The kind of strategy that you want to do is called a
fund upgrading strategy. The people that do that all have their own formulas, but generally, it revolves around comparing relative strength of the short term (3-month and 6 month) against the medium term (1-year, 3-year). They then also have rules to keep the results sensible -- like picking the best of of a certain sector that's doing well, and not all of them. For instance, when gold does well, you wouldn't want to end up with a portfolio of 5 funds all involved with gold -- a sudden price shock to gold would crater the portfolio.
Incidentally, there are a variety of investment newsletters that do an upgrading strategy. The best of these is NoLoad FundX Newsletter http://www.fundx.com/customerhome.aspx . It's been rated by Hulbert's as the best performing investment newsletter over just about any time period going out to 30 years. That's best performing newsletter period, not best performing of this strategy.
(Full disclosure: they started a mutual fund, FUNDX, about 10 years ago that uses their strategy. I own some shares in FUNDX the mutual fund.)
Ok, so, what to do with all this nice data you collected? I can't give you some formula that you can plug in and walk away. I do have some suggestions on how to build out your stratagy, though.
Clean the List:
First, clean that list of 100 mutual funds. Exclude any that you would not want to invest in, regardless of what the numbers say. Some ideas of what you could exclude: over specialized sector funds, active funds where the manager's recently left (thus making historical returns irrelevant), funds that, when you look at their prospectus just make you feel skeevy (your gut can sometimes pick up on incompetent BS artists where your head can't).
Second, categorize & group the remaining funds by investment style, sector, etc. This is so when you do have an investment signal from the strategy, you can avoid investing in duplicates.
Build the strategy
Let me start by saying that you should never finish this process. First because you can never know if you're right until after the fact, and second because the success rate of a set strategy will change over time. So, even after you've started investing with this, periodically come back and test new ideas, formula tweaks, etc.
Set yourself up with a good testing environment. You want to be able to run "what if I had invested this way" scenarios on your ideas to see what would have worked. So, either get a good back testing service, or create your own (you did say you were screen scrapping and stuff, so I'm sure you can program one:-) ).
Spend more effort on figuring out when to sell a fund rather than buy one. I was reading in one of Donald Trump's books (speaking of BS artists), something that stuck with me as wise: "Take care of the downside and the upside will take care of itself". So, when you run your scenarios, you're not looking for the one that would have made you the most money. After all, historical data is hindsight. You can tweak it to death to get the tippy-top return in hindsight, but the resulting strategy would have no bearing on reality going forward.
Instead, look for results that get a reasonable return, but minimize the drawdown (losses) and volatility (whipsawing) along the way. You want the strategy with the least drawdown and volatility for a given level of annual average return.
Start with a basic strategy, then theorize/build/test/tweak from there. For instance: invest when 3-month RSI exceeds 6-month, and 1-year exceeds 3-year, avoiding duplicates. Sell when 3-month RSI drops below 6-month RSI. See how that looks in testing, and go from there.
Make sure that that data you've collected accounts for dividends. Remember that when you receive dividends from a mutual fund (which are no doubt auto-reinvested) the price per share goes down by that amount, but you didn't lose anything. Depending on the fund, that can make a huge difference in change-in-price vs. actual-gain over time.