Sports Illustrated Effect

Sports Illustrated Effect

I want to explore why some players get us the Fantasy Football points season after season, and some are one season wonders. By looking into this I hope to find clues to help us Fantasy Football Managers pick the right players. Maybe the findings will be useful on a short term basis too, helping us to make some rules for picking players over 6 weeks and not just the whole season. 

In my last post, “Types of players not to choose in Fantasy Football“, I set the stage for this analysis by looking at players who earned lots of points for us managers in the 2011/12 premier league season, but did not deliver in 2012/13 and tried to find if there was something in common with these players. I used price drop as a way of spotting one season wonders. The analysis was not as fruitful as we would have hoped and left more questions than answers. However, it did give us two top tips.

  1. Players who do not play regularly are more exposed to price fluctuations than regularly playing players and should be avoided when choosing your team in August so you do not lose money early in the season; but these players are worth keeping an eye on as if they become regular and their price has sufficiently dropped then they could be a cheap source of points later in the season.
  2. ‘First team’ players with less 12/13 season game-time than previous seasons could be cheap picks in 13/14 if they are in teams where you feel they will play regularly. Nani, Rosicky, Downing and [at a push] Gervinho were our first spots but there could be more.

So I’d like to pick up where we left off, and ask “of the left over players in the analysis, why was each player’s 11/12 Points Per Million Per Game so similar to their 12/13 Points Per Million Per Game?”

To remind you, the table showed us

Name Team Price
Change
Points
Diff
Mins
Diff
PPMPG
11/12
PPMPG
12/13
PPMPG
Ever
Sessegnon SUN -1.3 -20 -238 0.58 0.656 0.672
Sigurdsson TOT -1.2 -25 -260 0.682 0.719 0.881
McClean SUN -0.9 -4 498 0.758 0.665 0.818
Silva MCI -0.8 -37 -290 0.591 0.573 0.593
Valencia MUN -0.8 -54 83 0.659 0.428 0.609
Cisse NEW -0.8 8 1893 0.899 0.39 0.566
Ramires CHE -0.8 -5 51 0.551 0.581 0.539

McClean and Valencia failed to score as many points in 12/13 as in 11/12 with the same or more game time, explaining the price falls. Indeed without the price falls the PPMPG would have been so much worse. Moreover, we can see the 12/13 PPMPG was poor when compared to the long term average.

Sigurdsson and Silva both scored similar PPMPG in 11/12 as they did in 12/13, so can we assume that if they had the extra 250 minutes of game time their prices might not have fallen? Judging by the long term PPMPG we should expect the prices not to have fallen so much.

Sessegnon did better in PPMPG in 12/13, and if he had achieved the extra 238 minutes would his price have fallen? His long term average suggests not and implies he performed as normal in 12/13. For Ramires, his long term PPMPG suggests his performance is right for his closing price, so he was probably over valued at the stat of the season.

Cisse is hard to judge due to only featuring in a season and a half in the premiership, however, an average long-term PPMPG of 0.566 is OK so perhaps a good season is just being balanced by a bad season.

So how important is long term average?

It seems from the four groups of players above that players might have a level; a rate at which they are likely to score points season in season out. Is this right?

Clint Dempsey on the cover of SITo answer this I would like to introduce a phenomenon called the “Sports Illustrated Effect”, or the SI Cover Jinx. To appear on the cover of SI the individual or team must have done something exceptional, for example, a striker scoring more goals than any other player in a world cup is likely to be invited to be on the cover. The Sports Illustrated Effect suggests that individuals or teams who appear on the cover of the Sports Illustrated magazine will thereafter be jinxed.

The Premier Fantasy League equivalent would be a player who had a great year and earned lots of points for us managers in a Premier League season, but then did not deliver the goods in the following season. They could be seen as Fantasy Football jinxed one-season wonders. I am sure we can all think of players who fit this criteria. Stephen Ireland, Jerzy Dudek, Papa Bouba Diop, Michael Ricketts, Roque Santa Cruz, Juan Pablo Ángel (to be fair he had two good seasons), perhaps even Michu from the first half of last season? Maybe Cisse and Giroud too?

So does this jinx exist and how can we spot it? Funnily enough, yes, it does exist and there is a mathematical proof for it. The argument goes

  1. a player will perform at a relatively consistent level;
  2. the level for some players is a lot higher than for other players, and this separates great players from good players;
  3. sometimes a good player will perform really well (at the level of a great player) and get noticed;
  4. we should not expect exceptional results to continue for ever, but instead should expect the form to regress back to the old level at some time.

Standard DistributionTo explain, the graph describes that probability that a player’s performance will be below average, average, or above average. A higher green line on the curve indicates a higher probability; and the highest probability is for an average performance.

It is important to recognise this “average” is the player’s average, not yours or mine. It applies to the player’s lifetime performance capabilities and simply says that on most days the footballer will perform at his average level.

Likewise, this average is different for different players, so the average for Gareth Bale will be different to the average for Hatem Ben Arfa, and the average for Phil Jagielka will be different to the average for John Arne Riise.

For any given footballer, there is a probability that he’ll perform exceptionally well in the next game. But if you had to make a bet to choose a player for Fantasy Football, the best thing to do is expect his performance in his next game will be just average. There is some correlation from game-to-game, meaning that if Ben Arfa played way above average last week, he might be above average this week – but statistically the tendency to repeat an above-average performance is not that strong, depends on the opposition, and your best option is still to expect an average performance.

Sports Illustrated JinxSports Illustrated is “implicated” in this theory because the magazine features players at their peak. The player then regresses back to their old level and Sports Illustrated get the blame [no more tipping players Chipper as we do not want an FFF Jinx!]

So the key to spotting a one season wonder seems to be finding a player who normally performs at one level earning so-so points per season, then suddenly performs really-well earning loads of points for one season, then drops back to the original level earning so-so points per season.

This peak of performing really-well could be one game, it could be a run of games, or it could be a whole season. Indeed the reverse is true in that a player may be on a bad run and we should expect them to return to average any time soon (for example, Carroll and Torres).

How does the Sports Illustrated Jinx help us?

Well if we know what is average, then we therefore know what is good, and what is great. We can use this knowledge to spot if a player is undervalued, or if a good player is on a great run. In Fantasy Football terms, an “undervalued player doing well” is the same as a “good player on a great run”. They are both players scoring more points than their Fantasy Football price suggests they should and we want them in our team to earn the most points for our £100 million.

So what is average?

We know in Fantasy Football terms we are looking for players who play regularly and who are scoring points above average for their price. We can work out what is “average” in terms of playing time and points scored over the last 7 seasons from the information published on the Premier Fantasy League website.

However, before we calculate an average lets look at the external factors that can impact the Fantasy Football points scored in any given season.

  • When a player swaps team he can get more or less game time, as we found with Clint Dempsey. His average PPMPG did not fall, but his game time did which can distort the stats.
  • Also when another player comes in, another player can also end up getting less game time which has an impact on the numbers.
  • When a player features in a new position, then the opportunity for points changes. For example, hands up if you remember when Bale was classed as a defender.
  • The bonus system changed in Fantasy Football two seasons ago and this might have an impact as it favored goal scorers.

So I am going to use a weighted average which favours recent seasons over previous seasons.

For points scored in a season, for the last 7 seasons, the weighted average number of points scored by average price is

Price GK DF MD FD
£4M 42 51.8 66.8
£4.5M 86.5 67.1 65.7 69.4
£5M 112.2 84.5 71.2 73.1
£5.5M 136.7 90 89.2 93.5
£6M 125.2 90.6 98.2 73.3
£6.5M 142.4 102 98.4 107.7
£7M 127 103.5 125.7
£7.5M 108.6 115 120.2
£8M 128.8 105.6
£8.5M 125 130.3
£9M 158.5
£9.5M 197.5 112.8
£10M 188.9 142.9
£10.5M 144.6 128.8
£11M 172.9 162.5
£11.5M 190.4
£12M 174

And for minutes played in a season, for the last 7 seasons, the average number of minutes played by average price is

Price GK DF MD FD
£4M 1052 1727.7 1303.8
£4.5M 2294.1 2093 1887 1331.4
£5M 2613.9 2326.5 1864.3 1764.1
£5.5M 3355.7 2274.5 2125.1 1927.6
£6M 2802.8 2134 2126 1394.9
£6.5M 3060.5 2196.1 2236.7 1993.5
£7M 2763.7 2008.9 2240.4
£7.5M 2167.8 2192.6 2192.7
£8M 2188.8 1914.7
£8.5M 2369.6 2070.9
£9M 2618.8
£9.5M 2963.5 1845
£10M 2630.7 2169.1
£10.5M 2399.6 2216
£11M 2383.1 2235.7
£11.5M 2353.7
£12M 2397.5

These points and times give us a baseline to measure players, and to identify players who are performing average, good and great. In my next article I will use the Sports Illustrated Effect theory and the statistical information above to try to identify the players who are likely to be the dodgy players next season, and ones who might provide good rewards next season, and to provide a way of measuring new signings. As with all stats, the information is only a guide to help you when picking your team for the 2013/14 season, but a very good guide non-the-less.

You can now read the final instalment of this series of articles “Choosing great Fantasy Football players“.


Thinking: fast and slowA little background

I was first introduced to this theory when reading “Thinking, Fast and Slow” by Nobel Prize winning economist Daniel Kahneman on holiday last year.

The book talks about why people are naturally not good at accurately estimating statistics, yet when we do we estimate stats we tend to believe are right (even though our estimates are usually wrong).

He introduced it to me as the statistical theory called “regression to the mean” which is summarised above. Kahneman went on to suggest

  • success = talent + luck
  • great success = a little more talent + a lot of luck

Success vs Great Success is a topic for another day. However, if you want to know more about his theories then get the book for your holiday/commute reading. Its a very good read and as a bonus each sale contributes a little to the running costs of the site so buying the book is in a good cause too.

Authors

12 thoughts on “Sports Illustrated Effect

  1. Great article! It’s passed the time well waiting around in the halls!

    Reminded me of my a-level stats, only you’ve got a way more interesting take on the subject than my lecturer did!

    I might have to purchase the book, but by luck I assume that means ‘success when there is a low chance of success’ & not luck being personal to an individual?! I’m not a big fan of the word the luck!

    I can’t wait for the next article. I may be asking too early for this answer, but, can your model predict a progressing average? So predict a young player’s average that is moving towards his average peak?

    • Hi Bowie,

      Thanks for the kind words. It would have been fun to use Football in some way when doing A-Level maths, rather than frequency of birthdays and other not-very-useful things.

      The model does not take into account progression as a player as maturity/experience sets in, which is why I use a weighted average which favour the more recent seasons over seasons 5 or 6 years ago. It would be interesting to do that, but probably over the top for a fantasy football website :)

      Success is success as a sporting individual – plays for big clubs, wins trophies etc and luck is the randomness of good fortune. The formula is designed as a tongue in cheek talking point to emphasis the point about regression to mean. I think you’ll enjoy the book.

      What have they said about your wrist?

      • Can you imagine the sort of people that would enjoy & use statistics more, if delivered in that or similar formats?!

        It would be hard to create a model fir it, too many variables, & slightly over the top. However, this may not produce significantly sound outputs, but if you put your model’s prediction at the end of a time-series graph of previous scores, it would allow you to see if you could predict a further boost beyond the ‘average’ you find.

        I have to wait a month for the results now! Now I’m radioactive I’m hoping I glow tonight!

  2. Holy crap, CR that is a true masterpiece you have just written there :)!

    Your article is extremely helpful in both identifying the obvious (picking up players with easy fixtures and ones on good form), but also mentions the understated (the value in cheaper players who are performing above their weight).

    The SI jinx is very similar to that of “The Madden curse”, which is that a player featured on the cover of Madden NFL often gets injured or regresses significantly in the year that game is released. I’m also glad you pointed out the caveats related to your theory to cover all necessary ground.

    Just wondering, but is there any form of diminishing returns in buying a player who was originally cheap and performing well, but then who’s value goes up as more managers bring in this player and they then start performing at or below their price? Or, are you better off recouping the increase in team value and re-investing it in the next player underpriced who may perform well, so as to increase team value quicker and allow you to ultimately buy better players?

    For example, let’s say Lambert started off the season at 6 million. As the season progressed, he performed very well and his price rose to 7.5 million, but at 7.5 million he is only playing to his price. Since we paid less, is it worth to keep this player OR are we better off buying say another striker such as Romelu Lukaku priced at 6 million if we’re to believe that he will eventually come good. Perhaps, I have answered my own question and it’s only worth dropping Lambert if we’re sure that we’re replacing him with another undervalued player producing at a level higher than their price…

    • Thanks HH.

      “is there any form of diminishing returns in buying a player who was originally cheap and performing well, but then who’s value goes up as more managers bring in this player and they then start performing at or below their price?”

      Players you purchase should be measured against their sale price, not current price; where as players you buy are measured against the purchase price. That way you know if it is worth selling or not. Make sense?

      “Or, are you better off recouping the increase in team value and re-investing it in the next player underpriced who may perform well, so as to increase team value quicker and allow you to ultimately buy better players?”

      Now that is an interesting question. Early in the season is it best to focus on the accumulation of pounds rather than points? That’s probably an article in itself and Chipper has strong views in this area so if he is free maybe he’ll write one.

      I think early in the season you can earn points and pounds by planing your next transfer in advance and then executing the transfer on a Saturday or Sunday night. This carries a risk as your new player might be injured in the week, but I’d expect it to work for most early season transfers. This is why ‘The Last Minute Transfer’ articles look at this & next gameweek’s transfer targets.

      Does that make sense? Do you agree, as there are many ways to play the game?

      • Cheers CR :).

        Definitely makes sense, I guess I was just thinking about how I kept Michu all season or another player longer than I should have because of the fact that if I dropped said player, it’d be more expensive to buy him back than to just keep him on my squad. However, it was difficult to gauge Michu’s production as he’d score goals in some weeks, then do nothing for a few weeks in a row, which made the decision tougher.

        Your site/stats will become invaluable if such a situation comes up next year as I can look at the player’s PPMPG and compare it to the performance graph you spoke about above.

        Would love to read that article by Chipper if he has time to write it. I really benefitted from taking advantage of undervalued players early on in the season to raise my team value (had Tevez, Michu, RVP, Mignolet, Begovic all from the outset last year), but struggled to find a consistency on when to let them go (in some instances I was not patient enough, others too patient).

        Also, despite this high team value, I clearly did not take advantage of it as I came 2nd in my money league and just outside the top 20,000 in the world. Thus it’d be outstanding to see a statistical comparison of earning points vs. pounds in the early season, and then furthering it by how to take advantage of the extra pounds to maximize points (if that makes sense).

Comments are closed.