Mayo Kickouts (pre 2020 Final)

Kickouts are a huge element of any game. Where keepers kick them to, who gains possession, and how, and what teams do with those possessions. Prime example being last year when Dublin outkicked the Kerry press by picking out Brian Howard on the sideline and then setting up McCaffrey’s goal.

The interest in kickouts (as with most things!) tends to peak when Mayo are involved. The overarching sense in the build-up to this final is that David Clarke’s kickouts can be “gotten at”. So in the best traditions of the blog let us preview the kickout battle on Saturday by reviewing what has happened previously between these two teams.

Overview

We have six Championship games in the database between Dublin & Mayo (2015 drawn semi-final and replay, 2016 drawn final and replay, 2017 final and the 2019 semi-final) and whilst the protagonists, be that the Mayo keeper (more on that anon), the managers or the outfield players, have changed the overall outline of the teams, and the game plans therein, have remained intact.

In the six games there were 259 kickouts with Dublin gaining 13 (136 v 123) more possessions*.

*Personally I think language here is important; “winning” is a very positive phrase indicating that teams have done something “right”. But you can “win” a kickout that was terrible (opposition has a 3 on 1 but the one player comes out with the ball). Or indeed by not doing anything at all (ball dribbles out over the sideline). Instead teams gain, or retain, a possession.

Mayo defended those Dublin possessions well – to the extent that despite having 13 more possessions Dublin only produced 3 more shots. Mayo were more efficient at moving their kickout possessions to a shot. But as ever Dublin’s more clinical finishing saw them convert those three extra shots to 2-02 on the scoreboard.

However from a macro view, roughly 0 – 01, and half a shot, extra per game is nowhere near the dominance that is generally attributed to Dublin’s kickout performances.

Dublin’s kickouts

The above table compares Dublin’s kickouts in the six games against Mayo with opponents in other semi-final and finals during the same period. So we are comparing Mayo’s efforts with very similar calibre teams in similar high profile games.

And in truth Mayo have performed well by comparison. They have restricted Dublin to a 79% retention rate whereas other opponents have allowed Dublin an 84% retention rate. Now part of that could be the likes of Tyrone dropping off, and allowing Dublin to have the short ones, but when we look at what Dublin did with those possessions we can see that Mayo restricted Dublin to 0.33 points per possession (ppp) whilst others allowed 0.35. Mayo also scored more, both actually and per possession, than others did on the Dublin kickouts they gained possession of.

Combining all that Mayo come out looking better than Dublin’s other opponents allowing them to net 0.19ppp vs 0.23ppp against other teams.

That is one comparison point – to Mayo’s peers. But how does that 0.19 ppp allowed compare directly to Dublin?

Mayo kickouts

Things are not so rosy. Yes Mayo stack up well when compared to their peers however Dublin are a rung above and it shows here. Mayo restrict Dublin to 0.19 ppp on their kickouts … but Dublin have held Mayo to 0.12ppp on theirs. When they get their hands on the ball Mayo are more or less in line with Dublin scoring 2 – 26, or 0.33 ppp on their own kickouts with Dublin scoring 4-23, or 0.35ppp, on theirs.

The issue, in the main, is the sheer volume of Mayo kickouts that Dublin win. Dublin have only allowed Mayo to retain possession on 28% of Mayo’s kickouts. For comparison Mayo retain possession on 82% of their kickouts against other teams. Dublin retain possession on 79% of their kickouts versus Mayo.

On top of that volume Dublin look to strike hard. When they do get their hands on a Mayo kickout they produce a shot 62% of the time; others produce a shot 53% of the time they “win” a Mayo kickout.

Mayo keepers

One point of distinction within Mayo’s six games is the fact that they have used two keepers. The six games break down as two for David Clarke, three for Rob Hennelly with the sixth split more or less evenly when Clarke replaced Hennelly after 41 minutes of the 2016 final replay

Now not everything is based purely on the keeper. To absolutely, correctly, compare the two keepers we should be overlaying games state, the opposition’s tactics, the positioning of the outfield players etc.

But we don’t have that. Let alone that I only started recording the length of kickouts from 2017 onwards so we can’t, from this vantage, tell if Clarke’s higher retention rate is due to a higher proportion of short kickouts. Instead what we have is as per previous tables – retention rates and scores from kickouts won/loss. And against Dublin Clarke comes out well on top.

2020

So if Clarke has done well against Dublin, despite obvious failings such as the ball over the sideline at the end of the 2017 final, how is he, and Mayo, doing in 2020?

The answer is – not great. In their last three games (against Roscommon, Galway & Tipperary) Mayo have only retained possession on 69% of their own kickouts. We have seen that this was 72% previously vs Dublin … but acknowledge that Dublin is a step ahead of the rest. If the trend were to continue you would expect Mayo to come up with …. ~65% in the final??

And the more we delve into the numbers the less appetising they become. That retention rate of 69% includes short kickouts which account for 42% of all of Mayo’s kickouts. Mayo retained possession on 93% of these, which in itself is poor, but all kickouts out past the 45 saw Mayo with a Retention rate of 53% and a net ppp of -0.13. The opposition combined has scored 1 – 04 from the 18 possession gained off Mayo kickouts out past the 45. Mayo have scored 0 – 03 from 20 similar kickouts “won”.

Looking at the chart for these kickouts out past the 45 (we are missing some whereby the cameras didn’t pan out quick enough to see where they landed) we can see some obvious trends. Mayo have avoided the “mid-mid” section out to about 55m which ostensibly means that teams shouldn’t be able to come straight through on goal from one broken kickout however Clarke’s lack of relative length (think of where Patton or Beggan can land them) means that the whole “mid” section between the 45 & 65 can be flooded and the opposition are able to key on anything that is not pinpoint. Or has any shape of an arc on it.

Mayo have historically done better than others against Dublin on kickouts. Not better than Dublin but better than their peers. Clarke’s kickouts have had much better outcomes than Hennelly’s. But the 2020 trend is not encouraging and whilst in a one off game one or two breaks can have a big effect on numbers Mayo will, collectively, most definitely have to come up with something much better than has been the case so far this year.

Short Kickouts Overview

Short kickouts. The bane of every traditionalist and subject of more opprobrium than steps, refereeing inconsistencies and red card appeals. Yet their prevalence continues to grow increasing from 46.7% of all kickouts in 2017 to 50.0% in 2019. That equates to a quarter of all possessions in a match.

So why, given that opprobrium, do teams persist with them? And do they work?

 

Why go short?

Short kickouts give a team possession (see Note1). And possession has been King of late with the average number of team possessions during a match dropping from 99.4 a game in 2015 to 90.3 in 2019. A reduction of ~9% over the past five years

The points per possession (ppp) by various kickout lengths also show that short kickouts far outstrip the returns for kickouts past the 45 (see Note2)

 

Short kickouts guarantee you possession in a game where possession is rarer; deny the opposition a scoring opportunity by kicking to a contest further out the pitch, whilst also being very productive compared to other kickout types. Why wouldn’t you go short?

Because that narrative is too simplistic. Incredibly so. A high proportion of short kickouts are conceded by opposing teams either through physical restraints (we know that no team can constantly press on the kickout) or for tactical reasons. Therefore basing efficiencies on whether a team gathered possession is too limited. It is what happens post gathering the kickout that determines how effective the routine has been.

(The best case scenario would be to know what percentage of possessions, emanating from short kickouts, were gathered versus differing defensive set ups. If a team defending the kickout drops you would expect the attacking team to take the easy option and roll a short one out every time. But it should be harder to score. If teams press up the proportion of short kickouts will drop as teams look to outkick the press. But we don’t have this granularity – see Note3 (more on what we do have later))

Here we look to introduce a new concept similar to the advantage that a server has in tennis (stay with me here!!). There is an excellent, if slightly out of date, article breaking down the nuances of the server’s advantage in tennis – https://www.tennisabstract.com/blog/2011/08/17/how-long-does-the-servers-advantage-last/ – but to quote that article

“… at some stage in the rally, the server’s advantage has disappeared. Four or five strokes in, the server may still be benefiting from an off-balance return. But by ten strokes, one would assume that the rally is neutral-that the advantage conferred by serving has evaporated”

The same rationale can be applied to possessions in football. At some stage the advantage of having a possession from a kickout, such as the ability to bypass a defensive shield or use a pre-set move, disappears. The type of possession moves from a kickout into what one might dub a transition phase and any scores from this (transition) phase should be removed from kickout returns.

How do we determine that inflection point? That point when a possession moves from “kickout” to “transition”? One way would be time – how long a possession lasts. Another is how many passes are in a possession (see Note4)? And it is the latter that we are going to use here. I have collected the number of individual player possessions, within a team possession, for a number of years now which we can use as a proxy for passes. The breakdown of player possessions per short kickout gathered are shown below

 

Half of all short kickout possessions end after six individual player possessions. How they ended (shot, turnover, end of half etc.) is not a consideration at the moment. That will come later. The tennis paper had a range over which the advantage of the server dissipates. Following that example I have created two kickout phases – “quick” possessions where there have been 1-6 player possessions and then “medium” for 7-11 player possessions. Anything with 12+ player possessions is deemed to have moved to the transition phase – any outcomes, given how long the defending team has had to set up, cannot be attributed to the kickout (see Note5). All scores emanating from short kickouts will be attributed to these three components – quick, medium and transition. It is up to the reader whether they want to put the middle portion – “medium” – into the kickout numbers or the transition phase. I have placed them within the kickout phase.

 

Short kickouts

So what happens on short kickouts? The team taking the kickout gathers possession 94.1% of the time. We will parse those below but for now, and the remainder of this piece, we will treat the 5.9% of short kickouts where possession was lost as one homogenous group. On these the opposition scored 0.67 points per possession (ppp)

That led to a shot

The first demarcation on short kickouts is whether the possession was progressed to a shot. In our sample (96 Championship games from ’17 to ’19 and in excess of 2,200 short kickouts) this happened 58% of the time.

There are five outcomes from any shot; a score, the ball goes wide, the ball drops short and the opposition gains control, the half ends, or the shooting team regains possession (blocked shot regathered, or the ball goes out for a 45). The breakdown, for these five outcomes, are laid out below

 

With a score, a wide, the end of the half or the shot regained the kickout possession has definitively ended. We can take the outcomes from these kickouts and determine our points per possession (ppp) metric. The question then becomes what we do with the 11% of turnovers within “quick” & “medium” (we already have determined that “transition” possessions are no longer considered within kickout returns). Should any scores off these turnovers be netted against the scores from the short kickouts?

My instinctive answer to this is yes – but only where the score is off a quick counter attack i.e. a shot is produced, after a kickout, with one to eleven player possessions (“quick” & “medium”), the shot drops short, or is blocked, and the original defending team now scores off one to six player possessions (“quick”). Why only “quick”? It is undoubtedly a subjective viewpoint but in the main with player possessions of 1-11 the original attacking team should be more or less set defensively. When the ball is turned over the impact on the defensive shape, from the kickout, will be very quickly remedied.

If we do this the ppp table for shots from a short kickout looks like this

 

No shot

To understand the effectiveness of any kickout strategy you have to take account of the scores the opposition managed to attain from the possessions they gathered. We have produced these numbers for (a) where the kickout team lost the kickout and (b) where the kickout team managed a shot from the subsequent possession. What of the instances where the kickout team didn’t manage a shot?

Again there are four main outcomes here; there was no shot because the ball went out for a kickout (without a shot), the half ended or the ball was turned over either inside, or outside, the opposition’s 45

 

Again “transition” possessions, where the kickout team had the ball for 12+ individual player possessions before losing the ball, are discarded. Possessions that ended at half time or where the ball trickled out over the end line without a shot are also discarded.

So the question becomes which of the turnover possessions, and scores from same, we take account of? A bit of subjectivity is required again. Personally I believe all turnovers, off short kickouts, where the ball was picked up outside the defending team’s 45 should still be considered as an extension of the kickout possession. And all scores off those turnovers should be netted against the short kickout outcomes (these are denoted in green above; see Note6).

For those where the turnover was picked up inside the defending teams’ 45 (denoted in orange) we have already set the rules – only those where the then defending team scores on a quick counter are considered. Our ppp outcome for these now looks like

 

We have now calculated all the component parts; a breakdown of kickout percentages, how long teams held onto the ball for after gaining possession, how many possessions were progressed to a shot, how many were turned over, what score came off all our various components.

A team takes 200 kickouts with the returns from same following the outcomes outlined above. What happens?

 

The kickout team ends up scoring 0-57 whilst the defending team scores 0-26 … a net 0.156ppp in the kickout teams’ favour. Still good. Still positive. But much closer to the net ppp in the original table above for mid-range kickouts.

By team

Using the new methodology how do individual teams perform on short kickouts? Unsurprisingly Dublin are way out in front aided by a double positive – they have gathered possession more often than anyone else (98.4% of the time) and are also just more clinical on their shots in general.

 

Tyrone do well here, in the main, as they are very good at getting the short ones away. They are second on the list gathering possession 97.9% of the time.

Kerry and Mayo hurt themselves here somewhat. The opposition has gathered possession 9.1% (Kerry) and 8.1% of the time – compared to 1.6% & 2.1% for Dublin and Tyrone – leading to them leaking a combined 0-15 here. This compares to the gold standard of Dublin; on their 251 short kickouts they have only ever given up 0-02 from directly losing the possession and one of them was in the 68^th minute when they were 9 points ahead against Galway in the 2018 AI semi-final (the other was Donegal in the 2018 Super8 game)

 

 

Note1: when talking about kickouts the term “won” is commonly used – “a team won 50% of its own kickouts”. I use the term myself; I interchange “won” & “gained possession”. But I shouldn’t. We need to move away from viewing kickouts as being won or lost as “won” indicates that the kickout was positive. But a team can win a kickout fortuitously (keeper kicks to a 1-v-3 for the opposition … but they all clatter into each other leaving his one player with the ball. This is a positive outcome on a terrible kickout). We need to think of kickouts in terms of gaining/losing possession as opposed to being won or lost.

Note2: this table is very simplistic. It nets, by length, the points scored on kickouts where possession was gathered versus points scored on kickouts where possession was lost.

Note3; partly due to time constraints on my part. But mostly due to the camera work of the various TV operators. Quite apart from the fact that one camera view is used for the vast majority of the game, so defensive shape cannot be determined, the operators tend to miss a number of kickouts due to the speed with which keepers get the ball away.

But even if we can’t see the exact defensive shape we can make some rudimentary notations as to how the ball was received. That was what I had started to do during the aborted 2020 league campaign. In the 11 games reviewed 72% of short kickouts were received uncontested thus further validating the need to go beyond whether the possession was gained to justify the tactic.

Note4; neither option – time nor number of passes – are fool proof. Any defending team can interrupt the possession by fouling and thus slow the possession down sufficiently to force it into the transition definition even if the time on the ball, or the number of passes, was relatively low.

Note5; these breaks are subjective. There is probably a statistical method available to split these groupings more accurately but my background is not academic enough to perform that. So we will, reluctantly, go with the gut feel based on the chart

Note6; we can tie ourselves in absolute knots here about what cohorts to include/exclude. But the volumes are small – 275 short kickouts ended up with the ball being turned over outside the opposition’s 45. Only 16 of these had 12+ player possessions. For simplicity’s sake they are all treated as one.