I took a look at temperatures:
Northeast USA weather
Chicago-Toronto-New York City.
Perhaps the thing would be to go with temperatures.
The very last digit of the temperature for 3 cities for example either at a given time of the day everyday or the daily highs and or the daily lows (Last Digits only), just something to think about.
This could be for testing a system or systems to see how well or bad it or they work with truly random numbers, numbers not controlled by any interested parties as the lotteries are.
This way there would be no such things as 3 to 5 predraws nor any other kinds if any of undue manipulations of the draws, with the lotteries you can never be completely sure that the draws are really random, with real not controlled by anybody random numbers you can find out for sure if you really have a winning system or not, no maybies and no buts either.
You would just need a data history of whatever lenght your system has to have.
Random numbers are random numbers that is they should all be the same.
Unless the numbers are not really random.
Before using any source of digits take a good look at a good size history of them to see if they really appear to be random or not, use your judgement for that, you should not get perfect expected stats, but the stats should not be completely off either, they should move in time between the extremes so in the long term they should be closer to the expected stats, but perhaps never completely quite right.
Besides that the main thing is that they be not controlled by any interested parties just in case to avoid any possible chances of foul play.
As to any system working or not, only the long term can tell, the pick 3 is really a straight system and there are 1000 numbers, so the true test might be 1000 draws.
But no matter the source of the digits, the randomness of any source of numbers will always be in question.
If you win with a source, but not with the lottery, they will say that that source is not random and that it is biased in your favor.