Tuesday, August 30, 2011

Shooting Stars

Anyone who has followed my Twitter feed over the past few months knows I am big baseball fan, specifically with regards to the statistical side of the game. I am also a big proponent of trying to use lessons from one sport to help us understand or improve another (the Premier League’s stubbornness in ignoring the success stories of officials, salary caps, transfer rules etc from other sports is a huge source on frustration for me). The statistical awareness in baseball is possibly the best of any sport in the world, with legions of fans (and team employees) dreaming up new ways to measure players and hopefully identify undervalued assets and commodities (the now over, and often incorrectly, used term ‘moneyball’ attempts to convey this thought process).

One term you will quickly come across if you ever decided to visit a statistically leaning site like FanGraphs, would be ‘regression to the mean’. I am no statistician so I apologize to anyone that is if the below is a bit simplistic, but, essentially, this means that over a period of time statistical anomalies will level out and a player will produce at a level in line with his talent and/or past performance. This doesn't mean that his future production will account for any past sub-par performance, but it will at least appear to improve to get back to the level he (or the league) usually performs.

Two things to note here. One, it is important to remember that while regression sounds like a negative term, it can go both ways. If a basketball player has a career free throw shooting percentage of 91% but starts a season 21 for his first 28 shots (75%), we would expect, over a longer period, that all else being equal, the average will regress back to his 91% level and so he should enjoy a period of shooting above his career average as things level off. Second, we are dealing with human beings and not robots so other factors do indeed play a role. A player might be unhappy with his contract, finding it hard to settle in a new rainy city (not looking at anyone in particular Carlos) or playing with an undisclosed injury, all of which will have some impact on his play. However, research done in the excellent Scorecasting  suggests that ‘hot streaks’ are generally statistical quirks remembered due to our own desire to explain statistical fluctuations, rather than the result of a player being ‘locked in’.

If you've made it this far, we can get into the actual meat of the post. Taking all forwards who scored more than 5 goals last year, we see that, on average a player took 6.1 shots per goal with 2.9 of them on target. The correlation between total shots and goals is 89% while shots on target is even better at 91%. Taking shots on target a step further, we see that the standard deviation of the sample is just 0.78, suggesting that the shots/goal ratio holds up across the board. Indeed, of the 34 players to score 5 goals or more, 22 scored a goal every 2-3 shots on target and only three players (Drogba, Rodallega and Lovenkrands) had more than 4 on target per goal (Hernandez was the biggest outlier in the other direction, taking just 1.8 shots on target respectively to score).

So if we accept that a high number of shots on target is going to lead to a lot of goals (it sounds obvious but it isn't 100% correlated) then the idea is that we can look at the shot totals to date and if a player is scoring well below his career average rate, we could imply some regression coming in the future (the same goes for a player scoring every time he shoots).

The season so far
A few players stand out from the early going, though be careful here as we are in seriously small sample size range. This review will be more useful after a few more games, but it's worth taking a quick look now:

NameGoalsTotal ShotsOn TargetOT/GCareer Rate
van Persie 

While Dzeko and Rooney's goal rate looks unsustainable (and of course, Dzeko won't score twice a game all year) if they keep having the number of shots they've managed to date, they are going to score some seriously high totals. 10 shots a piece on target through 3 games is off the charts and not even the trigger happy Ronaldo could match those totals. This suggests that Dzeko deserves as much, or more hype than Aguero given his lower price tag and while Tevez is obviously there as a constant threat, Dzeko is not simply tapping in the good work of others.

You'd like to see Aguero and Suarez up there shots on target per game as they surely can't continue to basically score almost every time they hit the target. This pair are two of the most promising players in the league but this evidence does put at least a small blot on their developing resumes.

Though he hasn't racked up big totals on target yet, Gyan's 10 total shots stand out among those players without a goal and if he continues at that clip he will surely finish with a good haul come year end. Andy Carroll hasn't been well received this year but he too, with 9 shots in a couple of games, could be in line for some adjustment over the coming weeks.

For the future
As I say, this is more something to re-examine in GW8 or so but it is worth looking at nonetheless. The extreme totals produced by Rooney and Dzeko look like they could be legit and Dzeko is very much on my radar alongside Aguero for the coming weeks. We will check back on these totals in a few weeks.

I now have a full set of stats for the new season to work with, so keep checking back throughout the international break as we try and find some clues as to where this crazy season fantasy season is heading next.


chemikills said...

Dzeko and Rooney are also averaging about 4 key passes a game.

A key pass for those that don't know is a pass that lead to a shot, so in essence it is the basis for an assist as a shot is the basis for a goal.

Such a large amount of key passes is certainly reassuring that these 2 will continue making returns (I've never had Rooney in my FPL team because I think he has a poor attitude and it's kind of a personal protest. But I must admit I'm close too caving :P)

Chris I'm hoping we also see some midfield analysis (defenders are much harder and too small a sample atm imo). I'm thinking of doing a similar analysis of key passes and assists from last season :)

konliney said...

Hey Chris,

How much importance do you place on wildcards? I accidentally played mine this gameweek, and feel a pressure to make significant changes to my team even though I was quite satisfied with how things were going in the first place.

chemikills said...

Also I noticed later Chris that you've used shots on target rather than simply shots and averaged that about half are on target, but there's also a lot of difference in how accurate our strikers are over all. Total shots would seem like a better measure to me

Chris Glover said...

chemikills - thanks for the comments. I used shots on targets as the correlation between these and goals is stronger. I will track both (as noted with Gyan) but Iwill lean on where history tells us the correlation is strongest.

Konliney - the wildcard is important but not so much as it used to be now that we have the second one at the halfway point. If you've played it this week I would obviously maximise its use but that shouldn't mean ditching Rooney just because you can. If you post your team, I;m sure everyone will be ready to make some suggestions :)

konliney said...

My Gameweek 3 team was:


My proposed team for the coming gameweeks:

Simpson-De Laet-Enrique-Kompany-Bosingwa

I actually don't have enough money for Sessegnon (missing 0.2m), so I'm contemplating swapping out Bosingwa (bought at 5.7m) for Shawcross (5.0m), which should make Sessegnon possible, whilst creating enough space for a Dzeko to Aguero swap later on.

Any thoughts? Also, who else would you consider as hot prospects for the coming gameweeks based on what you've seen so far, and also what you had in mind pre-season?

P.S Great blog btw. I'm new to the game, and I've found your strategies and analysis (ie GK-combo and Cheapest-Defensive-Link strategy) very helpful.

Craig said...

Very interesting analysis Chris.

Is there any value in looking at whether the correlation differs between shots in vs. out the box?

I also wondered if shots on target included penalty kicks and whether this may influence the analysis in that you would expect penalty takers should have a lower OT/goal ratio.

Eric008 said...

Hey, what do you think about Rosicky from Arsenals new lineup, he shoud have enough minutes and FPL points if Wenger completes Wednesdays transfers? I ask this because I have 5.9m and Rosickys value has already devolvedto 5.9...

My team for next GW:
Schwarzer, Given;
Bosingwa, Smalling, Kompany, Jos Enrique, Ward;
Sessegnon, Taarabt, Nasri, Moses, Rosicky;
Suarez, Rooney, Dzeko.

I think tthis team is very competetive in long term, am going to buy Begovic next gameweek instead Schwarzer because Begovic-Given have the better schedule in nex few GW...
Any thoughts?

FFFan said...

Great blog, gutted that I've only just discovered it. Regression to the mean should also imply that Wayne Rooney's kids won't be as good at football as him - so don't go putting bets on that!

Quick question re fantasy prem: how much attention do you pay to team values? With prices now rising, dropping the wildcard early this week (and making some shrewd transfers based on numbers of transfers in) could allow players to boost their team value by a mil or so - given the long game week. In theory, couldn't the extra half a million you could make this way be incredibly useful later in the season?

(My main worry with this approach being that you only get half of any sell-on profit)

Colin said...

Hi Chris,

I have been having similar thoughts to yourself regarding Dzeko. He seems to be playing the lone frontman role for City, with Aguerro withdrawn, so I would also suggest that Tevez is more of a threat to Aguerro than Dzeko. With this in mind, I have taken the plunge and activated my wildcard this week. My main competition in my league has got off to a flyer and I can't afford to fall any further behind. I know it's early in the season and this may seem rash, but I think sometimes it's important to be decisive. If nothing else I expect my new team to accrue a very decent profit in transfer value in the next few weeks, which will hopefully give me an advantage later on in the season.

Here's my team, any comments before Saturday very welcome.

Given, Vorm

R. Taylor, Warnock, S. Taylor, Enrique, Senderos

Young, Silva, Jarvis, Moses, Cleverley

Rooney, Aguero, Dzeko

I had to drop my exposure to the City defence to get the attacking players in. I'm particularly keen on Dzeko now (the four goals at the weekend were just a confirmation of what I'd been thinking about him.) He's currently leading the line of what is one of the most creative sides in the league. Yes there's rotation threat from Tevez, but I really don't see Mancini benching him with his current form. If that does happen I would then probabley drop him to bring in Defoe as my back-up plan. Regarding the defense, I'd really like to get exposure to the Stoke defense, so contemplating either Begovic for Given or Woodgate for Warnock; haven't really decided yet. I'm pretty hopeful that my shadow midfield players can also contribute. I see Cleverley as a steady Eddie, who will pick up the odd assist in the Utd midfield and at that price seemed like a good deal. Moses and Jarvis always have the potential to come up with something, so there's plenty of potential there, however I expect the front three to be my main points scorers. What do you think?

Gummi said...

@Chris: Again, excellent analysis and food for thought. As @chemikills points out, we also have to factor in assists (and bonus points) in our evaluation.

@Colin: I like the look of your team. Couple of points:
- I would be hesitant about R. Taylor, following Newcastle buying Santon. He could be losing his first team place.
- Senderos was on the bench in the last game, watch out whether this was a rest or rotation.
- I think hard before I hold two attacking players from the same team. You could for example, transfer in Suarez for Aguero and trade R.Taylor or Senderos (perhaps Woodgate or a City defender). However, with the Manchester clubs in such good form, I can see a good argument for holding multiple players in the same club (I have Silva, Nasri, and Dzeko).
- If you have the money, I would consider Anderson vs. Cleverley, as Anderson seems to be playing a little bit further forward.

@konliney: Can't really argue against that formation. In fact, it is scarily similar to mine.

@Eric008: I would dump Rosicky. He has never been prolific. There are more attractive options such as Brunt, Anderson, and Jarvis.

Gummi at fantasyformation.com

konliney said...

@FFFan- I actually tried doing the same thing as you... stocking up on players who I believed would see a price rise soon (might as well as I'd already activated my wildcard). However, I messed up my strategy pretty bad... I dropped quite a few players after they'd only seen a +0.1 rise in price, having forgot we only keep half the profit rounded down..

@Everyone- based on yesterday's deadline transfers, who do you think could turn out to be a real FPL bargain? For example, Crouch at could be the target man for Stoke at 6.9m, arteta will now (presumably) be pulling the strings for arsenal at 8.0m

Exquisite said...

Love your work Chris!!
Keep it up

EddieO said...

Colin: suggest dropping Aguero for Suarez and strengthening that defense. You really need to have three you can play confidently every week.

Eric: there are better options than Rosicky - who dropped in value for a reason. Cleverley, Brunt if WBA come good?

Konliney: I like that side but then its very similar to mine. :) Bosingwa looks better value than Shawcross at the moment but if you feel you need to keep Sessegnon (not one of mine) he's not a bad shout. Bendtner has just signed for Sunderland so this might lower Sessegnon's value?

Good luck to you all!

EddieO said...

Ah, one other idea for Colin and Konliney, I'd consider Nasri over Silva (I have the latter atm). Nasri is a differential, proven point magnet and should start regularly whereas Silva (quality player) has a well recognized propensity to assist the assiter.

mustardking said...

absolutely love the blog, chris. i'm really glad i discovered it just before the beginning of the premiership campaign.

i have to quibble with your explanation of regression to the mean, however. it does not mean that players who are underperforming will have final numbers in line with their talent level - it means that they should hit their natural talent level going forward. so if van persie usually scores once per match but doesn't score at all in the first two matches, it doesn't mean that he is more likely to score twice in upcoming matches. it means that he is likely to average one goal per match in the upcoming matches.

this was explained really well in a fangraphs column on albert pujols earlier this summer: http://www.fangraphs.com/blogs/index.php/regression-and-albert-pujols-slump/

i hope i'm not being pedantic here. i think it's an important distinction.