Mikeybit, I agree to an extent. The best we can do is try to make the best decision taking whats given. Right now we are operating almost as if %-change in price is random, and do not try to make predictions. We can however improve out decisions by looking at volume. As we increase our hash power, making predictions will be less important because we will be able to mine faster.
Geocurt, you're right. You can definitely use data to select the N most profitable coins. It is just a matter of picking the best one!
You have plenty of information available so building a model attempting to predict %change is possible. However, comparing alt-coins to real world financial markets is difficult since there is absolutely 0 regulation and the people that trade cryptos aren't necessarily the most rational.
To address your 2nd point, it is still important to look at depth of the markets. If we're mining faster, we're collecting more coins, therefore we need market depth to be able to offload those coins. If we're mining faster, we still don't want to switch coins all the time as a lot of hashpower is lost during switches.
I sent you a PM, if you'd like to get some feedback, let me know.