# Wide Receiver Gold Mining – Week 14

12The graph below summarizes the projections from a variety of sources. This week’s summary includes projections from: CBS’s Dave Richard, CBS’s Jamey Eisenberg, Fantasy Football Sharks, Fantasy Football Today, ESPN, Picking Pros, Yahoo Sports and Fox Sports.

## Standard Scoring Leagues

### Week 14 Wide Receivers

From this graph be sure to notice:

- Jeremy Maclin, DeSean Jackson, Doug Baldwin, John Brown and Pierre Garcon are the five players with the
**largest upside**(as measured from their (pseudo)medians). For these players, some projections are placing much higher valuations than others. If you are projected to lose this week by quite a few points and are looking for a risky play that may tip the balance in your favor, these are players to consider. - Terrance Williams, Kendall Wright, Allen Hurns, Vincent Jackson and Reggie Wayne are the players with the
**smallest downside**, which suggests that while their median projection might not be great, there is less uncertainty concerning how poorly they may perform. So, if you are likely to win by a lot and want to reduce your downside risk, these players may deserve extra attention. - On the other hand, Calvin Johnson, Julio Jones, Alshon Jeffery, Josh Gordon and Jordan Matthews are the five players with the
**largest downside**this week. If you are planning on starting them, it may be prudent to investigate why some projections have such low expectations for these players.

## PPR Leagues

### Week 14 Wide Receivers

From this graph be sure to notice:

- Calvin Johnson, Josh Gordon, Donte Moncrief, John Brown and Pierre Garcon are the five players with the
**largest upside**(as measured from their (pseudo)medians). For these players, some projections are placing much higher valuations than others. If you are projected to lose this week by quite a few points and are looking for a risky play that may tip the balance in your favor, these are players to consider. - Kendall Wright, Allen Hurns, Vincent Jackson, Terrance Williams and Reggie Wayne are the players with the
**smallest downside**, which suggests that while their median projection might not be great, there is less uncertainty concerning how poorly they may perform. So, if you are likely to win by a lot and want to reduce your downside risk, these players may deserve extra attention. - On the other hand, Julio Jones, A.J. Green, Alshon Jeffery, Josh Gordon and Pierre Garcon are the five players with the
**largest downside**this week. If you are planning on starting them, it may be prudent to investigate why some projections have such low expectations for these players.

Share

I realize I could do the work myself, but…

Have you guys checked actual performance against recent gold mining graphs to see how well they fit?

Hey Tim,

Haven’t done that yet. That’d be a great thing for us to look at after the season. Let us know if you examine the accuracy of the projections and want to contribute your findings!

-Isaac

[…] Michael Griebe’s Fantasy Football Analytics archive reveals a consensus of experts haven’t downgraded Demaryius Thomas from his No. 3 rank among wide receivers over the last two weeks. The experts expected Thomas to score 16 points on the road in Kansas City, with many allowing for a heavy upside. […]

Hello Isaac,

Nice website. I stumbled upon it searching about fantasy football projection models.

I would like to build my own fantasy football projection model and was wondering if you could point me in the right direction.

I’m not a stats guy so I am starting from scratch. I have a statistics tutor lined up to help I was just wondering where you thought the best place to pull the date from would be?

Also, if you had any advice as to how to approach it. It’s a long term project I see myself tweaking and learning for years.

I was going to use excel instead but it sounds like you’re a big supporter of R. From what I hear a lot of the better players have projection models using excel do you know any who use R?

Thanks

Arash

Hi Arash,

When you say you would like to build your own “fantasy football projection model”, what do you mean? If you are talking about creating your own projections, here are some of my thoughts:

https://fantasyfootballanalytics.net/2014/06/custom-rankings-and-projections-for-your-league.html#comment-7640

https://fantasyfootballanalytics.net/2013/03/isaac-petersen.html#comment-26364

Not sure what “better players” use Excel, but here is a discussion of the many reasons to prefer R to Excel:

https://fantasyfootballanalytics.net/2014/01/why-r-is-better-than-excel.html

Hope that helps!

-Isaac

What exactly does the 95 percent confidence interval tell you and how exactly is it calculated? Thanks!

Hi David,

A description of the 95% CI and how it is calculated is below:

https://fantasyfootballanalytics.net/2014/11/gold-mining-explained.html

Let’s consider that the sample of projection sources comes from a larger population of projection sources (i.e., all projections sources). If we repeatedly sampled from these projection sources, the CI would include the average (pseudo-median) of the larger population (which we don’t have) 95% of the time. For your purposes, I would interpret the 95% CI as the likely range of forecasts for a player (not the range of possible outputs).

Hope that helps,

Isaac

Once you have a pseudo-median, how is the confidence interval actually calculated? Is there a formula for it? I am trying to get my head around it from a mathematical perspective. My teams are terrible this year so I don’t really need the football advice. But I do teach high school stats. Thanks!

A typical confidence interval (e.g., around the mean) is calculated as the mean +/- the margin of error:

http://www.gla.ac.uk/sums/users/jdbmcdonald/PrePost_TTest/confid2.html

Calculating the confidence interval for a pseudo-median is somewhat different because it is a rank statistic. Basically, you rank the numbers from low to high. You calculate the pseudo-median by taking the median of all pairwise means. If the lowest number (Xi=1) is 0 and the second-lowest number (Xj=2) is 10, the mean of this pair would be: (xi +xj)/2 = (0 + 10)/2 = 5. You would do this for all possible pairs and then take the median of the resulting values to get the pseudo-median. The confidence interval is based on the calculation of a shift/jump parameter that satisfies the 95% critical region. The confidence interval includes the set: {Dij = (Xi + Xj)/2; 1 <= i <= j <= N}. For more info, see the source article here: http://www.tandfonline.com/doi/abs/10.1080/01621459.1972.10481279

In essence, a confidence interval is the margin of error around a point estimate. It gives us an idea of the amount of uncertainty in our estimates. Nate Silver has a book that does a good job of explaining why it’s crucial to include uncertainty in our estimates:

http://www.amazon.com/Signal-Noise-Many-Predictions-Fail/dp/159420411X/ref=sr_1_1?ie=UTF8&qid=1417941460&sr=8-1&keywords=signal+and+the+noise

Glad to hear you’re teaching stats in high school. Fantasy football is a great way to make stats more accessible and interesting to students!

-Isaac

The Garcon and Baldwin combo helped someone win a million $ last Sunday… Interesting.

Hi Isaac,

If you don’t mind I was hoping you could explain the jump/shift parameter calculations, or perhaps you know of another link to refer to (the link above requires payment)? I’ve had a difficult time finding this info online, and while I don’t mind researching, I thought I’d ask to save a little time. 🙂 In any case, the site is great and I appreciate all the help.

Cheers!

Hey Derek,

I’m traveling and don’t have access to the formulas at the moment. In the meantime, see the source article for more info on the calculation of the jump/shift parameter:

http://www.tandfonline.com/doi/abs/10.1080/01621459.1972.10481279

You might also examine the R source code for the function to see how it’s calculated if you’re familiar with R.

Hope that helps!

-Isaac