Spring review of rolling data

The current International Break has offered me the first chance I’ve had since November to take fresh stock of my spreadsheet’s rolling data, and see what interesting trends have emerged in the meantime.

First off, a quick reminder to you that I ran parallel models last season to test which number of gameweeks provided my spreadsheet predictions with the strongest correlation to actual scores, and I established that an 8 gameweek data range was optimal.  Ever since then, my spreadsheet’s predictions have been based on each team’s last 8 home or away games (whichever is relevant) for the upcoming gameweek.

Every team is given a weighting for attack and defence strength, based on their last 8 home and last 8 away games. The resulting ratios represent how many xG teams are expected to score or concede against a defence or attack with a rating of 1.00 (average). They are updated after each gameweek, and by charting the fluctuations in these ratios we can observe the directions in which teams are trending.

The best to worst 8 game sequences are coloured on a scale of blue to red, making it easy to see where in the season each team’s best and worst xG form was leading into gameweeks 1 to 30. For example, the most noticeable trends in the home attack strength table below are the resurgent ones achieved by LEI and MUN, and the downward ones experienced by LIV and CHE.

Appendix I

Unsurprisingly, given their well documented fall from grace, the most dramatic decline seen by any team in any category is the home soil attacking form of LIV, which is now below average for the first time in a long time.

average xG expected to score vs an average away defence

Worryingly, for those heavily invested in Bamford, Raphina, Dallas et al, the rate of decline in home attacking form for LEE is actually even more pronounced, due to their rating peaking a couple of gameweeks later than LIV. Their impressive looking improvement in the first half of the season should be disregarded as this likely only shows that my initial best guess weighting for a promoted side underestimated them.

average xG expected to score vs an average away defence

The alarming drop off in both home and away attacking form for AVL since the injury absence of Grealish is captured by my rolling data, with the talisman’s last home appearance coinciding with Villa’s peak in GW23. Watkins owners ought to be anxious for the club captain to return sooner rather than later.

average xG expected to score vs an average away defence

Trending in the opposite direction meanwhile are LEI MUN and WHU (since GW20, GW17 and GW22 respectively) who all went into the International Break at the height of their home attacking powers.

average xG expected to score vs an average away defence

Evidently, the defences of teams visiting King Power Stadium, Old Trafford, and London Stadium respectively can expect to be thoroughly tested by these hosts, who are currently ranked 1st, 2nd and 4th respectively for home attack strength.

The home form of LEI has been something of a revelation, considering they have tended in recent seasons to be much more effective on the road. I wondered if the Foxes have had a particularly favourable sequence of home fixtures lately, but this is seemingly not the case as their last 8 were against EVE, MUN, SOU, CHE, LEE, LIV, ARS, SHU.

Kudos to Ole Solskjaer too for seemingly turning around MUN‘s home form, which was really quite poor indeed in the first half of the season. WHU on the other hand have been relatively consistent throughout the season.

Turning our attention next to the home defence strength table, the three teams that have noticeably cut down the quality of goalscoring chances they concede to visiting teams are BHA ARS and NEW.

Appendix II

BHA have been a very frustrating team to follow this season from an xG perspective. They have failed to score in games where they have had chances equating to 3+xG, and managed to concede in games they’ve restricted opposing teams to very few scoring opportunities in.

Even so, their transformation from one of the most porous teams last season to now ranking as the second best home defence in the league, sandwiched between CHE and MCI, is nothing short of astounding.

average xG expected to be conceded vs an average away attack

Not quite so impressive, but arguably more surprising, NEW currently rank as the 6th best home defence in the league, having shown steady improvement throughout the season (see chart below), which maybe should temper expectations somewhat with regards to likely popular GW30 captaincy selections in the form of Kane, Son and Bale.

ARS have been a lot more erratic in their home defence form, but they have improved enough lately to occupy a season high rank of 7th best.

average xG expected to be conceded vs an average away attack

Thus far, we have only focused upon teams’ home form, so let us examine the performances of teams on their travels, starting with attacking strength.

Appendix III

The trends that stand out most in the away attack strength table above are the role reversals recorded by LIV, LEI and MUN. The upward trajectories achieved by ARS and WOL are significant, and also worthy of closer examination.

Arguably the only case that can be made for not selling Salah for those who still own him is the fact that LIV remain an attacking force to be reckoned with away from Anfield. In fact, as the chart below shows, their attacking process away from home has actually gradually improved throughout the season. Only the champions elect (MCI) carry more goal threat on current form.

average xG expected to score vs an average home defence

Earlier on this season, it was well understood that LEI and MUN were counter-attacking teams served better by playing away rather than at home. Well, for whatever reason, managerial design maybe or random variance, this pattern has well and truly flip-flopped. Compare the chart below with the one shown in the home attack strength section, and you will see the home and away weightings trending in opposite directions.

average xG expected to score vs an average home defence

The drop from the season best weightings achieved by MUN and LEI are among the three steepest experienced by any team in this department, albeit they are still ranked 8th and 9th best on current form.

ARS on the other hand have been rapidly improving, and are now ranked a season high third best in this category, which contributes to them being predicted by my spreadsheets to be the fourth highest scoring team over the next 6 gameweeks.

average xG expected to score vs an average home defence

WOL are currently predicted to be the sixth highest scorers over the next 6 gameweeks and, on the cusp of a favourable fixture swing, it is encouraging for would-be investors to see the recent turnaround in underlying attacking stats. The fact they approach the appealing upcoming fixture run in better away attacking form than at any point this season bodes well for owners of Neto.

average xG expected to score vs an average defence

For the purposes of the next team under the microscope I will show the same away attack strength table again, but this time with the teams colour shaded relative to all the other teams in the league.

Appendix VII

Having already referenced the improved home defence strength of NEW as a cause for concern for those planning to captain a Spurs player in GW30, it should be pointed out here that the case for doing so is weakened further by the lowly 16th place ranking TOT occupy in the away attack strength table above. Only SHU, NEW, CRY and BUR are reckoned to pose less goal threat on the road than Mourinho’s men, and it’s not as though the trendline below offers much in the way of encouragement either.

average xG expected to score vs an average home defence

And finally, here is the table for away defence strength, and the teams drawing my attention the most, for good reason and bad, are LIV, CHE, AVL and WOL.

Appendix IV

For those contemplating investing in WOL defenders ahead of their aforementioned favourable fixture run, the picture is nowhere near as promising as it is for their attackers. It appears at though the upturn in Wolves’ attacking process has come at the expense of their defensive process.

Excluding the promoted sides, the declines in home and away defence performance are the worst and second-worst respectively. Consequently, having started the season rated as one of the best defences, they are now ranked only 12th best at home and 11th best away.

average xG expected to be conceded vs an average attack

Apologies to the legion of you who own Martinez (41.5%) and Targett (15.1%), but only one non-promoted side has suffered a worse decline in away defence strength weighting than WOL, and that is their West Midlands neighbours.

The ‘V’ shape below is indicative of striking improvements having been made in the first half of the season, followed by a no less dramatic reversion to expectation levels slightly lower than they were pre-season.

average xG expected to be conceded vs an average home attack

On current trends, the top 3 away defences in the league are MCI, LIV and CHE.

average xG expected to be conceded vs an average home attack

Whereas MCI have been consistently good throughout the season, the appointment of Tuchel has contributed towards eradicating the disparity that used to exist between the performances of the CHE defence at home versus those away from Stamford Bridge. They are the most improved away defence in the league according to this review of rolling data.

Surprisingly, given they were last season’s runaway champions, it is LIV who are the next most improved in this department, and memories are revived of the preseason hype for Alexander-Arnold and Robertson. Now might not be the worst time to be owning either of this formerly much-vaunted duo.

Changing the lens through which we looked at the away defence strength table at the start of this section to one whereby the teams are colour shaded relative to all others, it can be seen that the next best teams in this category are LEI TOT WHU ARS MUN and BHA.

Appendix VIII

Once again, I have found this to be a useful exercise to undertake, with some genuinely surprising findings. Admittedly, this retrospective analysis of team performance is not by itself predictive of future results. To that end, these tables and charts should be cross-referenced with my model’s predictions for future gameweeks, which take into account the relative strength of opposing teams to be faced, so keep an eye out for my poker table screenshots later in the week, when they will become my latest pinned tweet.

This season has been a long old slog, but we are entering the home stretch now, so best of luck to you all on the run-in. Wishing you all a green arrow fuelled late charge to the finishing line.

Coley (aka FPL P0ker PlAyer @barCOLEYna)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s