I suggest that you start with a google search on "random number validation". I think you will find some statistical tools that you can run to see how random or not the data actually is. These statistical tests won't tell you what a predictive algorithm would be, but hopefully would give you some idea of whether or not such a thing is even theoretically possible. If the data is truly random, then looking backward over a relatively small interval you may (and probably will) see some pattern. But that piece of data would have no predictive value moving forward.

Of course a key piece of information would be "where do these numbers come from?". You haven't told us that. From your posts, I take it that this is not a roulette wheel.

You mentioned somewhere that maybe it takes 5 hours for the next number. There may be some correlation between time and the next number? I have a web application that reports tournament results. The number of hits per hour is definitely not random and has a lot to do with whether or not we ran a major tournament over the weekend. How many hits we will get in the next hour is highly correlated with how many hits we got last hour. Nothing happens for a long time - then something happens. The probability that something more will happen in the next hour is vastly increased.

There is a lot of Perl experience in the group, but we have virtually nothing to go on. Run some validation tests as suggested above, then tell us more about what you are actually doing.