Arsenal News
Arsenal News & Transfers
As featured on NewsNow: Arsenal newsArsenal News 24/7

Arsenal News, Only Arsenal, Blogs, Transfer News

Archives

September 2017
M T W T F S S
« Aug    
 123
45678910
11121314151617
18192021222324
252627282930  

Intentional misrepresentation, misunderstanding of statistics, or…?

 

By Sifarzone

The title of the article is setup for the reader to make their own interpretation of the facts presented below.

Recently, I have been miffed about the anti-Arsenal bias presented on ESPN, a worldwide, well-known and well respected network. However, such respectability seems not to have extended past their baseball, football (American) and basketball pages.

One such article (Untold also covered this article) claimed that 80% of Arsenal supporters wanted Arsene Wenger to leave, but just clicking on the link reveals that the only people that were polled were the Arsenal Supporter’s Trust (AST) a blatantly anti-Wenger group and the only surprise is that they did not get 100% wanting Wenger out.

I would like to stress that I am not saying that the ESPN writers have an agenda, but some of the statistical interpretations are either misunderstood and therefore improperly represented, or intentionally misrepresented to mislead readers.

On Monday, a post with the headline “Giroud may start for France Lacazette still better option for Arsenal” (1) appeared on ESPN, and used statistical reasons to explain why Lacazette is a better player than Giroud. Now I am not claiming whether Lacazette is a superior or inferior player, just that I believe the statistical explanations in the article were flawed and improperly presented.

  1. Lacazette has a higher scoring rate in league play.

On the surface, this is incontrovertible evidence as to who is the better striker. However, let us  take a look at the non-penalty goals, as that is a more useful comparison since Giroud is not the main penalty taker in the team and also penalties on the team can be taken by other players rather than just the striker. In terms of non-penalty goals, Lacazette has 18 goals last season (2)(3). Giroud being second option for most of the season had 12 non-penalty goals (4). In terms of minutes per non-penalty goal, Giroud played 1,194 mins (5), whereas Lacazette played 2,411 mins (6). So effectively, Giroud scored every 99.4 minutes in open play and Lacazette scored every 133.9 minutes. Basically as of last season, Giroud was the more effective striker in open play in a smaller sample size. Furthermore in this section, the ESPN writer claims that because Lacazette has a higher total number of goals in the last 3 years, he is the better striker. However, that fundamentally overlooks the difference in function of both Giroud and Lacazette, Giroud being more of a playmaker whereas Lacazette being an out and out striker. Not to mention that goals scored in the Premier League are not equal to those scored in Ligue 1.  Finally, their last argument for why Lacazette is better in this section is that he scores a higher percentage of the team’s goals, which is not terribly meaningful, given that we all remember the years when Henry was scoring 40% of the team’s goals and Arsenal still not winning the league ( Andy Johnson for Crystal Palace is also the same). So to be generous I conclude that out of three things (better striker last season, better striker over 3 years and percentage of team goals) they mention as being the reasons Lacazette is better, I generously concede the second point, so ESPN writer 1/3 correct.

  1. Lacazette is more versatile and can play more positions

No argument on this one, Giroud is definitely an old fashion hold up, playmaking striker, saying he can play two positions may be stretching it. So Lacazette definitely is more flexible, although is versatility necessarily better? Why not ask ESPN writers themselves? One of them wrote a critic of Alex Oxlade-Chamberlain’s positional flexibility (7), so who knows.

  1. Lacazette is a better playmaker

Here the writer mentions that Lacazette is the only player other than Messi that has scored more than 20 goals per season in the last three years and created 35+ scoring chances. My understanding is, Messi is in his own league, and not all goals and chances are equal, if they were, Lacazette will be considered as good as Messi by this metric. Also, the bulk of the argument in this section hinges on passing percentages, I would like to point to Statsbomb’s excellent article about how pointless passing percentages are without proper context (8).

  1. Lacazette and Giroud are both good at converting big chances.

This is just a wash.

  1. Lacazette is a proven penalty taker.

Here it is said that Arsenal have a poor conversion rate from the spot and hence needs Lacazette to help boost Arsenal’s penalty record, since Lacazette has scored 83% of his penalties. I wonder who else has a good penalty record and can help Arsenal? Oh, Giroud!  His conversion rate? 91% over all EPL and Ligue 1 seasons (9).

To wrap up, I would once again like to emphasize that I am not making a judgment on whether Lacazette is a better or worse player than Giroud. However, I believe that I have sufficiently pointed out that the statistics presented by ESPN writers are neither conclusive nor convincing. If they are not writing with malicious intent, then paid writers should do much better than throw out a few figures without properly contextualizing and explaining skin-deep statistics. For a properly analytical and in-depth look into why Lacazette could become a great player see Statsbomb’s analysis (10).

  1. http://www.espnfc.us/club/arsenal/359/blog/post/3148698/giroud-may-start-for-france-but-lacazette-still-better-option-for-arsenal
  2. https://www.transfermarkt.com/alexandre-lacazette/torenachminute/spieler/93720/plus/0?saison=2016&verein=&liga=&wettbewerb=FR1&pos=&trainer_id=
  3. https://www.transfermarkt.com/alexandre-lacazette/elfmetertore/spieler/93720/plus/0?saison_id=2016&wettbewerb_id=FR1
  4. https://www.transfermarkt.co.uk/olivier-giroud/torenachminute/spieler/82442/plus/0?saison=2016&verein=&liga=&wettbewerb=GB1&pos=&trainer_id=
  5. https://www.transfermarkt.co.uk/olivier-giroud/leistungsdaten/spieler/82442
  6. https://www.transfermarkt.com/alexandre-lacazette/leistungsdaten/spieler/93720
  7. http://www.espnfc.com/club/arsenal/359/blog/post/3147636/alex-oxlade-chamberlain-must-nail-arsenal-position-to-cement-future
  8. http://statsbomb.com/2017/06/passing-percentages-are-mostly-useless-quantifying-passing-ability/
  9. https://www.transfermarkt.co.uk/olivier-giroud/elfmetertore/spieler/82442/plus/0?saison_id=ges&wettbewerb_id=
  10. http://statsbomb.com/2016/07/who-is-alexander-lacazette-and-should-arsenal-spend-40m-on-him/

Untold Social

In case we ever have tech problems again, please follow us on Facebook – it’s where we go when things go wrong.   @UntoldArsenalToday

 

22 comments to Intentional misrepresentation, misunderstanding of statistics, or…?

  • Jerry

    Good article! I thought exactly the same when I read that ESPN article!

  • Goonermikey

    There are, of course, other things with the Giroud/Lacazette statistics that are not taken into account. Who had the most goal scoring opportunities per minute for example. What were the calibre of players around them and what were the quality of those chances.

    The other question ‘journalist’ are also completely unqualified to answer is “which player would be the better fit for the team and the system the manager wishes to play at any given time”.

    At the end of the day, whether ESPN has an agenda or not there appears to be fewer decent football ‘journalists’ around than correct transfer predictions made by said ‘journalists’!!

  • WalterBroeckx

    And the question not posed is: is Giroud better at defending set pieces like corners?

  • Shakabula Gooner

    Walter, the same point crossed my mind: that Giroud’s value to us, in defending set pieces may be overlooked.
    In any case, I also believe that the person to compare Lacazette with is actually Sanchez. He is the one that basically displaced Giroud in leading our attacking line. Giroud basically comes in when, in spite of his capabilities, the team needs the Giroud dimension as a possible option for subduing the opponent – after 60-70minutes of relying on Sanchez and co has proven fruitless.
    Thus, even with Lacazette bought, Giroud is still needed. However, if Sanchez isn’t leaving, would Lacazette be required? That is the principal issue but ESPN either deliberately or naivety failed to see the question in this manner.

  • Leon

    ‘And the question not posed is: is Giroud better at defending set pieces like corners?’

    So what’s the answer? I’ve Googled it and can’t get any info. Do you know?

  • Mad Eye

    Good points, well made. But perhaps an easy target.

    There is no doubt a reason why the author on ESPN is writing half baked transfer speculation articles rather than working for StatDNA or a similar outfit.

  • OlegYch

    nice work mate
    this is how journalism should be done

  • Gord

    Nice article.

    I don’t know if ESPN is respectable. I will grant you that ESPN would like people to think it is respectable.

    Approximately zero of the people writing sports articles appreciate the philosophy of statistics (and they want us to respect them?). A big reason why writers want to talk about goal scoring rates, is so that they can sort them numerically. But all of this data needs to be first examined for the likelihood of sets of values being statistically similar.

    If we are counting something, the first choice of a distribution to describe the phenomenon is the Poisson. Which has the property that the variance equals the mean.

    Your (drawn from ESPN?) rates for Giroud and Lacazette are 1 goal from open play for some time more than 1 game and less than 2 games. Well, if the important part is the 1 goal part and the variance equals the mean, we are talking 1 +/- 1 goals which quickly becomes a meaningless exercise as negative goals are not allowed.

    You mention minutes per goal from open play. Why use minutes? Why not hours, or multiples of Pi microseconds? If you want to scale goals over some time period, you want to find a time constant which results in the variance of that rate having the same properties as the Poisson does for counting.

    A person could talk goals per game. But a reason to talk minutes is that not all players, play full games. How do we convert minutes into games? Naively we could assume 90 minutes per game, but that ignores the time wasting the officials allow. If a striker plays an entire game and scores one goal, does it make a difference whether his team had 30% possession or 70%?

    Goals per season may have value with respect to a particular team. You mention Lacazette had 18 goals from open play and Giroud had 12. There is a perfect square near these two numbers (16). If I just blindly assume that both athletes should be scoring 16 goals per season, we see that Lacazette is half a standard deviation above that value and Giroud is one standard deviation below. It is likely that neither of those differences is statistically significant, and consequently what we really have to say is that they both score the same number of goals per season from open play.

    But now the writers in the medja have have nothing to write about.

    So much of sports data is just an accident of fate. Team rank at the end of the season is another example.

    At the moment, the EPL has about 6 top teams, and 14 rest of the pack teams. Three of those 14 get relegated every season. But really, there is (almost) nothing to differentiate the top rest of the pack team from the bottom one. Relegation is a lottery that nobody wants to win for the rest of the pack. There is no reason to assume that the top six behave in a single manner. They could behave as a unit, as 6 independent teams, or something in between.

  • Dammy

    Well written piece.
    I wonder why you haven’t ventured an opinion?

    I am a numbers man by trade and I like statistics. However, I am acutely aware that numbers can be manipulated to support any argument. I like the writing on ESPN to be honest, but I would say they are no more guilty than the remaining journo’s who use stats like “since 2012” or “in this calendar year.”

    I didn’t read the article but I believe I saw something similar reprinted elsewhere and dismissed it almost instantly. Giroud made mostly substitute appearances last season and cannot be directly compared to Lacazette for 2016-17. Perhaps Sanchez should as he played more games as a CF than Giroud? A different matter.

    However you can compare Lacazette to Giroud over 3 or more seasons at which point, the finer points like games played, injuries, minutes, position played become excuses. Aguero is almost always injured for spells every season but delivers the goods. As much as it pains me to say it, Kane missed 10 games last season and still picked up the golden boot.

    Lacazette looks an upgrade on Giroud not because of his age, but because of his goals record, pace, and finishing.

  • Sifarzone

    Hello all, thank you for the response to the article.

    Gord, that is exactly why I did not comment on who is the better striker, my point was that the ESPN writers have not really proven that with their numbers. I do agree picking one minute as the interval is a bit arbitrary, but for people like myself, average consumer, it is a more digestible unit. Also I did wonder a bit because not all minutes are equal (sub minutes vs starter minutes), just as not all games are equal, and again, the ESPN writers have proven nothing. At least with just the stats they presented.

    Dammy, I did not venture an opinion because as a graduate student, I unfortunately do not have the time or expertise to watch enough games or do enough stats to draw any conclusions. I would be the first to admit my limitations as an amatuer. I grant you that Lacazette passes the ‘eye test’ but the specific stats that ESPN have chosen do not prove he is conclusively better.

  • Gord

    Sifarzone

    What are you studying?

    Untold has a git project (or rather, I have a git project under Untold’s name) where I am dumping bit of code (mostly perl_ that I write for looking at statistical issues (github.com/UntoldArsenal/ua1).

  • Samuel Akinsola Adebosin

    Sifazone, what is the point you are trying to make? With this your comparison of Giroud’s stats to that of Lacazette, are you saying there is no need for Arsenal to sign Lacazette because in your own view, we have Giroud whose stats is similar to that of Lacazette?

    Well, I don’t know if that is the point you are trying to make by underscoring the Espnfc statistical comparison of Lava setter’s Lac a setters stats as against that of Giroud whose stats appeared to have been rated lower to that of Lacazette who got a higher stats ranking scores by the Espnfc Commentariat. Enh?

    Nevertheless, I am a believer in Arsenal reinforcing their attacking options in their forward-line this summer with at least one new top quality striker signing. While at the same time, I also believe Arsenal should endeavour by all means to keep Walcott, Giroud and Sanchez who were the main goals scorers for Arsenal for the last three consecutive seasons in all competitions. Therefore, adding another top goals scoring striker of Lacazette’s caliber could significantly improved Arsenal overall goals scoring tally next season to the extent of the Gunners lifting the Premier League title trophy at the end of next season’s campaign.

    For the LB Kolasinac who has come in and the 2 more who are strongly believed will be brought in by Le Prof to make it three in numbers brought in altogether this summer window, definitely, three Gunners in the Arsenal last season’s 25 man first team squad will be offloaded either on loan out or be sold out to other clubs. In this wise, I think Le Prof’s axe may fall on Welbeck(probably sold out), Iwobi(to be loaned out) and one of Jenkinson and Debuchy who are widely expected to be sold out with the sold out axe likely to fall on Jenkinson. Right?

  • Samuel Akinsola Adebosin

    Sifazone, what is the point you are trying to make? With this your comparison of Giroud’s stats to that of Lacazette, are you saying there is no need for Arsenal to sign Lacazette because in your own view, we have Giroud whose stats is similar to that of Lacazette?

    Well, I don’t know if that is the point you are trying to make by underscoring the Espnfc statistical comparison of Lacazette’s stats as against that of Giroud whose stats appeared to have been rated lower to that of Lacazette who got a higher stats ranking scores by the Espnfc Commentariat. Enh?

    Nevertheless, I am a believer in Arsenal reinforcing their attacking options in their forward-line this summer with at least one new top quality striker signing. While at the same time, I also believe Arsenal should endeavour by all means to keep Walcott, Giroud and Sanchez who were the main goals scorers for Arsenal for the last three consecutive seasons in all competitions. Therefore, adding another top goals scoring striker of Lacazette’s caliber could significantly improved Arsenal overall goals scoring tally next season to the extent of the Gunners lifting the Premier League title trophy at the end of next season’s campaign.

    For the LB Kolasinac who has come in and the 2 more who are strongly believed will be brought in by Le Prof to make it three in numbers brought in altogether this summer window, definitely, three Gunners in the Arsenal last season’s 25 man first team squad will be offloaded either on loan out or be sold out to other clubs. In this wise, I think Le Prof’s axe may fall on Welbeck(probably sold out), Iwobi(to be loaned out) and one of Jenkinson and Debuchy who are widely expected to be sold out with the sold out axe likely to fall on Jenkinson. Right?

  • Samuel Akinsola Adebosin

    Master Tony, Sorry for reposting my comment twice. This was as a result of truncating typographical errors which occurred in the paragraph II of my first posting.
    Thanks.

  • Sifarzone

    Gord, I am studying physics/biomed engineering. We do not do a lot of coding, but some. I will definitely be willing to contribute to your project if you do not mind me doing it in my limited spare time.

    Samuel, my point is that paid journalists should do better with their resources and not write articles that MadEye points out are easy targets because they are poorly analyzed statistically. I once again would like to emphasize I am not making judgments on the players, rather placing a judgment on ESPN’s analysis. As to who Arsenal should sign, I do not claim to have that expertise.

  • The_Ledge

    This from Fake Sky on Lacazette:

    His record of 28 league goals last term was more than the combined tally of Arsenal quartet Olivier Giroud, Theo Walcott, Danny Welbeck and Lucas Perez.

    Yes, but none of those FOUR was our top scorer! They might as well said “Arsenal quartet of Cech, Ospina, Cazorla and Bould!”

    What sort of stat is that?

  • Josif

    Excellent piece! 🙂

    I am not the biggest fan of Lacazette, to put it mildly. I understand he’ll become the new Arsenal player as Ornstein has tweeted about it but I fear he might find it difficult to deal with English defenders and the refs. The goalkeepers in France are a level or two below the ones in EPL. He won’t have 10 penalties at Arsenal for a season unless PGMO get Collina as their main man.

  • Gord

    Sifarzone, I think so far, I am the only one contributing. 🙂 If you grok Perl, wonderful. If you prefer other languages, go ahead. We all get busy. I’ve looked at some medical physics type stuff. My background is materials science and engineering, and at one time I was an Adjunct Professor of Pharmacy (Radiopharmacy and Bionucleonics). With that and a couple of dollars, I can usually buy a coffee somewhere.

  • Great piece Sifarzone.

    Believe me I don’t rate Lacazette…not that he is not good but rather he is not the kind of striker that should permanently keep Giroud out of the team. If it is not Mbappe(dreams), Aubemeryang, Lewandowski,Aguero…then let’s keep what we have…which includes Sanchez, Giroud, Welbeck, Perez(remember him?), Walcott and even Campbel.

    We then get a real monster of a DM to partner Xhaka…get another monster of a CB(6’3” tall), then bring Szczesny back.

    Our problem last season wasn’t in the attacking department…rather our defence sent us out of CL. When Wenger decided to play three at the back the team solved the problem but too little too late!

  • Jammy J

    “goals scored in the Premier League are not equal to those scored in Ligue 1” Why’s that?

  • Sifarzone

    Gord, can you tell me more about the project? we mostly use Matlab around here.

    Jammy J, “goals scored in the Premier League are not equal to those scored in Ligue 1” , it is not a hard and fast rule but generally accepted to be true, one way to see this is in the golden shoe race, each goal scorer from each league has their goals multiplied by a weighting factor to adjust for the differences in each league, this weighting is determined by the UEFA league coefficients. Currently the weighting for each goal in EPL is 2 whereas Ligue 1 is 1.5. So you can see that UEFA is telling us that they are not quite equal. This is just a guideline, but it changes as the UEFA coefficients change.

  • omgarsenal

    Tony…..let’s encourage Sifarzone to contribute more often….he/she? Writes well and very coherently with well reasoned and fact supported arguments! We need more of this on UA!

Leave a Reply

You can use these HTML tags

<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>