xELO: BGW29

by @FantasianPL

We have almost a full week of lead time and only 4 fixtures this weekend, so it seems as good a week as any to make my (triumphant?) return. For those who are new, or maybe just forget what this is all about, a refresher.

I made a model that takes xG results and uses an elo based system to adjust team ratings and make predictions for the upcoming games. I still call this all an “experiment” as I’m not certain how good it is, but it isn’t way off base. If you would like a more detailed explanation of the origin, check out: https://whitebeard.top/xelo-intro/ .

On to the predictions.

The first thing you may notice is that Brighton is top of the list, for both expected xG and xGC. Leeds are second for attack, but West Ham are second last, putting Arsenal as the second best defensive matchup. The one thing to note about the model (and most models) is that they can be slow to react. This is largely to reduce the potential for team ratings to fluctuate wildly based on single results. It is more likely that teams will get the occasional “odd” result than they will significantly change strength gameweek to gameweek. To get an idea of how certain teams may be trending (i.e. their team rating may not be caught up with their current “ability”) let’s look at the season and recent results.

A quick description about the image above. It shows the predicted xG per game the team would score on the season (bright orange) and in the last 5 gameweeks (dark orange). The actual xG per game is shown in blue. The longer the bar, the more xG was predicted/realized. Short bars indicate low attack ratings and/or tough opponents. 

The attacking results can be grouped as follows:

Hot on the season

Leeds have a reputation of a team that will go for it against anyone and based on their results, that seems to be true. Even at the start of the season I expected that their attack rating was probably lower than it should be, but it seems like it still has not reached where it should be. The one thing to note is that while Leeds has been outperforming predictions all season, there are some poor games mixed in with the great games.

Brighton are the talk of the town of late and especially this gameweek. It even seems justified based on the model and their results, they have been and are showing good form, outperforming their predictions on the season and last five. It is worth noting that they have consistently been better at home from an attacking perspective this season. 

Hot on the season, not as hot now:

West Ham fit this category with nearly spot on results in the last 5 weeks, while still being ahead of prediction on the season. They have seen an increase in attack rating, but this has plateaued in the last 10 gameweeks or so. It may indicate that their rating has stabilized and what you see is what you may get.

Aston Villa are also in this category, but slightly more extreme (in the bad way). Looking at their results, there is a stark change after gameweek 20. In the first 18 games of the season, only their gameweek 2 fixture was well below the predicted xG, the next 17 were at, above, or significantly above the projection. Since gameweek 21, one game has exceeded xG, two have been right at the precticion, and six have been below. (note: Grealish played until gameweek 24). This trend is a bit worrying and maybe we should consider mentally reducing their predicted xG.

About what you would expect:

In the last 10 weeks or so, Tottenham have been exceeding the prediction away from home and under the prediction on the road (both with some exceptions). This may not bode well for their trip to Aston Villa. 

Arsenal on the other hand have a mixed bag of results, but have many more games where they exceed the expectations in the second half of the season, even if there has been some inconsistency. It may be a case of which team shows up on the weekend.

Newcastle are similar to Arsenal in that their past results are a mixture of doubling, meeting, or halving their expectations. I wouldn’t be able to identify any pattern to it so, again, it may come down to which team shows up.

About what you would expect on the season, but hot now:

Maybe it is the impending relegation or something else, but people seem to be excited about how Fulham have been playing. While there isn’t a clear turn around in the results, a couple good and a couple great fixtures in the last five gameweeks has been enough for them to overperform their predictions. They seem to have done enough to put them back in our consideration.

In short, based on the recent results Newcastle and Aston Villa seem to be struggling the most, with the other six worth keeping an eye on.

This image is similar to the above (swap red for orange and green for blue), but showing xG conceded per game. In this one, a short bar is good. Also, everyone is performing better than their predictions. Yes not all consistent amounts or durations, but no stand out “bad” defences. Some very quick notes (roughly “best” to “worst”):

Brighton – Decent away, better at home

Arsenal – Better on the season than expected. No big underpermances the second half of the season

Tottenham – Could have been second, consistently slightly better than predicted as of late.

Fulham – Good as of late (not good before), but more often better on the road recently

West Ham – Fairly consistent on the season, maybe better more often on the road

Aston Villa – Generally good with the odd bad game. Good over the season.

Leeds – A bit surprised when I saw their recent results, but not great on the season

Newcastle – Not much better than expected. Low expectations, especially on the road.

So what do you do with this information? Well, first take a grain of salt and remember these are predictions that a) can be wrong and b) results can be odd regardless of the underlying stats. BUT, Brighton seem the obvious team to target with the best xG and lowest xGC predicted.

The Leeds attack seems to be a good option any week, potentially even against a “rejuvenated” Fulham.

Aston Villa vs Tottenham and West Ham vs Arsenal seem a bit of a mystery. A 2-2 or a 3-0 either way wouldn’t surprise me. Tottenham and Arsenal have the only premium attackers this week, which makes a punt on a couple of them reasonable on a Free Hit or otherwise. I wouldn’t ignore the popular Aston Villa and West Ham players, though.

Fulham look decent enough, but most importantly their players are cheap. Easy to fit in a Free Hit team or else bury on your bench for GW30 and onwards

Newcastle might play the spoilers this week, but largely seem like more of a superstitious pick than logical one. 

I hope this has helped and I hope to see you all for the next gameweek. Best of luck with your plans!

@FantasianPL