I've just completed the first-draft of our NFL projections - minus the rookies - and I wanted to share a few thoughts on my process. It's quite a bit different than what you'll see on most other sites - at least as far as I can tell.

I will put my projected average (mean) projections in for players to a point, but as I get deeper into the draft, I start to fudge a bit. The players with more ceiling get a boost, and the players with more floor get pushed down. Imagine it's as though I'm using the 60th rather than 50th percentile projections in Round 7, the 70th percentile in Round 10, the 90th percentile in Round 14. Why do this? If a player has a 10 percent chance at getting 1,000 yards and a 90 percent chance of getting zero, shouldn't I give him 100 yards? And if so, shouldn't he be below a hypothetical player who had a 100-percent chance for 200 yards? It depends.

If you're in a 100-team league, where 200 yards has value, then yes he should. But if you're in a standard 12-team league where 200 yards has no value, then the guy with even a 10 percent chance at being relevant should be ranked ahead of him. Perhaps an even better - and more realistic example - is the Patriots kicker. What happens if Stephen Gostkowski tears his ACL in July, and the Patriots bring in two veteran free agent kickers, each of whom has a 50 percent chance of winning the job. If we use their mean projections, both will be below the Giants and Browns starting kickers. But if you're drafting, you know whoever wins the job is very likely to be a top-10 kicker, and if you pick the wrong one, you can always drop him for any number of freely available options. Accordingly, no one in his right mind would draft the Giants or Browns kicker ahead of the potential Patriots one even if he had a 50 percent chance of being cut.

This is a problem for strict mean-value projections systems because they're set up to give you the best projections as if having good projections were an end in itself. In fact, so much have projections become ends in themselves, there are projections contests into which you can enter and get graded at the end of the year (we don't enter.)

But why then do we do projections at all? Why don't sites just post a cheat sheet - which is all customers really need - and call it a day?

The problem is fantasy football leagues have all kinds of different settings, scoring systems and rules. PPR differs from distance scoring, six-points-per-passing TD changes the value of QBs. There's no way to do just one or two or five cheat sheets, each of which you'd have to update separately, that can possibly cover all the permutations. So to make our cheat sheets malleable and customizable we base them on projected stats. If I say Philip Rivers will throw 29 TDs, then whether those are worth three, four or six, you can customize his ranking relative to the other QBs and players in the pool.

Put differently, the projected stats are a means to creating a cheat sheet that's adaptable to a wide variety of league settings and applications. That's the only reason to do them if you're running a fantasy site. To the extent that they make your cheat sheet worse and someone using them likely to take the Giants kicker over the potential Patriots one  they are a liability.

So I don't bother trying to make them 50th percentile across the board, or to provide the most likely year-end ranking for all the players. I'm making them to power the best possible, customizable draft list in formats that are likely to be played by human beings on planet earth in 2018. Consequently, I don't care if the Patriots receivers have 40 projected TDs and Tom Brady has only 34. I don't care if some team's receivers have only 14 collective TDs, and the QB has 19. It doesn't have to match because I'm not trying to win an award for perfect real-life projections. To do so in my opinion, is to put the cart before the horse.

There is a way to bridge this divide, but I don't think people will be up for doing it any time soon - I know I won't. You could project each player for 10 seasons, his 10th percentile, 20th, 30th, 40th, 50th... 90th, 100th. And you'd design your cheat sheet to pull the 40th in the first round (high floor), 50th in the second through fourth, 60th in the fifth... 90th in the 13th (these numbers are just ballpark approximations.) You'd make that the default, but you could also have the cheat sheet percentile projections customizable by round, e.g., if someone wanted to use 50th percentile the whole way they could. Or if they wanted an even more aggressive or conservative setting, they could do that too. (Maybe five rather than 10 different seasons per player would suffice, but it's still a ton of work - and updating - all for the end of making the system internally coherent but only marginally (if at all) more useful for fantasy owners looking for a good cross-off list.

I know from doing these a long time there's a great temptation to have everything turn out elegant and neat, for all the receiving TDs to equal all the passing ones, for each player to be ranked equally based on expected value taking into full account both floor and ceiling. I also know that such systems require you to scroll way down the list after Round 10 to find the players you actually want on your teams.