• How To
    • Win Your DFS League
    • Win Your Auction Draft
    • Win Your Snake Draft
    • Download Projections
    • Scrape Projections
    • Calculate Projections for Your League
    • Examine Accuracy of Projections
    • Identify Sleepers
    • Save Custom Settings
    • Use the API
  • Strategy
    • Fantasy Football is Like Stock Picking
    • Use Projections, Not Rankings
  • Projections
    • Our Projections
    • Who has the Best Projections?
    • Draft the Best Starting Lineup
    • Projections are More Accurate than Rankings
    • Points by Position Rank
    • Players’ Risk Levels
    • Value Over Replacement
    • Bid-Up-To Value
    • Player Value Gap
    • Gold Mining
    • Weekly Variability
    • Are Subscription Sources More Accurate?
  • Statistics
    • How To Learn R
    • R is Better than Excel
    • Do Stats Help in Fantasy Football?
    • Download/Run Our Scripts
    • ffanalytics R Package
  • Apps
    • Auction Draft Optimizer
    • Snake Draft Optimizer
    • Weekly Lineup Optimizer
    • Rankings/Projections for Your League
    • API
    • Other Tools
      • Stock Analysis
    • Error Logging
  • Testimonials
  • About the Site
    • About
    • Authors
      • Isaac Petersen
    • FAQ
    • FFA Insider
    • Privacy Policy
    • Terms of Service
  • Donate

Fantasy Football Analytics

Are Subscription Sources More Accurate?

21
  • by Dennis Andersen
  • in Projections · R · Theory
  • — 14 Apr, 2016

Introduction

In this article, we examine whether there are advantages to paying for subscription fantasy football projections. We tested whether projections from subscription sources have higher accuracy than projections from free, publicly available sources. There are arguments that subscription sources would possibly be more accurate as you may expect to get better accuracy as part of what you are paying for. We examined projections from the 2015 season for QB, RB, WR and TE positions.

Overall Accuracy

We calculated the projected seasonal points based on standard scoring settings as used in our Projections tool and compared with the actual points. For the aggregation of sources we used the regular mean. We have 10 free sources and 6 subscription sources. The free sources were: CBS, Yahoo, ESPN, FOX, NFL, FFToday, NumberFire, EDS Football, WalterFootball and RTSports. Because the subscription sources are not publicly available we chose not to disclose the names of the sources. For each of the groups we calculated R2 (higher is better) and MASE (lower is better) values as well as values for both groups combined. The results are below.

Overall accuracy
Source Type R-Squared MASE
Free .635 .548
Subscription .618 .568
All .641 .541

Based on the results, the subscription sources were less accurate than the free sources but add to the overall accuracy. In light of that, one possible reason the subscription sources were less accurate could be because there were more free sources than subscription sources. To investigate that possibility, we examined the accuracy of all possible combinations of 6 sources among the free sources. The results below show the mean R2 and MASE values for all the combinations. Reducing the number of free sources did reduce the projections’ accuracy, but the free sources were still more accurate than the subscription sources. In other words, even after accounting for how many sources of projections were included, free projections were still more accurate than subscription projections.

All Possible Combinations of 6 Free Sources
R-Squared MASE
Combinations of free sources .631 .556

Accuracy of Individual Subscription Sources

We then examined whether any of the individual subscription sources was more accurate than the crowd.

Accuracy of Subscription Sources
Source Type R-Squared MASE
Subscription 1 .567 .616
Subscription 2 .576 .608
Subscription 3 .588 .612
Subscription 4 .587 .615
Subscription 5 .599 .586
Subscription 6 .552 .641

None of the subscription sources was more accurate than the crowd of free projections (R2 = .64, MASE = .55) or all projections (R2 = .64, MASE = .54).

Position Accuracy

Let’s examine whether the results are different when we look at individual positions:

QB Accuracy
Source Type R-Squared MASE
Free .707 .416
Subscription .677 .434
All .705 .414
RB Accuracy
Source Type R-Squared MASE
Free .488 .673
Subscription .492 .686
All .503 .657
WR Accuracy
Source Type R-Squared MASE
Free .613 .575
Subscription .560 .633
All .616 .571
TE Accuracy
Source Type R-Squared MASE
Free .550 .583
Subscription .541 .592
All .559 .572

For the QB, WR, and TE positions, the free sources were more accurate than the subscription sources, while the subscription sources were slightly more accurate for the RB position measured by R2 but slightly less accurate as measured by MASE. For every position except the QB position, combining the free and subscription sources also increased the overall accuracy.

We also calculated all possible combinations of 6 sources among the free sources by position. As the results below show, the accuracy for the free sources did decrease.  However, as was the case with overall accuracy, the free sources were still more accurate than the subscription sources even after accounting for how many sources of projections were included.

All Possible Combinations of 6 Free Sources By Position
Position R-Squared MASE
QB .702 .422
RB .482 .688
WR .605 .585
TE .548 .589

Conclusion

We have seen that subscription sources are not more accurate than the free sources. In general, free projections were actually more accurate than subscription projections.  However, including subscription projections did improve the accuracy of projections both overall and for each position (except quarterbacks). As we have demonstrated before, individual analysts do not reliably beat the “Wisdom of the Crowd” and this analysis further supports that—none of the subscription analysts beat the crowd. So while free sources seem to be more accurate, on average, than subscription sources, it is combining them that adds to the accuracy of the overall projections. The most accurate projections combined free and subscriptions sources.  So if you are asking whether you should use free or subscription sources for your projections, the answer is: use both!

You can find data and script for the analysis here.

Share this:

  • Click to share on X (Opens in new window) X
  • Click to share on Facebook (Opens in new window) Facebook
  • Click to share on Reddit (Opens in new window) Reddit
  • Click to email a link to a friend (Opens in new window) Email

Like this:

Like Loading...

Related

Share

Tags: accuracyprojections

— Dennis Andersen

My name is Dennis and I am a Dane now living in the US who has been hit by a fantasy football bug. I have a background in Mathematical Economics and enjoy the opportunity to apply statistics and programming to something that can be used in a real life scenario.

21 Comments

  1. Jason says:
    April 17, 2016 at 10:07 am

    Great read! Keep these articles coming

    Reply
  2. Cris Benson says:
    April 18, 2016 at 10:42 am

    Dennis, when I look at the results I begin to wonder if the results reflect a positional bias? Are subscriptions services focused on RB based on historical performance? And are free services concentrated on overall value of the player and not the position? The great thing about this information is it opens up the discussion of what information is being valued among all the sources. Is it analytics, beat writers report, is it the analyst that serve on staff? What is so different? I value the opinions of the analyst just as much as the analytics. Analytics in football often has too many variables as compared to baseball driven analytics,

    Reply
    • Dennis Andersen says:
      April 19, 2016 at 12:02 am

      Hi Cris, I don’t think there is a positional bias. I think there is a lot of information available out there for all analysts if you go look for it. I think the differences reflects the individual analysts perception and interpretation of the information available. All the analysts have the information about previous performance, so I believe the difference come into play when you take the off-season information into account and try and apply that when projecting the performance for the next season. So the differences in projections probably reflect more of a difference in opinion about the players than difference in available information.

      Reply
  3. Jeremy says:
    April 24, 2016 at 7:55 pm

    This is a fantastic article Dennis. I really appreciate you guys putting this stuff out. This is something I have wondered for a few years. I assumed pay would be more accurate and if I was running an optimizer I wouldn’t want to contiminate the pay projection by adding them. Turns out I was very wrong.

    Thank you

    Reply
  4. Jeremy says:
    April 24, 2016 at 7:56 pm

    *adding free ones

    Reply
  5. Timothy says:
    May 3, 2016 at 8:05 pm

    Interesting article. I really like the work you guys do on this site.

    It’s not obvious to me from what you’ve shown that there is not any single subscription source that outperforms the crowd. Can you show the table of individual performance compared to the performance of the crowd? (similar to the ones here: https://fantasyfootballanalytics.net/2016/03/best-fantasy-football-projections-2016-update.html )

    The reason I think this matters is because subscription sources are usually contrasting their performance with that of other subscription sites. So in terms of their marketing claims it doesn’t matter how a group of sites perform, it only matters how each one individually performs.

    Thanks again for taking the time to share your analysis with us.

    Reply
    • Isaac Petersen says:
      May 4, 2016 at 7:42 am

      Hi Timothy,

      Great question. We updated the post with the accuracy estimates of the individual subscription sources. None was more accurate than the crowd.

      Hope that helps,
      Isaac

      Reply
      • Timothy says:
        May 4, 2016 at 10:26 am

        Thanks, Isaac! It’s nice to see more confirmation that the individual sources do not reliably beat the crowd. Keep up the good work!

        Reply
  6. Jason says:
    May 5, 2016 at 7:09 pm

    These accuracy results can be misleading, since it very much depends on which list of players your comparing to. I could do the same study but my group of players may vary and get different results. Case and point, I have done a similar study and used 15 different sources for 2015 (based on 200 players). FFA was the most accurate site for preseason results with a .409 r2 – the average for accuracy among all the models was .347 and the low was .174.

    I applied this to positions to see if any position was easier to predict than others. QBs were by far the most accurate in the 70% range. All the other positions fell in the 60% range. However, despite FFA’s strong overall accuracy, this is how they scored: QB – 7th, RB – 5th, WR – 9th, TE – 15th, K – 5th, D – 6th. I haven’t really figured out if being accurate at position is better than being accurate overall.

    I also applied this accuracy study to a week by week time frame. FFA once again was the top model, placing in the top 3 every single week. They also had the highest weekly accuracy of .572 over the 16 week season.

    This site is outstanding. I have done my study every year for 4 years, FFA is always a top producer in accuracy. Very impressive work.

    Reply
    • Isaac Petersen says:
      May 5, 2016 at 7:42 pm

      Hi Jason,

      Good point—the accuracy depends on which positions and players are examined. We examine the accuracy of all available players we can find projections for among the most common positions across people’s fantasy football leagues (QBs, RBs, WRs, and TEs). Nevertheless, you can examine our (and others’) accuracy for other positions using our tools:
      https://fantasyfootballanalytics.net/2015/07/accuracy-of-fantasy-football-projections-interactive-scatterplot-in-r.html

      In general, we (and others) are more accurate for seasonal than weekly projections, which makes sense because there are fewer plays (and hence, more randomness) inherent in weekly than seasonal projections. If you have more accurate sources of projections that we might be able to include, please send them to us, and we’ll do our best to include them!

      Thanks!
      -Isaac

      Reply
  7. Jason says:
    May 6, 2016 at 12:25 pm

    Three more to consider are: FBG (Football Guys – Draft Dominator is free till before the season starts – you can export those projections), FS (Fantasy Sharks) and FP (Fantasy Pros). CBS does IDP projections, I could send those to you for both the season and the week.

    How would you like those, via email, what format (xlsx, csv). I’m still struggling to learn R, so I’m still an excel guy, until I finally make the plunge.

    If you’re interested, I built a model that is a stand alone, it quantifies a projection solely off of historical data. Been using it for 2 seasons. It does a pretty good job and rivals your weekly results. It’s a lot of work to maintain because it’s based on last 3, 5 and YTD (scored and allowed) data for every position, including IDP. But in 2015, it received a .408 r2, which was only .001 less than FFA on my study. It also did a better job when judging accuracy on individual positions. Regardless, it’s still in beta. If you would like more models for the aggregate. I’m open to helping.

    Reply
    • Isaac Petersen says:
      May 6, 2016 at 8:34 pm

      We have plans to include many of those projections. To the best of my knowledge, we already include all sources of projections included in the FantasyPros projections. What’s the link to the CBS IDP projections? Feel free to send me an email if you want to help out!

      Thanks!
      -Isaac

      Reply
  8. Jason says:
    May 9, 2016 at 2:28 pm

    My league hosts through CBS, but unless you are in a IDP setup, not sure CBS publishes them for the public. But since there are so few sites that create IDP projections, every one counts right? I’ll email them to you once they become available.

    Reply
    • Isaac Petersen says:
      May 9, 2016 at 8:12 pm

      Hey Jason,

      Thanks for your help! The challenge is that we’d need a way to update the projections when CBS updates them. Do they not have them publicly available?

      -Isaac

      Reply
  9. Michael says:
    August 14, 2016 at 12:28 pm

    Hey Isaac, have you considered adjusting for the number of “analysts” involved in developing each single site’s projections? For instance, if CBS’ projections are basically the average of 2 individual’s projections and ESPN’s projections are basically the average of 5 individual’s projections, would you want to look into weighting those projections differently in your final average projections?

    Reply
    • Isaac Petersen says:
      August 14, 2016 at 5:31 pm

      Hi Michael,

      We weight based on historical accuracy. Presumably, the more sources are included, the higher the historical accuracy, so it should be accounting for the number of sources (to the extent that it matters).

      Hope that helps,
      Isaac

      Reply
  10. Huey (@hueykwik) says:
    September 18, 2016 at 1:04 pm

    Why did you exclude Defense?

    Reply
  11. Randy D says:
    December 4, 2016 at 8:12 am

    Another free source that you might want to consider adding to your weekly rankings database is Justin Boone from The Score. His overall accuracy has consistently been rated in the top 10 over the last few years according to Fantasy Pros.

    Reply
    • Val Pinskiy says:
      December 6, 2016 at 5:42 pm

      Thank you for the suggestion but we need projections, not rankings, for our tools.

      Reply
  12. Mike W says:
    August 25, 2021 at 1:04 pm

    Is there a way this could be broken down into perhaps finding WHERE projections are most/least accurate? Like perhaps certain sources are most accurate for the top 10 or so QBs, but the next 22 are just all over. Or conversely very accurate on game to game basis, but just fail to capture the monster games and duds. I’m thinking even one major early season ending injury could drastically skew someone’s accuracy if they had projected that player for a career season.

    Reply
    • Noah Schwartz says:
      September 21, 2021 at 7:18 pm

      Hello Victor,

      We try to pull data from as many sources to give as accurate as possible projections, thus using a wisdom of the crowds approach.

      Reply

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

  • Previous story Which are More Accurate: Fantasy Football Rankings or Projections?
  • Next story 2016 Fantasy Football Projections
  • Tabs

    • Most Popular
    • Recent Posts
    • The ffanalytics R Package for Fantasy Football Data AnalysisJune 18, 2016
    • 2015 Fantasy Football Projections using OpenCPUMay 28, 2015
    • Win Your Fantasy Football Auction Draft: Determine the Optimal Players to Draft with this AppJune 14, 2013
    • Win Your Fantasy Football Snake Draft with this AppSeptember 1, 2013
    • Post-Combine Mock: Team Needs and TargetsMarch 10, 2025
    • Fantasy Football Weekly Cheat Sheet: Week 18 (2024)January 3, 2025
    • Fantasy Football Weekly Cheat Sheet: Week 17 (2024)December 26, 2024
    • Fantasy Football Weekly Cheat Sheet: Week 16 (2024)December 18, 2024
  • FFA Insider

    Logo
  • Categories

    • About the Authors
    • Articles
    • Auction Drafts
    • Draft Optimizer
    • FFA Insider
    • Gold Mining
    • How To
    • In the Media
    • Luck
    • Package
    • Projections
    • R
    • Risk
    • Theory
    • Tools
    • Trade Strategy
    • Uncategorized
    • Weekly
  • Facebook

  • Twitter

  • Our Partners

    R-bloggers

  • Support us building things... Even a cup of coffee ($1.99) helps us stay awake!

  • Subscribe to the Fantasy Football Analytics mailing list (no spam).
    Loading

        © Fantasy Football Analytics

        %d