Methodology for Fantasy Rankings

This post will probably be long, tedious, and boring. But I think it's very important that my methodology for ranking fantasy players is laid out in full. The process below applies to a standard 12-team, mixed-league, 5x5 rotisserie league, although the rankings could be easily adjusted for a league with more or less than 12 teams and AL- or NL-only. With that said, here is the process:

1. Create projections for every relevant player.

I wish I could tell you that I had some magic formula, but my projections are probably no better (or worse) than anyone else’s. How do I come up with them, you ask? Well, I read the Bill James Handbook, I look at Baseball Prospectus’ PECOTA projections, and then I think long and hard. Basically, my projections can be boiled down to the following formula:

33% Bill James projections + 33% PECOTA projections + 34% my own little head

2. Find the historical average and standard deviation of each scoring category on a fantasy team basis.

I do this by using the last two seasons’ worth of data that I’ve collected by playing in fantasy leagues. That is 14 leagues, 12 teams per league for a total of 168 teams. I’m hesitant to go further than two years back for fear of changes in the way the game is played (fewer home runs now versus 2001, for example).

Here are the results:

3. Find how many standard deviations above average each player makes an otherwise average team in each category.

At this point, it would probably help to use an example. Let’s go with Victor Martinez since catcher is the only position where I’ve actually done projections so far. Here’s my preliminary projection for Martinez (subject to change):

145 G, 79 R, 23 HR, 105 RBI, 0 SB, .301 BA

Now, if Martinez was on a team with eight otherwise average fantasy players, here’s what that team’s stat line would look like:

764 R, 192 HR, 767 RBI, 97 SB, .286 BA

Now, we can calculate how many standard deviations above average that team is in each category. I usually multiply these by 10, just so they’re easier on the eyes. For Martinez’s “team,” here’s how it looks:

R: -0.83 HR: 0.54 RBI: 2.56 SB: -3.41 BA: 2.46

4. Add the standard deviations to arrive at Fantasy Value.

This is pretty simple. For Martinez’s “team,” it looks like this:

(-0.83) + 0.54 + 2.56 + (-3.41) + 2.46 = 1.32

At this point, you can rank every player based on his total projected Fantasy Value. This would be a fine place to stop if everyone played the same position. However, some positions are deeper than others and we have to take that into account with the next step.

5. Find Fantasy Value above a replacement-level player at that position.

So what makes someone a replacement-level player in fantasy baseball? Basically, it’s the best player you could reasonably expect to find available on the waiver wire. This is very difficult to determine most leagues have utility positions and bench spots. Based on my experience, I've assigned a certain number of players as generally being owned at each position. Yes, it's somewhat arbitrary, but that's just the way it goes sometimes.

The other complication is that if there is an extremely large drop-off at replacement-level, it could skew the results. To adjust for this somewhat, I use the average of the Fantasy Values of the three players closest to replacement-level at each position.

The replacement level at catcher turns out to be (-13.33). Thus, Victor Martinez’s Fantasy Value Above Replacement Player (FVARP) is 14.65 (1.32 - [-13.33]).

6. Rank players by FVARP.

There you go. Position-adjusted rankings.

Is it a perfect system? Of course not. There are plenty of flaws, mainly arising from the projections which are going to be wrong to some degree because I can’t see the future. Also, defining replacement-level is a tricky game and guaranteed to create some complications. I’m always open to suggestions on improving the system.