't expect anything useful to come out from doing so.
If you want to read more about this, check my website, I have a detailed PDF that describes more things.