A Rookie at Doing Projections

[mikedemelo 0]

This is going to be my 7th season playing FF. I've done quite well over the years winning the league title twice and finishing in the top 3 two other years. Generally I depend on various sources to pull together my projections and then build my draft plan and tiers from there. This year I decided that I was going to start doing my own projections. I've always shyed away because I was never sure how people came up with the numbers they did. I've asked people before how to project and the answer I generally get is "What do you think X player is going to do this season?" Well, that doesn't help. Projecting JLewis to rush for say 850 yards and 4 TDs is not just pulled out of the air. There is obviously a 'method' to projecting. This 'method' is what I would like to learn so that I am in complete control (for the most part) and whatever happens is a result of what I did and not the result of someone else's good or bad projections. So here's where I began....

 

For every team I recorded the 2005 stats for QBS, RBs, WRs & TEs.

 

For QBs I also determined completion %, TD% and INT%

 

For RBs I calculated the percentage of the time they were used for rushes (# rushes for player X / # team rushes) and their TD%

 

For RBs, WRs & TEs I calculated TD% and the % of total receptions.

 

I then used a 3-year weighted moving average to project the number of rushes and passing attempts for 2006. Based on these numbers I then calculated number of passes, rushes and receptions for each player and then determined yards & TDs.

 

This was my first attempt at projections. Overall I am quite pleased but I've already found a number of holes in this system. How do you account for players on new teams or for new coaches with different systems? What about age? What about players who had an extraordinary year in 2005 or a really bad one that is completely out of character? What about if a player was injured last year? And how do you project for rookies?

 

Any feedback on the system I put together or tips for improving are greatly appreciated. I'd really like to be able to put together a solid method to use moving forward.

[Wildman 0]

This is going to be my 7th season playing FF. I've done quite well over the years winning the league title twice and finishing in the top 3 two other years. Generally I depend on various sources to pull together my projections and then build my draft plan and tiers from there. This year I decided that I was going to start doing my own projections. I've always shyed away because I was never sure how people came up with the numbers they did. I've asked people before how to project and the answer I generally get is "What do you think X player is going to do this season?" Well, that doesn't help. Projecting JLewis to rush for say 850 yards and 4 TDs is not just pulled out of the air. There is obviously a 'method' to projecting. This 'method' is what I would like to learn so that I am in complete control (for the most part) and whatever happens is a result of what I did and not the result of someone else's good or bad projections. So here's where I began....

 

For every team I recorded the 2005 stats for QBS, RBs, WRs & TEs.

 

For QBs I also determined completion %, TD% and INT%

 

For RBs I calculated the percentage of the time they were used for rushes (# rushes for player X / # team rushes) and their TD%

 

For RBs, WRs & TEs I calculated TD% and the % of total receptions.

 

I then used a 3-year weighted moving average to project the number of rushes and passing attempts for 2006. Based on these numbers I then calculated number of passes, rushes and receptions for each player and then determined yards & TDs.

 

This was my first attempt at projections. Overall I am quite pleased but I've already found a number of holes in this system. How do you account for players on new teams or for new coaches with different systems? What about age? What about players who had an extraordinary year in 2005 or a really bad one that is completely out of character? What about if a player was injured last year? And how do you project for rookies?

 

Any feedback on the system I put together or tips for improving are greatly appreciated. I'd really like to be able to put together a solid method to use moving forward.

 

I think your method is already very solid, especially the way you broke down the percentage of attempts, scoring, ints, rec, by position. Very very good. You should be showing others how to do this.

 

As for age? I wouldn't do much to factor it in unless you want to find 10 Rbs that played for 10-12 years as as starters that have a lot in common that you can quantify and then look for common trends up or down by year. I think you'll find the results too unpredictable though. Injuries? Look at the prior year as your baseline. Coaching changes? Look at the team the coach comes from and compare to the team it's going to..are the players better on the previous team or better on the new team. Look at 2-3 years of stats if you can. For instance Martz to Detroit will probably put Williams in the range between Bruce and Holt in terms of numbers, but I don't think the Lions have another polished receiver with the talent to produce like the Rams secondary receiver over the years. Less numbers, more common sense here.

 

Rookies...check out my rookie impact series as a start. Unless the players are starting, I don't bother with projecting them too seriously. Click on my name at the top of the article to find the other positional articles. This should help you get started with a sense of what happens historically for rookies. With starters, use the previous year starter on that team as the baseline add more productivity if he's starting over a journeyman or guy on the downside of his career in termsof actually numbers, make it lower if the guy was a free agent and signed elsewhere or was hurt. Those are a couple of general thoughts.

 

Overall, I think you're doing very well...I might be asking you for projection advice!

[BigGuy 0]

Wildman - I have a followup question if you don't mind. Do you have a feel for how much variation there is in the results of the the projection methods. E.g. How much is the Compiler/Draft Buddy the ones used on seemingly good sites. I pretty much used the FFToday Compiler last year tempered with common sense and gut feel. I did well but was in free leagues with some bad teams.

 

Have you or anyone actually validated the different models against year end results?

 

Scott

[Wildman 0]

Scott,

 

Good question. If I understand the specifics of it correctly, I think you're asking what rate of error the projected numbers had to the season totals. It's always a nice idea to look at this information and I know I've thought about it with my own projections. In fact, I began embarking on this type of analysis for my Crank Scores and Projections but didn't really know what would be the best way to do the analysis. I still want to do something like this, but I want to make sure I'm being fair with the analysis...

 

As far as I know this hasn't been done. I imagine for a few reasons and I'm not speaking for Mike Krueger who does the FFToday projections--just a wild guess:

 

1. The time involved to look at information.

2. Broadcasting your error rates when no one else does. Even if you are better than everyone else, it's not a great marketing strategy for a product if no one else is held equally accountable.

3. Having an agree upon method to list the variation and define what is poor, fair, good, above average, excellent for this industry. Unless we begin grading every site's projections on the same scale and then rate ours to theirs, I'm not sure how good this would be. Of course what I just mentioned seems like a pretty good idea. Not sure I'm ready to sink a few hundred bucks into purchasing subscriptions to get their final projections though

4. Looking at the past when you are getting ready for the future.

 

Anecdotal evidence obviously isn't as nice as statistical proof, but I can tell you that Mike Krueger and Mike MacGregor are very competitive fantasy owners that use their projections and tools that are on the site to compete with quality fantasy sites and they have a very high winning percentage and pretty good record of making championship games, if not winning them.

 

Any suggestions you might have in this department would be appreciated...