Well, baseball works the best because there are by far the most statistics to work with in projecting numbers. In fact, you don't even need to make your own projections in baseball. You can go look at Streamer projections or ZiPS projections and just plug those numbers in. Those numbers are very much different from "expert projections" in football, hockey, or basketball. They are based upon significant sample sizes and a much more repetitive process (batter facing pitches/pitcher facing batters). Sure, circumstances change somewhat in a given at bat or from year to year with the protection one has in a lineup, but it is way, way easier to predict (accurately).
Rotisserie vs. head-to-head is also important. Take the LGP Fantasy Baseball league. I now have three prior seasons of standings that I can look at. If it took 250 HR to finish first in HR in one season, 242 in another, and 255 in another, I can pretty definitively declare that if I have a lineup that can hit 250 HR, I should be able to count on 11 or 12 points in that category. It doesn't matter to me if I get 100/250 HRs in the first month of the season or evenly throughout. I know I need to have a lineup that can combine for 250 HRs when all is said and done. With head-to-head, you cannot predict what you need in a given week because it depends upon what your opponent does. The best that you can do (in my experience) is calculate the average weekly score for each team in your league from the prior season and work with that. Don't even bother trying with head-to-head categories (vs. points). That is just far too much data to assimilate into anything usable.
I am also in a fantasy football league with a salary cap (auction style). Despite what I said above, I still do exactly what you are discussing before entering the auction draft. I look at the average weekly scores from the year prior and loosely determine what per week score I need to get. Lets say that I want 100 points per week from my team. If I play 16 weeks in a season, that means I need a team with 1600 points on the year. If I have $100 to spend in order to get 1600 points, that means that I need to get 16 points/$1 I spend. That is your big number that you can work with.
If Aaron Rodgers produced 160 points last year, I can spend up to $10 on him. Any more than that and I am overpaying, any less and I am getting a steal (allowing me to overpay on someone else later in the draft). I create a spreadsheet which assigns a FMV price for each player based upon last year's stats. This is a very useful guide for heading into your auction draft.
Still, there are several flaws.
1. You are counting on last year's stats. Again, this is very tough to do in football and hockey.
2. You are counting on last year's stats.
3. You are looking at total production, which ignores week-to-week consistency, which is critical in head-to-head.
4. Your model is premised upon total production and you could go out and put up 1800 points and miss the playoffs if the weeks/matchups fall the wrong way.
Still, its useful. If Matt Stafford is an $8 player based upon last year's numbers and he is due for a massive progression in TD (and therefore points), I need to be aware of that to know that I should be willing to maybe spend a bit more on him. In other words, the spreadsheet doesn't replace your research on how this year will differ from last year for all the players. In baseball, you could theoretically never watch games or the players that you are choosing and pick a team based purely on stats and projections and have a shot at winning.