Fulham 1-5 Arsenal: By the numbers

38

When I saw this match on the calendar I was excited. Fulham are a fun team, they play expansive attacking football and have a dodgy defense. They are like Arsenal, but with players that aren’t as good. It was the perfect recipe for goals.

This match lived up to my expectations in terms of good attacking football and was the perfect send off into an international break.

Fulham 1-5 Arsenal: By the graphs

Fulham 1-5 Arsenal: By the numbers

9 – The number of wins for Arsenal in a row total

6 – The number of wins for Arsenal in a row in the League

1 – The number of spots that Arsenal are above Spurs in the table

4 – The number of spots that Arsenal are above Manchester United in the table

1 – The number of “goal of the season” type goals that Arsenal scored in this match. Holy back heel that goal was amazing from start to finish. [I am not going to lie, I also loved the first goal from Alexandre Lacazette, he is so good at the receiving the ball with his back to goal, turning and shooting with power. That kind of goal is pure talent and I love it.]

15 – The number of days we get to enjoy this match and the great goals scored until the next match against Leicester City.

Today was a great day to watch football. The stats say that this was a close match, but I think that from watching it, it was far more comfortable than what it says. I have not always been happy with the performances that have accompanied the victories that Arsenal have gotten during this streak but today I am quite satisfied and can finally see some green shoots from Arsenal. It wasn’t perfect, the defense was still a bit suspect and it would be great to create more chances but now isn’t the time to dwell on those.

Having true wide players on the pitch is helpful

3 – Shots for Henrikh Mkhitaryan, led Arsenal.

5 – Touches in the box for Henrikh Mkhitaryan, 2nd for Arsenal.

20 – Final 3rd touches for Henrikh Mkhitaryan, tied for most for Arsenal.

0.51Offense Value Added for Henrikh Mkhitaryan, led all players.

4 – Passes completed into the box by Alex Iwobi, led all players.

7 – Touches in the box for Alex Iwobi, led all players.

19 – Final 3rd touches for Alex Iwobi, 3rd most for Arsenal.

0.21Offense Value Added for Alex Iwobi, 2nd for Arsenal.

Henrikh Mkhitaryan and Alex Iwobi were both outstanding for Arsenal in this match. I think it is pretty fair to say that they aren’t at the same level as a Pierre-Emerick Aubameyang or a Mesut Özil but sometimes the ability to fit into the system is more important than pure talent. Today in the kind of 4-4-2 (4-2-2-2?) that Unai Emery Used both were a perfect fit for the wide creative spots to support the two strikers. I’d really like to see both players given an extended run in the team regardless if it is in this system or the more traditional 4-2-3-1 that Emery has used in the earlier League matches.

Arsenal’s hot finishing

78% – The percentage of shots that Arsenal had that were on target against Fulham

47% – The percentage of shots that Arsenal had that were on target this season. Arsenal lead the League in shooting accuracy.

11 – The value of Arsenal’s chance quality this season (xG) for all shots.

14.6 – The value of Arsenal’s chance quality this season (xG) based just on their shots on target.

7 – The amount Arsenal have over performed their chance quality (xG) for all shots this season. [Not including the own goal]

3.4 – The amount Arsenal have over performed their chance quality (xG) based just on their shots on target this season. [Not including the own goal]

I think expected goals has a lot of value, it is one of the reasons that I post it after each match in these breakdowns. That is not to say that it doesn’t have drawbacks or that it is perfect. What Arsenal have done this season with their shooting accuracy helps to explain some of how they have been able to over perform their expected goals (and expected points).

With expected goals, every shot is given a rating for the quality of the chance regardless of what happens after the ball is struck. What happens after the ball is struck is important and I believe that there is a skill in being able to shoot accurately (maybe not THIS accurately for the whole season), currently Arsenal are at the top of the list and are much higher than the long run league average of 30%.

Arsenal are a talented attacking team and could end up higher than League average but I am not sure that they will continue to get almost half of their shots on target for the season. With their attacking talent perhaps they will continue to be better than average at shooting accuracy and continue to do better than the naïve expected goals would expect.

Another alternative that I have considered while watching them this season is that there has been a distinct tactical shift in how they attack low block defenses. Under Arsene Wenger, Arsenal passed teams to death in that dreaded “U” shape before trying to break through with a one-two passing combination or a through ball between the lines. Both of these are picked up pretty well by xG, but under Unai Emery there seems to be more of a focus on attacking the wide areas of the eighteen yard box and then working on cut backs, these shots are very good chances and I believe that they are underrated on the quality of the chances that they produce.

It will be interesting to watch if this is a hot period of finishing that cools off, or maybe a tactical shift that isn’t accounted for in the rating of the chances that Arsenal produce or perhaps even that Arsenal just have players that are much more accurate than average. This is something that will be revisited in this column as the season progresses.

@oh_that_crab

Sources: Whoscored, Opta, My own database

Stats Glossary

38
Leave a Reply

avatar
newest oldest most voted
Red-Sky
Red-Sky

The stats suggest it was an entirely different game. Which is a serious flaw.
It doesn’t reflect this new direction we’re head in. That is we do not need to hold the ball for long or create as many chances to win games.
We’re getting more clinical, much faster in countering the opp, and our pre-planned combinations upfront gives direct routes to the goal.
I never saw Fulham that dangerous or worrisome during the 90 mins. Of course you expect the opp to create chances or gain possession at times, but we minimized the risks to some degree, and for us it always seemed a question of making the right decisions and pass correctly going forward. Once we did that, the game was ours.

Obinna
Obinna

Yeah I noticed that how we attack packed defenses has changed now. Doubling up on the fullback, getting to the byline and the cutback(which is halfway a goal already) such is the quality of most of our cutbacks that it goes in on any contact. We’re no longer tedious in attack. You get the feeling that we can score at anytime if we get the attacking move right. Even when we’re not playing that well. Its no wonder people say our performances don’t match. But I’ve had enough of the opposite. We’re effective these days and that’s what pays off in these day and age. COYG!!

Muigai
Muigai

Regardless of tactics, a team wins games by sheer attitude and self application. That is Man City for you. A driven player will tackle hard, run hard and dribble with purpose. Talent wins you games 5% of time. The rest is down to how hard you want to sweat: the duels you will win, the players you will get past, the dangers you will overcome. Football has become a lot scientific lately, but these are the basics that will never go away.

A Different George
A Different George

I’m sorry, this is mostly nonsense. Given equal quality, or even quality that is relatively equal, attitude and hard work matter a lot. But the hardest-working team in the fourth division, one that “tackles hard, runs hard, and dribble with purpose,” will lose to a full-strength Barca by five goals 95% of the time.

Onenil
Onenil

Arsenal are so good they smashed your statistical model!

red14
red14

One of the few things that exites me is we tend to play much better in the second half. Maybe Emery can analise the opponent’s play in 45 minutes, then adjust the tactics.

One thing is sure, he can give good advices in the break. In todays football the most important thing is tactics, and discipline. We are getting better every match, so im really optimistic. Its a shame we couldn’t get a draw at Stamford Bridge, i think if we played that match yesterday, we could’ve gotten a draw.

Anyways, its really exciting to see a coach with that much of tactical knowledge, I really wanted to see a disciplined side, beacause we do have class players. You never know, if we keep going, we can fight for good things. It seems that Liverpool is losing a bit of momentum, and I think Chelsea are going full throttle now, but they have a CL also, so maybe they’ll be tired for the second part of the season. City is very good again, though.

We are looking like saving some energy, giving just enogh to win our games (except yesterday for sure), then be better with every game. We must keep going, nad maybe fatigue kicks in for other teams.

Very optimistic right now… COYG

historical truth
historical truth

Chelsea don’t have CL, they play thursday football this yeas as well 🙂

Cliff Bastin
Cliff Bastin

That Ramsey goal is like Wilshere vs Norwich but stretched out over the whole pitch.

Glenn
Glenn

I wonder if it’s worth having an xG model that considers where the pass is coming from, and whether the actual statistics (proportion of shots from low crosses/cutbacks that are scored) matches our impression that these are better chances than e.g. through balls to that position.

Scott Willis
Scott Willis

The xG does take into account where the pass and the type of pass. One of the things that is missed tho are cutbacks that are dangerous but don’t turn into shots.

A Different George
A Different George

And, as you have pointed out, it does not take into account the identity of the player. Francis Coquelin had some wonderful chances that, I am quite confident, Thierry Henry would have buried.

Tomislav
Tomislav

When was the last time we had 6 league wins in a row or 9 in all competitions?

Tomislav
Tomislav

didn’t think it was 6 in the same period two seasons ago…

Three Steps Sideways
Three Steps Sideways

The last time we won 6 consecutive league fixtures in one season was at the start of the 2016-17 season between 27th August and 15 October. The fixtures were (I’ll put Arsenals score first, regardless if home or away) –

Watford (A) 3-1
Southampton (H) 2-1
Hull (A) 4-1
Chelsea (H) 3-0
Burnley (A) 1-0
Swansea (H) 3-2

Arsenal also won their last 5 matches that season, and the first match of the following season, so it’s arguable whether you would consider that a continuation of a run of form. Those matches were –

Man Utd (H) 2-0
Southampton (A) 2-0
Stoke (A) 4-1
Sunderland (H) 2-0
Everton (H) 3-1

We then beat Leicester City 4-3 in the 2017/18 opening fixture 🙂

The last time we won 9 on the bounce was exactly the same as the 6 win league streak during the 2016-17. Interspersed between the league fixtures we won 2 Champions League matches and won League cup match.

Nottingham Forest (A) 4-0 (20th September)
Basel (H) 2-0 (28th September)
Ludogorets (H) 6-0 (19th October).

Tomislav
Tomislav

Lets hope we can extend this run 🙂 I definitely have a good feeling about this team at the moment. Thanks.

SweGooner
SweGooner

The first word that comes to mind i exquisite! #coyg

rogerbij
rogerbij

Thanks, I always check this article 😉 I dont think your model is broken as some suggested – the model is based on prior statistical means across many PL games, so it just shows that either a) we have more quality in a given game b) we got statistically lucky.

Your model perfectly shows that Fulham had a high 21 shots but that most of them were outside the box and had low xG value. So this adds up to look like something, but Shots on Target xG shows they were shooting blind most of the time. So we a) were clinical in front of goal, b) perhaps a bit statistically lucky some cracking strikes went in (eg Lacas second) c) their shots were often crap quality and crap execution.

I would love it if our attacking quality allows us to stay above the mean xG across the season, and there is a good chance of that. What Im more concerned is that we are conceding much fewer goals than xG predicts too. Is that correct based on my weekly readings?

I remember the year we won the FA cup against chelsea and Wenger switched to a back 3, and we beat the xG conceeded in the last 10 games and won the cup. But then next season we reverted to mean xG conceeded and came 6th…so it cant last. Man Utd managed to sustain this trick all last season (thanks to de gea and the post) but look at them now…

Some feedback on the visuals – on the pitch, Im a bit confused with the colors – grey shot, pink goal, red? Why not just 2 colours for shot and goal? Also pink for a goal seems an odd choice and my eyes have trouble catching this. Maybe a lighter pitch colour and no need for the outline on the circles? Last, I dont think the ‘expected points’ graph is that useful since it predicts impossible points, like 1.5. You could ditch it imo. Thanks!

Scott Willis
Scott Willis

The colors for the graphs correspond with the teams color, and I picked pink because it is in general a contrast with every teams colors.

Taking out the green might not be a bad idea.

I do the expected points as way to illustrate in general the simulated results as points. In the same way that expected goals have decimals so does expected points.

Elliott Wave
Elliott Wave

I regard your work a lot and i am sure it will only get better as the season goes by. I would like to see a column that shows Arsenal season to date expected ponts compared to opposition season to date expected points.Then i would compare it with the actual figures ( after 8 rounds we have 18 points and the opposition have just 6 )

Scott Willis
Scott Willis

Arsenal vs Manchester City: Actual Points 0, Expected 0.3
Chelsea vs Arsenal: Actual Points 0, Expected 1.5
Arsenal vs West Ham United: Actual Points 3, Expected 1.9
Cardiff City vs Arsenal: Actual Points 3, Expected 1.9
Newcastle United vs Arsenal: Actual Points 3, Expected 2.1
Arsenal vs Everton: Actual Points 3, Expected 0.7
Arsenal vs Watford: Actual Points 3, Expected 0.5
Fulham vs Arsenal: Actual Points 3, Expected 1.2

Total: Arsenal 18 vs 6
Expected: 10.1 vs 10.1

rogerbij
rogerbij

Aha, I see what my brain did… we scored so many goals (pinks) that the red dot we didnt score looked odd… and our low quality chances have a grey outline around them so I didnt see they are actually tiny red dots. (they looked grey like fulhams). Perhaps a case to remove the grey circle outline? Thanks for your comment

Unid
Unid

Another thing that might be helpful about attacking down the flanks is that, unlike through balls which can be easily intercepted and lead to counter-attacks, cutbacks usually just get cleared, often times for a throw-in or up the field but without any real accuracy.

Scott Willis
Scott Willis

or often for a corner or even for an own goal at times. Cutback are awesome.

Dave Crawford
Dave Crawford

Lacazette said on arriving that he won.t shoot if he doesn.t think he can score. He wants to be efficient with his shooting. Maybe this explains the higher than average shots on target stat to some degree. Been quietly impressed with Laca this season. Had my doubts about him at first but he is definitely winning me over..looked a £50M player yesterday

A Different George
A Different George

He’s a much better player than I thought; I was expecting good pace, accurate shooting and some good runs off the ball, based on limited knowledge of Lyon in European competitions and for France. But he is much better on the ball, can link up like a traditional number 9, and is generally a very intelligent player. And an even more clinical finisher than my already high expectations.

TeeCee
TeeCee

I have to say I also love that first Lacazette goal. I’ve seen the Fulham keeper criticised for it, but that’s incorrect.
Laca foxed him by swivelling to his left, at which point you would expect him to shoot right footed to the far side of the goal which is where the keeper, quite correctly, goes. Laca instead pivots further to pound the thing into the bottom near corner, leaving the keeper wrong footed.
I reckon that Laca knew exactly what he was doing there and deliberately went for the more difficult turn and shot that had less chance of being intercepted.

Scott Willis
Scott Willis

Yeah I really liked that goal. It was such great skill to control the ball, turn, feint like the shot was going to the far post and then quickly close the hips to lash it into the near post. It was great.

A Different George
A Different George

If the keeper had done what the pundits are saying he should have, I think Lacazette would have lashed the ball (or maybe just rolled it) into the far corner. I think he made his decision about where to shoot after the keeper committed himself.

Mathew
Mathew

How can SoT xG be higher than xG?

If Arsenal could have expected 2.48 goals just from their shots on target, how does adding the off target shots reduce the expected goals by more than half. It’s like shots off target have negative xG, but that doesn’t sound logical.

I see the glossary claims that “With [SoT xG], shots that end up on target will have a higher value than the corresponding regular xG model”, but I can’t work out why that would be the case. If, per the graph, Ramsey’s goal had a ~10% chance of being scored, why would ignoring a few wayward shots make it more likely to go in?

Everything Matters
Everything Matters

Maybe because if a shot is on target from that range it’s significantly more likely to go in. So if the general all shots xG model includes shots that are off target, a shot from that spot would expect to go in 10 percent of the time= 0.1 xG. But if the shots on target model looks at only the shots on target from that spot, and say 30 percent of shots that are on target from that spot go in= 0.3 SoTxG. I can’t say for sure that’s why, but since the author didn’t answer yet, it seemed like that’s basically why.

Scott Willis
Scott Willis

As Everything Matter states below, in general the SoT xG is about 30% higher compared to regular xG (shots outside the box are closer to 40% because in general they are blocked at a higher rate).

Because you are ignoring the chance that it will be missed or be blocked the value is higher.

Shots off target, because they weren’t within the frame of goal, by definition cannot be scored. So they are assigned a value of zero and aren’t negative.

If you have 3 shots from outside the box the xG for each is around 0.03. Say one is on target and the other two off target you’d have an SoT xG of 0.09 for the shot on target and 0 for the other two. In this situation the xG is the same. But say now you get two on target and the difference is 0.18 vs 0.09 so because you put more shots on target the value of that is higher than regular xG. The reverse works as well, say all miss you have 0.00 vs 0.09 on the xGs.

I hope this helps.

Mathew
Mathew

If I follow correctly, SoT xG is not just (or perhaps at all?) ignoring a shot missing the target in this particular match, but ignoring shots missing the target in the benchmark data. So, if the benchmark for a particular shot type/location has 100 people taking shots, of which 30 were on target and 8 resulted in goals, that would make the xG 0.08 (8/100), but the SoT xG would be 0.27 (8/30).

Is that about right?

Scott Willis
Scott Willis

Yes

Steve Bartlett
Steve Bartlett

An excellent, statistically substantiated article. Thank you.

Scott Willis
Scott Willis

I am happy you liked it

euler_of_the_world

@oh_that_crab Even though the XG model adds value and that this match is an aberration (5 goals were scored from 7 Shots On Target and 9 shots on total whereas on an save rate of 80% it would have been 1.8 goals), one of the reasons why XG seems to be trailing is because a team which is 1-0 down is likely to be open at the back and the same shot conceded will have a higher probability of conversion. Shouldn’t you consider the score at the time of the shot event? Also, expected value models usually fail when modelling low probability events.

So like any expected value, a lot of low probability events are adding together to give you a higher value than it should have happened. I see it all the time when sales teams report a sequence of low probability sales which are unlikely to happen, but considered together will give an impression of a sales pipeline which is not there.

However, I noticed when you give Shot on targets only XG, the accuracy improves dramatically. It doesn’t make sense to have a SOT only XG model, as it is just an extrapolation of common sense.

Scott Willis
Scott Willis

Score effects do definitely effect things but for the most part they are captured in the model. When trailing teams are less selective in there shooting and the reverse is true for teams winning.

I agree on the variability of a low scoring sport, ideally you’d want about 5+ matches to be able to draw conclusions. With a single game you can tell stories but it is really hard to make big conclusions from it.

ArmsAndEther
ArmsAndEther

The “Dreaded U-Shape”… Truly!