I haven’t written one of these in a while because I’ve been focused on other work, but there’s plenty to discuss.

How Do We Fix Spurs This Summer?
First, you get a decent manager. Van Gaal is the name that’s been batted about, but there’s no reason for Spurs fans to be despondent if he doesn’t come. There are plenty of other sociopathic geniuses out there to choose from, the trick is finding the right one.

Here are three: Thomas Schaaf, Thomas Tuchel, Markus Weinzierl. Yes, they are all German, but they are also all actually good at this whole managing and tactics business. Uncle Tim is not.

Find Germans disagreeable? Fine. I think the best name I have heard for the Spurs job was floated as a possibility by the Telegraph this morning: Mauricio Pochettino.

Sound defensive system? Check.

Better than expected offense with decent, but not great players? Check.

Used to working with a Director of Football system and ownership that isn’t completely attached to reality? Also check.

Don’t underestimate what Pochettino actually did with Southampton. Before he came along, they were definite candidates for relegation under Nigel Adkins. After he came along, they were nowhere near relegation, and have taken fairly regular scalps from big teams in the Premier League. They are a very comfortable 8th this year, even after wasting £27M on Wanyama and Pablo Osvaldo. His tactical system is a lot like what AVB ran at Spurs, except Southampton players actually score when they shoot.

Beyond the Spurs manager issue, it’s probably best for me to cede the floor to Tor-Kristian Karlsen, as I agree with just about everything he said yesterday on Twitter. (Note: read bottom up on each image.)




The attacking talent is mostly there, and I still think the buys they made last summer are pretty good, but evaluating them is complicated by the fact that guidance from the top has been seriously remiss. On the other hand, there is a problem with the Spurs defense. They need 1-2 center backs this summer and have to completely re-evaluate their fullback situation. Neither Dawson nor Kaboul (who used to be outstanding) have looked capable of playing against anyone in the top half of the table. Meanwhile, Kyle Naughton, who is 25 now, doesn’t belong anywhere near the starting lineup. That leaves the oft-injured Danny Rose (decent) and Kyle Walker (good), plus The Janimal, who might actually rip the still-beating heart out of the next manager that forces him to play left back.

I don’t know where Spurs will get the money to re-tool this summer, but fix the manager issue and add quality and pace to the back line and they will again be good enough to compete for Champions League places. Hell, everything kind of fell apart on them this season and they are 2 points above Manchester United in the table. All is not lost.

Something is Actually Wrong at Arsenal
I’m going to start by assuming you read the piece from this weekend by Ben Pugsley. If not, click over there and come back to me.

I haven’t written much about Arsenal this season because I really did not want to be that downer fan standing on the side saying, “Hey guys… something’s not right here.” I just wanted to enjoy the ride and hope that they continued to over perform.

Writing about Tottenham’s self-destruction? Good fun!

Writing about Liverpool’s analytical revolution and march up the table? Also good fun!

Writing about how Arsenal’s baseline numbers are falling off a cliff this year? It just irritates people.

This kind of needs to be said though.



I can’t…

What the hell is going on here? *Goes to sob in the corner*

Forget Manchester United’s transition from Fergie to Moyes – that is the saddest chart I have ever seen. [Disclaimer: I’m an Arsenal fan.]


Yes, it fails to account for shot quality for or against, but basic shots  and that differential captures the broad strokes. Going from a 7.2 shot differential per game – a number that would generally compete for the top of the league – to a 1.5 over the course of five seasons is staggering. Even more staggering is that Arsenal have done it under the same manager (and are still about 80% likely to finish 4th).

The first reason most Arsenal fans would point to is injuries. The problem with this theory is that Arsenal always have a lot of injuries. In fact, they are almost always near or at the top of the injury table in terms of player games missed due to injury, and this happens every season and is not just because Abou Diaby is broken.

Because Arsenal always lead the injury table, I’m lead to think there is definitely a problem with regard to their fitness regime. Either they are training wrong, have the wrong expectations in terms of how much training and game time most players can endure before they are likely to pick up injuries, or they have too few players for their typical game load (in short: the squad needs to be bigger). I am not an expert in sports science and I don’t know how things operate specifically at Arsenal, but the fact that this problem is constant means they need to do something different going forward or the problem will always exist.

Another major issue with Arsenal this season is quite simple: They have no zoom. Pace. Meep-meep.

With Theo Walcott out, Arsenal have zero players in the attack that actually scare opposing defenders, either when running on to throughballs (an Ozil specialty) or running at defenders off the dribble. Compare that to City (Aguero, Navas, Silva) or Liverpool (Suarez, Sturridge, Sterling, Coutinho), and the differences become stark.

Lack of pace means tighter marking to the players up front. It also means less space to exploit between the back line of the defense and the midfield bank of four, which makes it harder to Arsenal’s intricate passing game to operate, and means you have less space for all of those talented attacking midfielders to do their thing. It also means slower recovery speed from attack to defense, and it creates a whole host of secondary issues that aren’t entirely obvious.

I actually saw the esteemed Arseblog and Orbinho discussing the quality of Arsenal’s crossing this morning, and Giroud’s ability to get on the end of them and my jaw dropped. Given how Arsenal have played in years past under Wenger, this whole crossing business should be a tertiary concern. However, because the entire squad lacks pace going forward, it suddenly matters. That’s definitely not part of the plan.

So there’s no pace in the squad, the talent level in attack is a huge question mark, and multiple important players are constantly injured and have been for years.

Things are very much not right at Arsenal, but they used to be. Money is no longer a problem, but maybe the correct perspective to diagnose and treat the issues is.

We’ll know more next year.

Failed Stats Experiments
The theory that tackles and interceptions have no statistical correlation to goals against and the like bugs me. It’s correct – there is no correlation from either – but defense is also tremendously important. It likely means that we’re measuring the wrong stuff from games, or we haven’t figured out how to adjust the stuff that we are measuring in a way that is important. I can also tell you from my player research that having higher levels of Int and Tack in your stat sheet seems fairly significant, but at the team level? Meh.

With this in mind, I started thinking about how to “fix” the base stats collected. If a team has the ball, they can’t make defensive actions. Therefore possession is a limiting factor on potential defensive actions! A team like Bayern, who has had more than twice as much possession than their opponents, has far fewer opportunities to even perform defensive actions unless they go rogue and start stealing the ball from their teammates. A breakthrough!

Or so I thought. I did the regression analysis on Tackles plus Interceptions per game at the team level and compared it against various metrics it should impact like Shots Conceded, Goals Against, etc and found…

Normalizing by possession ends up giving defensive actions the same correlation as possession itself.

Which probably makes sense, but I was hoping for more.

It’s entirely possible I screwed things up or missed a different way to attack this problem, so with that in mind, I have included the data set at the end of this article for all of you to play with. In the meantime, I’ll keep working on this, because I am sure defense matters… I just don’t know how to prove it yet.


[table id=62 /]
  • adc

    Isn’t the whole Arsenal thing down to a score effect? Arsenal *this* season have (or did have) the third highest PDO of the league. So are they not simply scoring early (by chance) and hence playing a lot of the time of their matches in the lead? Which is accounting for fewer shots and less possession?

    • tknutso

      If that’s the case, why aren’t City and Liverpool showing similar issues? And why the 5-year decline?

      • adc

        No idea. But, and this is wild speculation, it could be that City and Liverpool, for some reason, do not seem to be massively letting up once going in the lead, hence their rather large scoring totals this season. It’s a completely different topic, but I’ve said in the past that defending a lead (or more rather attempting to tighten the game), if you are demonstrably the better team, is a questionable strategy.

      • gpm

        My opinion is that the drop off in the various stats are primarily related to the personnel available (or not) for Arsenal.

        Based on who was fit at the start of the year (no Walcott, Ox or Podolski), Arsenal’s game was intricate passing first and then crossing. These strategies were necessary because of Giroud’s inability to create his own shooting opportunities – opposed to everyone on Liverpool’s front 3 and Aguero. This situation has become painfully apparent over the last few weeks whereas when Walcott was healthy there was an uptick in the shooting statistics week-over-week.

        I’d also add that losing players – first Cesc, then Nasri, RVP, even club-footed Gervinho – took away a major dimension of Arsenal’s offense. They lost the man who could create shooting opportunities for everyone and then 3 players who were able to create for themselves. Glaring need for some more direct players in the squad.

  • toshack


    Interesting as ever and good inputs on both Spurs and Arsenal.
    On defense I return ot Xabi Alonso who said that tackling is the last resort, when everything else has failed. Good defense, to me, is the ability to “divert” an attacker to either loose speed, to enter less dangerous areas or to make a less dangerous pass (compared to what the attacker would do if non-challenged). The ability to make attacking passes go to less dangerous areas or to stop passes go to dangerously placed players (i.e. force passes to either go to the defender or to players that have defense cover or to force a bad cross that can be dealt with easily by the defenders or the keeper). This has to do with team structure, tactical set-up, geometry, mutual coverage etc. And how do you measure that (individually)?

    • tactistic

      Not sure if Alonso was really referring to those tackles addresses in the article. I think he talked about the “sliding tackle”, which has a high risk of resulting in a bad foul, whereas the term here is general expression for dispossessing an oppononent. Have a look at this. http://www.whoscored.com/Glossary

      I agree to you that much improvement should be achieved in defensive stats. The bulk of the common stats only respect ball events, but defence is based on movement and positioning OFF the ball. Maybe it is worth looking at the “coverage shadow” (=the space where the opponent’s player on the ball can’ pass because of the defender’s positioning; don’t know how to translate it from German). But it is very difficult to ascertain it (you’ll definitely need a software), and I don’t know how to use it.

      • toshack

        I guess we’ll never know exactly what Alonso meant. My interpretation is that he was comparing several “defensive tools” and found that tackling was amongst the least useful. The full quote in the book I read is: “I can’t get into my head that the football development would educate tackling as a quality, something to learn, to teach, a characteristic of your play.” “How can that be a way of seeing the game? I just don’t understand football in those terms. Tackling is a [last] resort and you will need it, but it isn’t a quality to aspire to, a definition.” The book concludes that to Alonso tackling happens when something goes wrong, not right. I tend to agree.

  • Vignesh

    On validating defensive performances…
    I haven’t done the data work to support my hypothesis. Typing out loud, so to speak…

    Why restrict defensive actions to tackles and interceptions?
    How about also including “defensive actions” that are not attempts to “forcibly” regain the ball.
    Clearances made also seem an important defensive action – they stunt an opposition attack (either temporarily or forcing a turnover). From memory, a lot of clearances if they weren’t made (especially headed ones) would have resulted in a shot conceded.

    Defensive stats are currently defined in terms of positive defensive actions made. However, maybe we can also attribute defensive performance to “failed” attacking actions by the opposition. While they may not make sense at an individual player level, at a team level they might shed light on the defensive style / performance of a team.
    Eg: unsuccessful passes (or FT passes) made by opposition could work out as loose proxy for defensive pressure. Failed Dribble attempts by opponents could also be another proxy for defensive pressure.

    Would love to hear your thoughts on this.

    Insightful writing as always.

    • tknutso

      A quick note on the defensive stuff. I’m aware of the other concepts and angles of attack and will likely explore some of them this summer. I covered it just to add info about what I’ve done and suggest that some things don’t always pan out in a useful way.

      It’s possible I should have just dumped that in its own post 😀

    • tknutso

      Speaking of that, we’re working toward examining “Opposition passing percentage” in general and in the Final 3rd, but have not finished that work yet.

    • tactistic

      I’ve also thought about using the opponent’s offensive stats as a indicator for a team’s defensive achievement. I can imagine two problems:
      1) There might be even more random variation compared to “positive” actions”. I can’t prove it, but i think so because the opponent’s strength and their random variation has a g´bigger impact. But if statsbomb is going to do some research we could know more soon…

      2) It is very difficult to measure an individual player’s impact on it. I can imagine something similar to a Plus-Minus-Rating, but that might be imprecise. Perhaps that isn’t necessary because there are reliable individual player statistics (would be interesting to read something about it) or because it should be used only as a team value. Though, it is always worth to know how/if a player influences the stat.

      Meanwhile, it’s a pity that there exists nothing comparable to (second) assists for goals/shots conceded. If somebody invents it, it could be put into a relation to clearances (=Goals/shots avertes vs. goals/shots caused).

  • plechazunga

    re: Spurs manager

    If not van Gaal, how about Bielsa? Or maybe Jorge Sampaoli after the World Cup? I would love to see him coaching in a top European league.
    Or maybe even Spalletti, which would be a bit ironic, but still, he is quality. (Or Spalletti to ManUtd! 🙂 )

    Also, I know this ain’t a jukebox, but it would be great if Statsbomb wrote an article on the Top 5 most underrated managers of the past few years. 🙂

    (What is the most important stat evaluating a manager anyway? Win%? Points/Match? But that’s too contextual.)

  • Mitch

    Longshot prospective manager for Spurs: Roger Schmidt of the Salzburg Red Bulls.

    He’s done some fascinating things with his high press, even going tit-for-tat with Bayern Munich in a friendly earlier this year, despite a giant talent disparity. Also pretty close to winning his league handily.

  • Derek

    How can tackles and interceptions be important stats for players and unimportant stats for teams? That makes no sense to me. The other way around kind of makes sense as it seems impossible to credit individual players with credit for good team defense by counting tackles and ints.

  • Errorr

    I would also like to add that Arsenal may also overrate their own ability mitigate injuries of off injured players.

    I am also starting to suspect that the European development system overworks many young players that ultimately never develop because they are doomed to recurrent injuries. It may also be why players peak so early compared to other sports.

  • Tuiuan

    I think one way of trying to discover what makes a good defense is to pick up stats of historical great defenses or historical good defensive managers(Mourinho, Rafa Benítez, for example) and see what makes then great.

    Maybe this is approached like the NBA. You have some stuff that you can’t allow your opponent to do(shoot corner 3, allow shots inside the arc, foul a lot) but the way you do it can vary(you can decide if you want to press the ball-handler on pick-and-roll or not, if you want to send very aggressive help or not, etc)

  • Anthony

    Minutes per turn over (minutes not in possession / [tackle + interception]) correlates fairly well with shots conceded (coefficient of .53)
    Not sure if there are other forms of turnover stats that could be incorporated to get a better correlation (incomplete passes (would that equal interceptions and balls hit out of play?)).

    Maybe also having time in possession in oppositions half would give a greater correlation, on the basis defensive units could concede possession outside of their half as a positive defensive outcome (shelling).

    Shots from poor locations are a positive outcome for a defense and could be included in the turnover count to further refine it and hopefully improve correlation.

  • Stuart Ingham

    I think what is missing from defensive statistics is what tennis(inaptly) calls ‘unforced errors.’

    I say inaptly because ‘unforced errors’ in tennis refers to the number of times the opponent hits it out or into the net. However, these errors are very obviously forced. When the opponent feels the need to try more adventerous shots, or are slightly stretched when playing their shots, they are far more likely to be innacurate. This is why tennis commentators often talk about the unforced error stat as something that both players can influence. They are aware than on average a Andy Murray or Novak Djokovic make their opponents produce more unforced errors than other players and this is part of their skill set.

    This is important for football because good defending often doesn’t result in a tackle or an interception. It instead results in the opposition taking logn range shots or attempting passes that go straight out of play. (Bayern were takign an unusual amount of pot shots yesterday which is a sign of how well Man U were defending). Without adding these ‘unforced errors’ to the more noticable dispossesions, you completely lose a sense of true defensive performance.

    • toshack

      Good comment Stuart,
      You are putting into words what I was aiming at in my earlier comment, with a clear example, thus better :-).
      A difficulty here is that we want to measure a lot of “actions” that all ends up in something not happening (i.e. not conceding a goal), whereas in attacking we want to measure a lot of “actions” ending up in something that does happen (i.e. goal scored).
      The link between a through ball and a goal is realatively easy to measure. But the link between an well positioned DM and a poor attacking pass that results in no goal (“unforced” error) is so much more difficult to measure.

  • Pingback: Replica shirts? Grow up… – Le Grove – The Arsenal opinion blog()

  • Woolwich Peripatetic

    The Arsenal stats could be a maturing of playing style, concede more shots from easily defended positions, take fewer shots from more dangerous positions.
    Just a thought. If I were the manager of a football team I would love to see my forwards shooting from high probability areas (and not trying their luck from distance) while conversely I would want my opponents reduced to trying hit-and-hopes.

Improve Performance and Productivity in Your Club:
State-of-the-art Football Analytics