Expected Points

Football (and Hurling to an even bigger extent) operates in the universe of small numbers. Dublin & Kerry have played each other in just 25 Championship games in the GAA’s 125+ year history. In a very similar period the Yankees have played the Red Sox 2,130 times!! Comparing era’s and teams in that historical rivalry carries much more weight than the one on this island.

Every year football team’s seasons seem to be defined by one or two very small, possible random, events. In 2015 alone there have been some great examples of this;

Westmeath were 8 points down at half-time in a seemingly one-sided Leinster semi-final. Meath capitulated and Westmeath went on to win by 4. The first time the have ever beaten their near rivals. The following day’s papers were full of headlines like; Tom Cribbin’s tough love pays off for Westmeath, following is very public criticism of his players during the national league. This all despite the media being heavily critical of the ‘rant’ at the time, in hindsight it seemed they were happy to change their mind. By most accounts Westmeath’s season was heralded as a success despite one second half performance and going on to only score a further 13 points in their remaining 2 games.

Fermanagh another team being picked as a ‘weaker’ county punching above it’s weight this year, beat Antrim twice, granted had a very good win against an out-of-sorts Roscommon team and had a moral victory against Dublin.

At the other end of the scale Cork had a terrible year, resulting in the resignation of their manager. Despite pushing Kerry to the pin of their collar in Killarney, Kerry required a lucky inspirational point in the dying minutes to force a replay. Cork couldn’t seem to raise themselves for the replay or the qualifier game against Kildare 7 days later.

Dealing in these tiny sample sizes makes it very difficult to separate the signal from the noise. How much of that Westmeath 2nd half display was down to a managerial rant 3 months previous, or tactical changes at the break? Was Fionn Fitzgerald’s point lucky or calm execution of skill honed on the fields in the Kingdom.?

Expected Points (ExpP)

Shots is perhaps the easiest performance characteristic to attempt to separate the signal from the noise. To do this we can examine what the ‘average’ footballer would expect given the same shot location. This metric is similar to those found in most sports. Soccer has expected goal metrics and Basketball has an Expected field goal metric as just 2 examples.

I have broken the pitch into 5 x 5 square zones. Giving 13 zones long and 18 zones wide (234 zones in total). We then take the average return (Scores / Attempts) for each zone and work out what the expected return for the ‘average’ footballer from that zone would be.

expP-pitch-map

Because of the difference between shots from play and placed ball we need to create 2 set’s of ExpP. Actually we create another one for Penalty’s.

Let’s look at an example; Taking a free in front of the goal and on the 21m line (zone 5;8) gives a 100% expected return for the ‘average’ footballer. We would not expect anybody to miss from here. If, hypothetically, a team took 10 frees from this position we would expect them to score 10 points.

If we look at the same zone but from play we would expect the ‘average’ footballer to score 50% of the time. If our hypothetical team took 10 shots from this location we would expect them to score 5 points.

We can plot the shot location of every shot for each team – we add all the expected returns to see what a team was expected to score based on the chances they created.

In any one game a team will score more or less than the ExpP but this is randomness – the ExpP should give us a more accurate reflection of what the true value of the chances created was.

Fionn Fitzgerald

So to bring us back the the Fionn Fitzgerald shot we can look at the true value of that shot. When I plot that shot it goes into zone 9;13. The average return from here is 27%!!. To put this another way; that shot only just over 1 in 4 times.The reason we like sport is because sometimes we don’t know what will happen and there are no second chances under the same conditions. Cometh the hour cometh the man and all that. The shot sails over the bar, Kerry live to fight another day, their manager becomes a tactical genius again and the show moves on. However Cork can count them somewhat unlucky based on that single event. If you pause the game at that point Cork are the overwhelming favourites every-time.

Donegal v Mayo

Let’s take a look at the model on a team level. If we take the recent game between Mayo and Donegal we can see what each team was likely to score based on the chances they created. Dealing with goals presents a slightly different problem than other sports with a success/failure scoring system. If a shot is taken inside the red zone (above) I am making (a possibly unrealistic) assumption that this was a goal scoring chance. For chances in these 12 zones I multiply the expected return x 3 to more accurately account for goal chances. This is not perfect but it’s a start.

Donegal: The model shows Donegal’s ExpP was 13.3 (they scored 11). That’s pretty close. The reason for the over estimation is probably that Donegal had 2 goal-scoring chances with an ExpP of 3.2 but took neither.

Mayo: The model shows Mayo’s ExpP was 17.3 (they scored 19). Again not very far off. The Lee Keegan goal brings the actual above the ExpP. My model assumes this is a shot for a point (as it falls outside the 12 goal scoring zones) so therefore underestimates the ExpP.

Just to check I looked at the Monaghan v Tyrone game.

Monaghan scored 14 (ExpP = 11.7). Shows just how off they were on the day. Even with a bit of luck they fell short.

Tyrone scored 15 (ExpP = 15.9). As good as Tyrone were on another day they could have scored slightly more.

Final Thoughts

This is only a starting point, in no way am I saying this is perfect. But it’s a hell of a lot better than spouting on again about desire and workrate.

 

By |September 30th, 2015|Uncategorized|0 Comments

It’s a Squad Game

It’s a squad game. Invariably if a team brings on a Sub and he scores a vital point we all hail the strength in depth of a squad. Donegal have too small a squad, the Dubs have an abundance of talent and Kerry’s bench is packed with All Stars. But how do teams use their squad and is there any truth to the cliché that it’s a squad game.

The average Squad size in 2015 is 25. Sure you have some variance. Sligo at 21 and just 8 subs used in their 3 games does seem low. Westmeath & Roscommon top the pile, having used 29 players each in their 4 games. What about the superior 7 counties?

In the first graph you will see that the number of players used is not all that different. Tyrone played 7 games which might explain their high number and Monaghan do look a couple of players short but the others; Cork, Donegal, Dublin, Kerry & Mayo are all very similar.

Squad size 2015

How Are they Used?

Ok so just because a team used X number of players doesn’t mean much on it’s own. The second image shows how many players have started every game and how many have played max game time.

players-used

The stand-out numbers  here is that just 5 Kerry players have started every game. This is half the number for Dublin, Cork, Donegal, Monaghan & Mayo. It really does show how willing Eamonn Fitzmaurice is to change his side from game-to-game (or does he just now know his best 15?). Only Sean Kavanagh played max game time for Tyrone! Tyrone managed to use a lot of different players giving them good game time, starts and still reach a semi-final.

Squad Depth?

What does squad depth really mean? Players used, starts, game-time? Or do we just weave a narrative after the fact?

 

By |September 1st, 2015|Uncategorized|0 Comments

Dublin Kickout Map

There is so much talk about Dublin’s Kickouts I thought it would be worth sharing some data that people could explore. This visualization contains *some Dublin Kickouts from the last 2 seasons.

*TV companies don’t show all kickout locations, especially at the speed Cluxton kicks it out.

By |August 25th, 2015|Uncategorized|0 Comments

The Dominant Dubs – Kickout Analysis

There can be little argument than Stephen Cluxton has changed what it means to be a goal-keeper. His All-Ireland winning free taking aside his skills from the kicking tee, picking out runners with an accuracy previously unseen, have changed the role completely. So how dominant are Dublin at Kickouts?

Dublin Kickouts – Winning %

Dub-Kickouts-Own

  • Since 2011, in 22 Championship games, Dublin have only had 3 games below the national average.
  • Only once have they lost more than they have won – Kerry 2011
  • 69% is there worst performance in the last 2 campaigns.
  • Last season teams at best won 1/4 of Dublin’s Kickouts.

Opposition Kickouts v Dublin

They are many games where Dublin kick the ball out a lot less than their opponents. So how do they fair when facing opposition Kickouts

Dub-Kickouts-Opp

  •  As you can clearly see – teams struggle. Laois winning just 25% of their own kickouts in 2012 is a stand-out.
  • Only 10 out of 22 games have the opposition beaten the average of 60%
  • Some teams have done quite well, noticeably Mayo (2013) & Laois (2014) but those big performances are few and far between.

While their dominance on their own kickouts is something of considerably note – it shouldn’t be underestimated how effective they are at stopping the opposition securing primary possession. Gives an idea of the task facing opposition teams visiting Croke Park.

The Death of Gaelic Football – Fact from Fiction

There is only one way to play football – like it was played in the good old days. When were those good old days – and what were was the game actually like? As has become custom, there is plenty of discussion about the death of football. Hand passing, cynical fouling and blanket defences are the problem. Perhaps amazingly Joe Brolly would like Mickey Harte (and every other manager) to entertain the masses. Yes Inter-County Football is essentially an entertainment product but it’s not Mickey or Jim or Rory’s responsibility to entertain is it?

I’m sure Brolly is not adverse to the limitations of eye-witness testimony and it is worrying that he commands such an audience and bases his thoughts on the current game on his own eye-witness testimony. Tainted by every conceivable bias and very strong beliefs that there is simply only one way to play football, he is dealing in opinion rather than facts, which is his prerogative I suppose.

I wanted to look at the facts. Let’s deal with the death of football and see what that looks like – maybe we can even come up with some ways to fix this dreadful game!

Let’s start at the most basic of points. Scoring is UP

  • Championship 2014 was the highest scoring championship ever.

Ok that cheats a bit – there are more games than ever. But it’s the highest scoring championship since the qualifiers. So how about average scores in Championship games since 1970. It still UP.

average-points-per-game-championship

If football was truly getting too defensive wouldn’t you expect to see scores decreasing? Instead we find ourselves not just with the highest scoring Championship ever but the highest average per game.But maybe scoring is not entertaining?

What about winning margins. No Change!

avg-winning-margin

So we have the highest scoring championship – and again despite the rhetoric the gap between winners and losers is growing. It was the same in 2014 as it was in 1970 and has fluctuated wildly in between 4 – 9 for the last 40+ years.

Not all scores are created equally. Just because scoring is up that tells us nothing about what type of scores. Surely now there is a bigger % being scored from free than from play? My database doesn’t break this down for every game, however we can work with samples(televised games). Looking at games from the last 3 years (league games and at least 500 scores per year) we can see the breakdown of scores.

score-breakdown-league

From this we can see that the breakdown of scores has changed – is it a dramatic change? I will leave that for you to judge.

What about Championship games?

score-breakdown-championship

More scores came from play in 2014 than the previous 3 years. These are very minor changes – but hardly a signal for the death of football?

Other Facts

Is hand passing up? YES

Is foot passing down? YES

Is Shooting up? Yes

Did the number of fouls decrease in 2014? YES

Did ball in play time increase? YES

I have no issue with what direction the game takes – that is up to higher powers than me. You might not like hand passing and what to see more high-fielding and that’s fine. My argument is please be clear what you mean by the death of football. Is it just the style you don’t like? That’s a very subjective opinion to hang the future of a sport on. Any analysis of where the game is at and monitoring its future should at least look at the facts and not just rely on eye-witness testimony.

The death of football looks distinctly like it’s past!

By |March 31st, 2015|Uncategorized|0 Comments

2014 Football Top Scorers

Below is a list of Top Scorers from the 2014 Football Championship. While it is a fun table to look at at the end of the year, the format of the Championship makes the Top scorer an odd measurement of performance. After all you need to be a good forward on a good team to have any chance of making this list.

Top-Scorers-2014

Cillian O’Connor had a great year but he played New York and 2 games went to Extra Time, so his opportunities to score were quite high. A very simple way to evaluate forwards is based on their time on the pitch. The table below simply takes the Minutes Played by each forward and divides by their total score. This gives us a new calculation of Minutes Per Point. So on average how many minutes did each player ‘need’ for each score they got.

top-scorers-mins-per-point

  • Sean Quigley jumps to top of the list with a very impressive Score nearly every 6 mins of play.
  • Cillian O’Connor really did have a phenomenal year. Even removing the New York game leaves him at the #2 spot with a Mins Per Point of 9.35
  • Michael Newman doesn’t make the Top 10 scoring list but appears here at number 5 with a score nearly every 11 mins.
  • Also worth noting that there are only 3 All Stars on this list, perhaps some data might help when they pick the 2015 list!!

It’s not a perfect measurement but does offer some additional context that the raw scoring chart fails to pick up.

*Update – Below is the Table just counting scores from play.

top-scorers-mins-per-point-play

By |November 18th, 2014|Uncategorized|0 Comments

Comparing Football Forwards

It can be hard to rank forwards in any meaningful way. Teams play such a diverse range of opponents and styles that simply looking at Total Scores or Shooting % can be a bit misleading and counter-productive. I wanted to look at a way that would allow for a better comparison of forward play, accounting for both their style and efficiency from play. Copying some of the techniques used by Kirk Goldsberry in his much acclaimed Basketball paper we can come up with some nice and simple techniques to achieve this.

shooting-splitsThe idea is quite simple. Break the ‘scoring zone’ into 3 very distinct areas.

Close Range: 30m arc around the goal

Mid Range: 10m arc around close range

Long Range: Any zone not included in the above.

The image on the right shows what this looks like overlayed on a pitch.

Within each range we can look at two things.

  1. What % of shots does forward X take from each zone. This is called the Frequency Split. 
  2. What is forward X’s Shot % (Scores/Total Shots) in each zone. This is called the Efficiency Split.

The table below shows both the Frequency Splits and Efficiency Splits for 4 marquee forwards.

4shooters-table-efficiency-frequency

To take Bernard Brogan as the example.

  • In terms of shot distribution, Bernard Brogan took 59% of his shots in the Close Range, 31% in the Mid-Range and only 9% from Long Range
  • In terms of shot efficiency, Brogan shot 48% from Close Range, 35% from Mid-Range and 36% from Long Range – ‘surprisingly’ below average (of top 20 shooters) in the Close and Mid-Range zones.

By contrast we can see that Michael Murphy shoots 60% from the Close range but only takes 33% of his shots from that zone. This reflects a very different type of forward. Murphy’s role is much broader than Brogan’s and this is reflected in the Frequency Splits. All other things being equal, if Murphy was getting the same shots as Brogan we would see a 12% higher conversion rate.

Obviously the standout forward in terms of efficiency is James O’Donoghue with 70% from close range and 53% from Mid-Range. The fact that he takes such a large amount of shots within the Close-Range helps a lot and should worry any opposition. I have less data for  O’Donoghue than other players and it will be interesting to see how he gets marked in 2015, but looks like you won’t find him taking many shots from +40 meters.

Cillian O’Connor had a phenomenal year this year and his numbers are impressive in the Close-Range, but push him out a bit further and defenders might see a bit more joy as his return dips below the average.

Conclusion

This method allows us compare styles through the Frequency Splits. This differences in playing style can be dictated by the player themselves or the management but either way it offers us hard of evidence of just how different these 4 forward are. Combining the frequency and Efficiency splits allows us compare different forwards given similar shooting opportunities. Perhaps surprising is that Bernard Brogan is below average in the Close and Mid Ranges. There is obviously more to his game than just shooting, but given the same opportunities the average forward would convert more often. Murphy’s role is also intriguing, no doubt he offers something anywhere he plays but, based on these figures, it’s quite obvious how devastating he is close to goal.

Would be interested to hear your thoughts….

 

* Data is based on 180+ games covering League & Championship from 2011 – 2014
By |November 10th, 2014|Uncategorized|0 Comments

Peak Age of a Footballer

All sports have a peak age but what is the peak age of a Gaelic Footballer?  Ideally you would look at game time per squad for the last 10 years, but that info is simply not available. Maybe one day!

One possible way is to look at the age at which a player was when he received an All Star Award. Now I know the All Star’s have their detractors and results would show that it favours the All Ireland finalists more than it maybe should, however it can be used to see what age the ‘best’ players are.

Gathering the list of All Star Winners since 2001 (start of qualifiers) and matching to their DOB (or at least year of birth) it is possible to see the peak age at which players receive an All Star. As you can see in the graph below. The peak age is 25, with the Average Age being 26.

all-star-winning-age

Odd that the 26 bucket has a few less than 25 & 27, but it’s not a massive sample. What is clear is that there is a big uplift once players reach the age of 23 and starts to drop off after 27. So it’s fair to say that based on this players start to play their best football at 23 and their peak performance will last until 27. There will always be exceptions to the rule but it is interesting to see a clear pattern.

It would be really interesting to see if this Age has changed over the decades? Is the peak age much younger than it was 20 or even 40 years ago? Work for another day.

Average Age of Squads

I wanted to see how the All Ireland Semi-Finalists stacked up against the peak age effect we saw. Do these teams have an average squad age of somewhere between 23-27? Well the Answer was hard to believe.

semi-finalisits-2014-average-squad-age

All 4 squads have exactly the same average squad age of 26 – the average age of an All Star! This is 1 year so it might have been a fluke but… If there are other teams out there with ambitions of challenging for Sam in 2015, it will be worth looking at their squads, will they have enough players in the peak age bracket and more importantly are those players any good?

Averages can hide a lot of variation and there are different ways to measure this. Donegal have the biggest variation in ages, despite the fact that they have an average of 26 the spread of ages is much bigger than in other teams. But what about Kerry’s ‘old’ squad? Actually not so, there spread of ages is very similar to Dublin’s. These figures take into account the entire squad rather than the starting 15.

Position Specific

When we breakdown the figure by position we do begin to see bigger differences (graph below). As you would expect GK’s peak at a later stage than any other position, but still under 30 which might be surprising.

all-star-average-age-position

No surprise that the FB and FF line are the young guns. Speed is probably king here, so age is on their side.

The half-backs and half-forwards peak slightly later than the full-back and full-forwards but there is really not much in the differences between positions.

 

By |September 25th, 2014|Uncategorized|0 Comments

Gaelic Football World Rankings

Most sports around the world have some sort of ranking system. It’s not a new phenomenon but has some real merit in Gaelic Football. The team that lifts Sam has achieved all they need – but how do you asses all the other teams that have competed. New managers come and go but have they improved the team?

The structure of the GAA season does not lend itself to direct comparison from year to year. However if you develop a clear ranking system for teams you can assess how 1 team has done year on year. Good or bad has the team improved?

There is an explanation of how the system works at the bottom of this page, but it is based on a very similar system to FIFA world rankings. Points are awarded for a Win or a Draw with additional weightings being added for the Competition, Opposition Ranking and Historical Weighting.

Here is the 2014 Table as of 18/09/2014 – This will change with 1 game to go!

2014 Table

world-ranking-2014

Historical Table

Want to see how teams have moved up and down over the last few years? Check out this historical table.

historical-world-rankingWinners and Losers

  • 2009 is based on just 1 seasons worth of games – so it is best to concentrate on the later years where more data gives a better picture.
  • It is clear to see the effect Jim McGuinness has had on the Donegal Team. Ranked 19th at the end of 2010 they have yet to finish outside the top 4 under his stewardship.
  • The other big movers are Monaghan. Since Malachy O’Rourke has taken charge they have climbed from amongst the worst ranked teams into the top 5.
  • It might be surprising to see in 2011 for example that Donegal finished 1st and Dublin 2nd but there were more points available for Donegal as they were coming from a much lower base.
  • Armagh look like one of the most inconsistent teams. But 2014 has seen them break into the top 10 for the first time.
  • Derry can feel a little hard done by –Despite 2 successful league campaigns in 2013 and 2014 they have really failed to inspire in the Championship, hence the low ranking.

I’m sure it will create plenty of discussion – no system is perfect but it’s as objective as any system in determining how well a team has done over the years. And remember this is just a bit of fun!

Ranking Explanation

So how does it work?

This is based heavily on the system FIFA use to rank International Football Teams and works like this.

100*(Result x Match Status x Opposition Ranking x Historical Weighting)

Result Explained;

Win = 3

Draw = 0

Loss = 0

Match Status Explained;

Not all games are created equally to we apply a weighting to each game type.

Championship Game = 5

Qualifier = 3

League = 1

This makes a championship game 5 times more important than a league game. Most teams are judged on their championship form so I think that gap is justified.

Opposition Ranking Explained;

At the end of a season teams are ranked 1 – 34 – this ranking forms the basis for the Opposition Ranking in the subsequent year. This means beating a team ranked higher will get you more ranking points than beating a team of lower ranking.

Historical Weighting Explained;

The results below cover games played between 2009 – 2014 (15/9/2014). Each season is weighted so that results in 2014 are worth more than in 2013 and so on.

 

By |September 18th, 2014|Uncategorized|0 Comments

Home Advantage & Regression Towards The Mean

This is just a very quick follow up to yesterdays post. Call it hedging my bets – but this is an important concept. When calculating home advantage in the qualifiers between teams from the same division it turned out that home teams win 81% of the time (16 games). In Gaelic Football terms 16 games is about a quarter of a season so it does seem relatively big but in reality 16 games is tiny. There are just too many other factors (ignoring home advantage) that could be at play in such a small sample size.

So how accurate is the 81% figure. Well it definitely tells us what happened in the past but can it be used to predict the future? The one big thing that tells me this figure wont stay this high is that when we look at all the studies on home-field advantage across any number of sports we almost never get a figure this high. Could home-advantage be that important in GAA – I can’t see why it would be any different to American football, baseball or Basketball. In fact when we look at League home advantage in each of the 4 divisions separately we don’t get close to these numbers.

There is a concept in statistics called regression towards the mean. In any data set you would expect to see fluctuations. Even if you toss a coin it will not always go Head-Tails-Head-Tails etc… You might get a run of heads and then a run of tails. Getting a run of tails after a run of heads is not unexpected, as such, its just that if you toss a coin for long enough you will eventually end up with 50-50 split. A run of heads or tails is likely just the data regressing toward the mean.

So what has that got to do with home-advantage. Well 80% seems unnaturally high, not unlike a run of heads. So without predicting which matches will go which way, or even if we will see the regression start this weekend, I would expect home advantage % to return to a more normal level. So for any punters, don’t just blindly back the home team this weekend.