October Mailbag

Time to go through and answer some questions from my favorite people my readers

Photo by sue hughes on Unsplash

Time for a mailbag. Thank you so much for the questions.

From Mike: “I would love to know how much xG Jesus generates for himself each game. Like if he gets the ball on the edge of the box he could shoot for a really low xG chance, but he often takes it past 2 or 3 defenders and seems to manufacture a higher xG chance.

Also think this could contribute to his apparent underperformance on xG. He seems to manufacture himself decent chances but perhaps not be in the optimal situation to take the chance - like he's a bit off balance, or has to toe poke, or he's taking the shot on his wrong foot, which I think would penalise him from an xG perspective.”

This is an interesting question and one that goes with the current analysis of Jesus and his finishing.

To answer the question, he has created six shots for himself or a teammate where he completed a dribble first, with 3 of those being for himself.

The second part of the question I think is certainly plausible. There are several shots that stick in my memory where they are in a very good location but he is just off balance or unable to get clean contact. This would match my thought that one of the situations that got him so many good scoring chances with Manchester City was his ability to get on the end of the low, hard crosses across the face of goal. He was good at getting a touch on them but perhaps not the best at getting full clean contact on this type.

Unfortunately, I think the evidence points to a player who is very good all around but probably a 10-20% below-par finisher. That could still mean he is a very good but perhaps just short of an ‘elite’ striker.

From Mide 0: Given how good some of the top 7 teams look, do you foresee a scenario where 75 - 80 points will become the minimum required for a top 4 spot? Will the bar for top 4 get higher as the top 7 get better?

I very much think that we will have a legit ‘Top 7’ pretty soon. Newcastle haven’t exactly gone the Chelsea/Manchester City route with a helicopter drop of cash on all the players where they lapped everyone else, rather they have just become one of the top spenders. So far their spending looks smart and well-targeted. Kieran Trippier was on the older side but his ball progression stats are severely underrated. Bruno Guimarães looks legit and I feel like that is a miss for Arsenal. Sven Botman looks like a good young ball-playing defender. Alexander Isak is still a good striker prospect.

Given that there are now 7 teams who will think that they belong among the top 4 or better, yeah the points expectation for the position will probably generally go up. The only factor that might limit that is how far the teams are able to distance themselves from the rest of the league. With the relatively flat spread of money the ‘other’ teams are still among the richest in the world and they have owners that want to see them compete, even if the ambition isn’t quite the same.

So the competing factor of teams that aren’t easy to beat might lower some of the up for all 7 teams to all meet their expectations.

I will agree it seems less likely that points in the mid 60’s (66 has been enough twice in the last 10 seasons) and more where getting to 75 is needed to be realistically in. For context this season 72 is the average finish for 4th, getting to 75 points would mean top 4 in about 90% of simulations. So we are getting close to that now and I don’t imagine it will get easier over the next few years.

From Cornballer: Is there anything interesting that can be gleaned from radar comparisons or similarity scores looking at this season and last season? Aside from the obvious, have any players’ roles changed significantly?

I love an excuse to break out some graphics!

One of the surprises for me (maybe it shouldn’t because I predicted him to be player of the season) is Gabriel Martinelli.

I did not anticipate him becoming this good of an all-around winger. His uptick in creating for teammates and ability to beat players and then provide a final ball has gotten much better. Overall he seems to keep his head up more often and takes the chances to set up a teammate rather than always hunting for his own shot.

Next is Granit Xhaka who has seen his role change quite a bit from a more deep midfielder to a more advanced midfielder.

His stats are still mostly the same but with more of the final product type. He has roughly the same number of shots but the average quality and distance has improved a lot. This has come at the expense of some ball progression from deeper but it is still impressive who well he has taken to the ‘8’ role. It is still probably the most obvious marginal upgrade in the first eleven but the player that is needed to improve that spot is getting better and better.

The last one I want to highlight here is the other’8’ Martin Odegaard.

His role hasn’t really changed but I think he has added some things that people did want to see from him. His passing is lovely and it remains so, but the big thing that was missing from his game was some actual goal threat. He has added nearly an extra shot per 90 and done so while keeping the shot quality about the same, that has lead to him moving from a 0.15 to a 0.25 xG per 90 player. Over the course of the season that is about 3 goals! A nice improvement from him and hopefully something that continues.

From Austin S: I'm curious why the 538 model was so pessimistic on Arsenal at the beginning of the season. We ended 2021-22 with 69pts (albeit 64.7xpts per understat), and an SPI of 80.2. We began this season on 538 though w an SPI of 78.5, xpts of 61 and xgdiff of +12. What in the 538 model had us doing so much worse than last year's 69pts or even understat's 65 xpts? And how did we start w lower SPI as we really didn't lose much but brought in some valuable players while we have more other players entering their prime than leaving? I know 538 does something to account for player value...not sure if it looks at age progression of important players. Seems like a model input that wouldn't be too difficult to add to improve the R2!

It was bit peculiar that FiveThirtyEight was so down on Arsenal at the start of the year. I too had Arsenal starting fifth (it would have been hard with a backward-looking model to do otherwise) but much closer to Chelsea and Tottenham. Their methodology for the start of a new season looks like this:

They made the choice to use Transfermarkt value which I might be a little iffy on, Transmarkt has a lot of great information but it is not without issues for things like how much time is left on a player’s contract. Personally, I use another iffy method of trying to use wages as my method as I think that has a stronger correlation than transfer value. This isn’t perfect as wage data is notoriously unreliable as well. There isn’t really a great option here to try to look forward for teams that make moves in the transfer market.

From @lhj192: Do Arsenal need to surround the ref like Mourinho’s Chelsea when he is having a disaster class of a game?

I don’t know. I am pretty strongly of the opinion that refereeing is hard and it is actually pretty amazing how well referees are able to do. The speed of the game is crazy and they only get one angle and no replays to try and make decisions. The laws of the game are also not hard and fast leaving different interpretations, plus there are things that come from the league that they want to focus on. I would like to see some tweaks to VAR personally now that we have had it for a few years.

My preference is that we go to a challenge-based system, I think I would start with say two challenges per match. If you get your challenge right you get to keep it. On challenge, the referee goes to the monitor and has a look. The everything gets checked stuff kind of sucks, and it would make coaches think a little more about what is worth challenging and I think overall would lead to the biggest mistakes and most controversial calls getting corrected which is really what we would want anyway.

From @shareeeeezy: Is Monte Carlo method employed for your league projections?

Yes. My league projections take the goal probabilities that are produced from a Poisson distribution and then use random numbers to simulate the goals scored for each team for each match, this is used to generate the results of a simulation. I then repeat this method 10,000 times to get what is presented in my charts. Speaking of simulation…

Week 14 Simulated Odds

This weekend we are back to a full Saturday including four different timeslots for a full morning of soccer for myself.

The early Saturday match is one I am not setting an alarm for.

Leicester has seen a bit of an improvement in their results but they are still big underdogs here. I would love to see them get a result but it’s not especially likely.

the 7am kickoffs are decent. I think the one that will probably get my viewership is Brighton vs Chelsea.

This obviously has the Potter storyline but also features two solidly good teams with question marks about them.

The 9:30 match features the most even odds on paper, it could be interesting. I think at a minimum it is worth having on in the background.

The last match of Saturday will be an interesting test for Liverpool.

They are still big favorites here but Leeds are a sneaky tough team and one where they could cause problems to a team that has defensive weakness like Liverpool. If Liverpool drop points again they could be in real trouble for making the top four.

Sunday the main match I care about is Arsenal.

On paper, this is the second easiest match they will have all year. A perfect opportunity to reset and get right after some performances that have raised questions. This match is pretty pivotal in how the narrative and vibes go for the next couple of weeks.

Lastly is United vs United.

I have to say I remain a little skeptical of Manchester United even as they have clawed their way back up the table against a team that is performing about the same as them so far this season.

🤠 Yee Haw!