Ian Watches Soccer with a Computer - A Weekly Goals Added Roundup
/by Ian L
Viz by Mike Imburgio
Soccer, when viewed through a lens pulled back wide enough is basically an exercise in futility. There are so many ways that players are prevented from accomplishing their most basic directive of kicking the ball over that goal line. Some of these are personal failings. Some are systemic. Some are manufactured. Some, of course, are just your no frills varietal of lousy luck.
Some say we love the sport in spite of this. Others insist we love the sport BECAUSE of this. And indeed, the overwhelming unlikeliness of it all is what makes it such compelling viewing.
The computer next to me, of course, has no thoughts on this subject. It cares not a whit for the “why” behind any piece of event data it computes. It is not moved by any particular narrative. It can probably do a good job telling you which team is likely to emerge victorious in any given matchup we feed it, but it does not care one way or the other.
As such, the computer picks out some really surprising things to highlight. I am constantly looking at this printout, pulling the video, and saying to nobody in particular: “Really? THAT?”
So, that’s kind of what I’m trying to do here with this whole thing. If soccer is in fact the beautiful game, it stands to reason that even if you’re just a computer dispassionately watching soccer as a series of coordinates moving a sphere around a grid, maybe we can find the beauty in what you’re trying to show me.
Maybe not though.
Anyway, let’s go on a little journey, you and I, and this computer too. Let’s see what moments of soccer genius it has conjured for us this week.
Daniel Rios - Receiving - + 0.49
I sure didn’t have “Charlotte pastes Philadelphia with the shield on the line” anywhere near the top of my pile of expected outcomes, but MLS is gonna, well, you know the rest. We catch this play already in progress, as Kamil Jozwiak breaks through Philadelphia’s midfield and is rewarded with a ton of space down the left flank. Jose Martinez relentlessly pursues and the presence of one of the league’s best ball winning midfielders has Jozwiak increasingly coming to the realization that his possession of the ball is not long for this world. He quickly moves to rid himself of it and plays a very poor pass towards Daniel Rios. I get the basic idea, but even if it’s played ideally, it’s hard to see a positive outcome here for Charlotte. Perhaps a better idea would have been to cut the ball back to Swiderski who is moving into some intriguing space, but that’s not what happens and it doesn’t matter in the end.
Jack Eliott moves to cut out the pass successfully, but Rios also isn’t going to make that easy and forces Eliott to try a one touch clearance. That clearance winds up not being a clearance of any sort, and the ball winds up with Swiderski anyway because soccer is ultimately an exercise in futility and that’s just how it goes sometimes.
Swiderski deftly dances through a challenge from Glesnes and then plays another poor pass into space for Jozwiak who does very well to catch up to it and keep the play alive. If, like me, you’re wondering where Philadelphia right back Nathan Harriel is, the answer is “not ideally positioned to deal with this”. Instead Glesnes has to go out and close down Jozwiak, who plays a flat simple ball across the six yard box that doesn’t look threatening until it very much is. Rios makes an extremely good run here, and the computer loves it more than the shot that happens after (+0.073), and the pass to him (+0.061). I’m going to give the computer some props here. I wouldn’t have thought much of the run if it didn’t point it out to me.
So, let’s all go back and just watch Rios. You know, REALLY pay attention to the run and movement here.
Rios is just trying to find any kind of space to exploit here at the start, and it’s difficult because Philly’s marking is solid. Rios calls for the ball, perhaps a bit prematurely, but you can see that he’s trying to collect the ball in stride and carry it between the two center backs. It’s an ambitious ask that probably doesn't even work if the pass is perfect, which we’ll never know because the pass is far from perfect. You could probably excuse Rios for giving up at this point, but he stays with the play and his pressure on Eliott forces the center back to deal with it quickly and the scuffed clearance winds up back with Charlotte. Rios pushes Eliott a bit off balance and then shields the ball from him allowing Swiderski to skip past him without issue. Now, Charlotte is in a very dangerous position. Rios is making a great run into that newly vacated by Jack Elliot space, and Bronico (though currently JUST offside and ultimately furious he didn’t receive the ball from Swiderski) is completely unmarked as well. Swiderski’s return pass to Jozwiak is poor enough that the timings of both runs are now moot.
Rios doesn’t shut off for even a second though. He curls his run back to shade to the far post to be on the keeper’s back side. Wagner has moved to patrol that zone, but he doesn’t seem to notice Rios until Rios lifts up his hand to call for the pass. Wagner catches that out of the corner of his eye and turns his head to assess the situation but by the time he looks back towards the pass, it’s too late. Rios darts right in front of him and scores the opener.
All in all, it’s hardly a beautiful goal. The passing is bad, the defense is confused, and a better word to describe it would be “chaotic.” The only guy who seems to be on top of everything is Rios, and because sometimes soccer is fair and just and good, he is rewarded with a goal and the admiration of both me and the computer.
But that’s what’s interesting about this computer. All of that garbage I talked about in the last few paragraphs ascribing intention and adding color to something as unremarkable as a well-timed run, the computer cannot see any of that. Even if it could, it would not care. From the computer’s point of view, there is no difference between Rios doing all that clever running and positioning and just walking there casually. As far as this computer is concerned. Rios just teleports there. All the computer knows is that Daniel Rios being exactly where he was when he was was the most important thing to happen in that play, and even though I think so for different (arguably dumber) reasons, we are in complete agreement.
2. Daniel Rios - Receiving - + 0.47
Ok computer, we get it, you love Daniel Rios. This was obviously a big night for young Daniel Rios. He’s already secured a hat trick by the time we start this play up, and this goal makes it four on the night. What do you call four goals in one game? Double brace? No, that’s terrible. We’ll workshop it.
Ok, so remember how we really got into all the minutiae in Daniel Rios’ mazy run in the last clip? We’re not going to do that here, because this is just bad defending by a beaten team in stoppage time. Nothing wrong with the run to be clear, it’s very good. Just straightforward. This is a classic “the computer” goal. The computer loves this stuff.
Anyway. Nathan Byrne, McKinze Gaines, and Alciver work the ball well. Alciver attempts a 1-2 with Nuno Santos, but Santos doesn’t play the return ball, so it was more of a “1” really, but hey, it all worked out in the end. My favorite part of this play is the nice little ball Santos played over the top to Gaines here. The computer was “meh” on that one (+0.0034 g+), and that’s because the computer likes possessions in dangerous central positions to stay there. The computer doesn’t understand that this was the best option given everything available to Alciver.
Except… Well, the computer can’t see this, but yeah I bet it would have loved Alciver to go ahead and pass to Rios there, and while that’s a tougher pass to be sure, and probably somewhat of a senseless risk with a three goal lead in injury time… ok computer fine.
All that being said, it’s a solid run from Rios, who manages to find space and the ball in a high leverage situation. That’s really all you can ask for from your strikers. So while this is somewhat forgettable from an aesthetic angle, it’s a very well-worked, professional goal from a player having the night of his life.
3. Tristan Blackmon - Interrupting - +0.39 g+
And now for something completely different. This is the first clearance I’ve seen pop up near the top of one of these lists, so I’m sure it’s a world changing clearance. And…kind of?
The danger in this play creeps up on Vancouver and pounces like a house cat stalking a fly. Fagundez opts to take the corner short to Rigoni, and it looks like your standard short corner. Rigoni returns it to Fagundez who should be applauded for avoiding the third rail of short corners (the corner taker not moving onside and getting caught offside receiving the return pass. FM players know what I’m talking about). Not only is Fagundez onside, but he has a lot of space, and uses it to play what appears at first to be a very strange and non threatening ball. Cascante rises up to meet it, and somehow plays it dangerously back across the face of goal towards an onrushing Fagun….wait a second… Was this a designed play?
Huh. Yeah, you watch Fagundez here continuing his run to that exact spot, and I think this one comes from the training ground. I’m not sure if Fagundez was meant to shoot here, or if this weird touch he does towards Driussi was the plan but it very nearly works out. That is until the computer’s star of this sequence, Triston Blackmon, shows up to hoof the ball away from goal.
Comparing Major League Soccer to top European Leagues is a topic I find tiresome and disingenuous, so my apologies for indulging in it, and I’ll do my best to keep it brief. It’s not always the craziest top 2% of plays that is the big difference between leagues of varying qualities. It's the stuff like this. Clearances that are actually clearances. Clearances that actually relieve pressure instead of just sort of redirecting it. The computer is unusually impressed with Blackmon’s clearance because he took the ball from an incredibly dangerous spot on the field and moved it somewhere not very dangerous at all. The computer doesn’t even take into account that he also took it off Driussi’s foot. On Driussi’s foot and 3 yards away from goal is one of the last places in Major League Soccer you want a ball to be if you’re not Austin FC.
It’s a solid piece of defending too. Reviewing a bunch of these plays over the last few weeks, I’ve seen so many defenders just caught looking the wrong way at the wrong time. This play moves quickly and changes angles a number of times, but Blackmon manages to keep his eye on the ball without losing track of Driussi. It would be so easy to mishit this clearance or deflect it into your own goal, we see it happen every week, but Blackmon’s reflexes are quick, his strike is true, and he manages to get the ball high and far, forcing Austin to completely reset their attack.
I’m not sure I can go as far as the computer does here and claim this was the third best play of the week, but I do appreciate it showing it to me. It’s pretty cool to see the difference a seemingly pedestrian play like that can make.
4. Lucas Cavallini - Receiving - +0.31 g+
Ok. So, I’m going to be real with ya’ll. This is where me and the computer really disagree on something, and I’m going to do my best to explain why the computer is seeing and appraising things the way it is, but it’s wrong here. I cannot sugar-coat it.
Driussi misses a shot wide and he can’t believe it (it was a terrible shot, and his shots have not been terrible at all this year so his surprise is not unfairly earned). While he and his teammates are cursing their luck, Vancouver restarts very quickly and they are off to the races.
The speed at which this entire play moves is spectacular, and the computer can see this speed. The computer loves speed. Cubas finds Cavallini who sprays a lovely ball outside to a completely unmarked Luis Martins. Now check out Gauld who is already at full tilt running into space. Hard to pinpoint exactly when he clocks Cavallini’s run, but he clearly catches it out of the corner of his eye. He doesn’t even have to look up and lays an absolutely gorgeously timed and weighted ball right across the face of goal, and Cavallini is right on the spot to sweep it home. A beautiful team counter attack with plenty of credit to go around.
So uh…why is Cavallini receiving all of it? The computer has, frankly, really biffed this one. That bit is just my opinion of course. As a viewer of soccer, all of the credit has to go to Gauld here. That lung-busting run, dragging a center back out wide with him. That perfectly hit ball, and here’s how the split went:
Gauld Pass: 0.0328742
Cavallini Reception: 0.314489
Even if you add in the reception value for Gauld (0.00842251), he’s getting like 1/10th of the credit for this exchange. Computer? COMPUTER?
Note: If you don’t care about any of this and just want to see more good soccer plays, skip to the next one.
In Goals Added, every pass is given value and that value is split between the passer and the receiver. How that pass is split depends on a number of factors, but in this case it feels like Gauld is a bit hard done by it.
However, he’s not actually being punished here, and in fact, one of the reasons that he’s taking such a small share of the goals added value on this play is to protect players like Ryan Gauld, who try a lot of high risk/high reward passes. Yes he only got a small amount of the credit, but if that pass finds nobody that assigned value would be considered negative goals added. By giving him less goals added burden in the split, trying and missing a bunch of passes like this won’t tank his goals-added value beyond repair. High usage risk/reward guys like Gauld have lots more opportunities to collect g+. Gauld gets more touches, more dribbles, more passes, and more receptions than most forwards around the league (though, unhelpfully for this example Cavallini is actually one of the highest-usage forwards in the league).
So that’s one side of the low score. It’s balancing and math and all of that and it can and probably will be tweaked at some point as more data is collected.
Here’s the part where I have to be mad at the computer. The computer just didn't really care that much for that pass. Why? Because that pass very rarely gets to its target. Let’s remember here that the computer cannot see the shape of the defense, the way Gauld dragged Cascante out wide leaving more space for Cavallini. It doesn’t see that it was the best possible pass you could have asked for at that exact time and exact situation. The computer looks at where the ball was kicked from, where it was kicked to, and says, that pass has a 9% chance of being successful, because only 9% of passes from there to there have been successful in the past.
The fact is, the computer isn’t able to see or know that its usual calculations don’t actually work here. This is one of those rare times where I am smarter than the computer. And yes, I’m not too proud to admit that I trash-talked it for being so wrong and dumb.
The other important thing to remember is that this is an infinitesimally small event in the grand scope of all the soccer event data that is collected by not just us but many other parties. So yes if we look at just this one exact instance, analytics has somewhat failed us, but if Ryan Gauld and others keep making that pass successfully over time it will become more valuable as it won’t seem like such a low percentage effort. Goals Added, and I’d say analytics at large is a very big picture kind of enterprise. Individual plays that buck the general probabilistic trends make the computer look stupid, but if you take the whole of soccer data we have into account, the computer is actually right. That’s a very low percentage pass and it won’t work out nine times out of ten.
So let’s score this one for the humans. I’m sorry Ryan Gauld, that run and pass were dope and I award you 20 Ian goals added.
5. William Agada - Receiving - +0.29
Oh good, the moment that crushed my favorite team's playoff hopes into dust. I’m so delighted I get to look at this 100 times and describe it in excruciating detail for you.
Ok here goes. Let’s start with Yeimar. Yeimar has had a problem dawdling on the ball this entire season. It is a strange turn of events from someone who was so sure-footed and dominant last year, but here we are again. He has a few choices here. He can try a pass out wide to Nouhou. This is a risky venture, because it’s a difficult pass that his body shape would transparently telegraph, and also because Nouhou would receive the ball, which is very rarely a good thing for Seattle. His two closest options are Leyva and Atencio, but it’s already too late for either of those passes to be safe, as Agada has positioned himself well to discourage both of those passing lanes. You can’t see Arreaga in the video but he’s just off screen between the edge of the box and the touchline behind Nouhou. That’s another tough one because he has to get it past Agada and also Russell. His best option is to play the ball out wide to Alex Roldan. You can’t see him until the pass if halfway there, but he’s in a lot of space on that flank when the play begins. Instead, Yeimar opts for something outside of the box and passes to Remi Walter, who at the time of this writing was not and never has been employed by the Seattle Sounders.
Lodeiro was presumably the target here, but I’m not sure what Yeimar was envisioning was going to work out. No Seattle Sounders player is well positioned to deal with any of this sudden danger. Walter sprays it out wide to Salloi. Agada’s positions himself right behind Yeimar’s shoulder. Yeimar can’t follow Agada because, although Alex Roldan tried his best, he just wasn’t able to get back to try and prevent Salloi’s cross. Yeimar loses track of Agada who just casually ghosts into the available space, and it’s already 1-0 it just hasn’t happened quite yet. Salloi’s pass to Agada actually isn’t great. It’s a bit behind him, and it’s at a somewhat uncomfortable height and velocity that makes bringing it down way more complicated than it needs to be. Agada is up to the task though, bringing it down with his left, and then slotting it past Frei.
Since joining the league, Agada has put up more open play xG than any other player. He wasn’t a player that impressed me when I gave his resume a look-over, but clearly he has the ability to be a dominant forward in this league, and he demonstrated it by ruining my summer. Thanks William. Well done.
Alright. I’m tired and so is the computer. It has soccered so very hard today. We had some laughs, some fun, and some disagreements. I am now putting it in sleep mode. If androids do indeed dream of electric sheep, I have to wonder what this computer dreams of?
Daniel Rios apparently.