Friday 3 April 2015

Liverpool: Formation change to form change

Back when I was still wet behind the ears when it came to all this football analytics malarkey (and perhaps I still am…), about this stage of Brendan Rodgers’ first season in charge I had a look at how some Liverpool’s attacking metrics had changed over the season. After the season had finished I did the same thing from a defensive point of view.

Overall, it was a season that started off pretty insipidly, with poor results on the back of sterile possession, very little creativity, poor shot conversion and an error strewn defence. Then pretty suddenly there was a change, and Liverpool very quickly became one of the best attacking teams in the league and even the defence started to show some solidity. Sound familiar?

I’m sure you all know that Liverpool switched to a 3421 formation in the first of their games against Manchester United this season, although you may not remember that it was Liverpool’s 16th game of the season. Prior to that, Liverpool had a string of poor performances and disappointing results. Whilst Liverpool lost 3 goals to nil against United, and needed a very late equaliser against Arsenal in their next game, the performances were significantly better, and Liverpool went on a 13 game unbeaten run, which included 10 wins, finally losing in their last match, again against Manchester United.


Since those early posts I’ve built on the metrics I used, and have combined attacking and defensive metrics together, so as this has been a season split into two very different halves for Liverpool, I thought it might be interesting to do a similar analysis with the new metrics to see just how much their underlying performance has changed. I’m not the first to do this, Dan Kennett has written a piece over on the Tomkins Times, but I’ll be doing mine in a different way, so will hopefully still be interesting.

Like those early posts, I will look at the metrics on a cumulative basis, and will again show the 6-game moving average so that we can see more clearly how Liverpool’s form has changed over the season. I’ve collected data for four seasons now, and so to add some context, I’ll also show what the average has been for the Champions, the 4th placed teams, and the 18th placed teams.

I’ll start by looking at good old TSR. For those that don’t know a team’s TSR is the proportion of shots that a team has taken in all of its games, it correlates very well with points won and goal difference.




From the graph we can see that in terms of TSR, Liverpool’s season didn’t actually start off too badly, after nine games Liverpool’s TSR was at 0.63, the kind of level that we would expect of a team fighting to get into the Champions League places at the very least. However from that point Liverpool’s 6-game average TSR took a nosedive, and a run of games that included losses to Newcastle, Chelsea and Crystal Palace, as well as a scoreless home draw against Sunderland saw Liverpool’s TSR form go south of 0.5 and was approaching the level we might expect to see from a team fighting against relegation. Two games later, after the matches against Manchester United and Arsenal, Liverpool’s TSR form is back up to where we would expect it to be where it has more or less remained since.

As you can see from the graph above, there is very little difference between the domination of shots on average for the team that finishes 4th and the team that ends up as champions. The thing that separates them is the quality of chances that the champions create and their defensive strength, so next I’m going to look at Liverpool’s CEDE Score over the season so far.

CEDE Score is a metric I put together to measure a teams’ efficiency at both creating good quality chances and restricting their opposition from creating good quality chances. To calculate it, I add a team’s Creative Efficiency to their Defensive Efficiency, a bit like PDO for those of you who know what that is. Now, to calculate those two metrics, I use Opta’s Big Chance stat, which is a chance that one would expect a player to have a good likelihood of scoring from, usually because that player does not face any defensive pressure, such as in a 1-on-1 with the keeper or a free header. Creative Efficiency is the percentage of a team’s shots which are Big Chances, whilst Defensive Efficiency is the percentage of shots against which are not Big Chances. I’ve explored CEDE Scores in a bit more detail here, but it’s probably useful to know that the average CEDE Score for a team is 1.0.



So what do we see? Whilst Liverpool’s season from a shot domination perspective has actually been ok apart from that short term blip, their efficiency at creating and restricting Big Chances at the beginning of the season was nothing short of awful. After 12 games Liverpool’s CEDE score was below 0.9. This compares to the average for the 18th placed team of 0.97, and the lowest over a whole season in my sample was 0.92. After game 12, their 6-game CEDE Score form started increasing, and reached a peak of 1.11 at game 23 before it started dropping back down towards the average. Despite the improvement, over the season as a whole so far Liverpool’s CEDE Score is still only at 0.97.

Although the CEDE Score is good for looking at how efficient teams are at when it comes to Big Chances, it is missing a key ingredient, which is the volume of Big Chances. In my last piece I showed that the Big Chance Ratio (“BCR”) is pretty good at explaining what has happened as it has a higher correlation with points scored than TSR. The BCR is calculated in the same way as TSR but with Big Chances only. As we can see from the graph below, although Liverpool enjoyed decent shot domination in the games at the beginning of the season, it was their CEDE Score which was the driving force behind their BCR, which was equally poor. For the first 10 or so games the BCR was hovering around the 0.4 mark, essentially relegation fighting territory, but when the shots started to dry up the BCR dropped with it down to around the 0.3 mark and a 6-game average low of 0.26. Although it was only 6 games, and I’m sure that there has been plenty of occasions when teams have put together a worse run than that over 6-games, but the worst BCR over a season was 0.3. Then from the Man Utd game we see the rapid increase, and 7 games later the 6 game form BCR had shot up to 0.8 and peaked at 0.88, where over 6 games Liverpool created 13 Big Chances and gave up only 2.



The final metric I’m going to look at is one I also introduced in greater detail in my last piece, and one I call CQR+ (this is an acronym for Chance Quality Ratios Added Together). In the Premier League about half of all goals come from Big Chances, with the other half coming from what I term Normal Chances. A team’s CQR+ is their BCR and their Normal Chance Ratio (“NCR”) added together. CQR+ has a slightly stronger relationship to points than the BCR, however it is a much more repeatable metric, which means that it is a better indication of a team’s actual strength.



So what is the story of Liverpool’s season? For the first 9 or so games, Liverpool’s level of play was essentially about average, they had strong shot domination but weak efficiency rates. Over the next 6 games, Liverpool’s form had dropped to the level that one would expect of a team at the bottom of the league on the back of falling shot numbers and still weak efficiency. Then Liverpool switched to 3421 and 6 games later Liverpool’s form was what you would expect to see from a team trying to win the title, their shot domination had increased back to what you might expect, more importantly, their efficiency also improved. Over the last few games Liverpool’s form has dropped, however over this run of games they’ve played Manchester City, Manchester United, Spurs, and Southampton, so a cooling down of the numbers is perhaps not surprising.

Following the loss to rivals Manchester United, and with only eight games remaining, Liverpool now only have an outside chance of qualifying for the Champions League, and after the United game and the Swansea game before that, some are saying that Liverpool’s 3421 has been found out. This may be so, or perhaps they came up against two teams with the players and the discipline to be able to counter it. Either way, Rodgers has shown that he can get this team performing to a very high level, and that he will change things if needs be, and if he can get Liverpool's performance levels back to where they were just a few games ago, it could take the battle for 4th to the wire, however Liverpool fans may well ebd up looking back at the first half of the season and think what might have been.