Last week I conducted a review of rolling data to see what trends have emerged during the early stages of the current season. Many were identified, and I recommend using that blog as a reference point before making any transfers in the near future. Of more immediate relevance to these FPL Poker Tables blogs, however, was something that became abundantly clear during the course of that exercise. Namely, the inaccuracy of the initial weightings given to the promoted sides in terms of their attack and defence strength at home and away.
In only two out of a possible twelve instances were my initial best guess weightings found to have overestimated how much more difficult LEE, WBA and FUL would find life in the Premier League. The two exceptions were the home attack strength of LEE, and away attack strength of FUL, which have both fared better in xG terms than anticipated.
Consequently, my model has had to play catch-up since the start of the season, constantly modifying its evaluation of these teams’ true capabilities. Inevitably, this has had knock on effects on the accuracy of my spreadsheet’s predictions thus far. And this is why I’ve generally not published my spreadsheet’s predictions before the second international break in years gone by.
The subtext here is a plea for your patience whilst waiting for my model to reach equilibrium. In truth, I never expected that to occur before GW16 anyway, because only then will my model be drawing exclusively* on data drawn from this season.
[*Except for 1 home game each from last season for BUR and MCI, and 1 away for AVL and MUN, because of their GW1 blanks.]
Thereafter, the results could be well worth waiting for though, if last season is anything to go by, when I secured seventeen successive green arrow gameweeks in a row, climbing ~4.2 million places in the process!
As stated previously, another complicating factor for my model this season has been the ridiculously high number of penalty kicks being awarded (41 so far compared to 19 during the same period last year), and this is a theme that appears repeatedly this week.
Supporters of SHU were warned last time of a potentially devastating storm by the name of Chelsea headed their way, and so it proved, with the Blues succeeding in living up to their star billing in my last expected goals table where they were predicted to score 3 goals.
This time around, there are no teams predicted to score 3, and only five teams (WHU, LIV, BUR, MCI, CHE) deemed more likely than not to score at least 2 goals in GW9.
Not for the first time this season, BUR sound the dubious prediction klaxon, which is probably due to them ending last season in uncharacteristically attacking fashion once safety from relegation was assured. As I pointed out in my 2019-20 season review of rolling data blog, this came at the expense of their defensive process, and it was always likely that Sean Dyche would revert back to type at the start of a new season. BUR have only played twice at home so far this season, which means three quarters of the data used by my algorithm dates back to the ReStart period when they were playing a much more cavalier brand of football. It will take a few more home games for my model to realign to the new reality, but I expect the downward trend in the Claret’s home attacking strength ratio shown in the graph below to continue.
I have a confession to make about the BIG mistake I made last week when I rounded down my spreadsheet’s prediction of THREE goals for CHE to two, and TWO goals for MUN down to one, in the Correct Score Forecast table last week. I have to hold my hands up here to being too trigger happy with a new concept I was in far too much of a hurry to roll out. Namely, the significance I attached to the xG difference between teams playing against each other. With hindsight, I was likely swayed by a desire to justify rounding down the 2 goal forecast for WBA (vs TOT) that I felt embarrassed about.
I did the same with WBA in GW7 too, and the outcome vs FUL seemed to vindicate the policy. There is a simpler explanation for my model overvaluing the WBA attacking strength though, which is they started the season with an overly generous interpretation of their Championship form last year, and it’s taking time for that flattery to fully filter out. WBA have really struggled away from The Hawthorns and, as the graph below shows, their away attack strength weighting has been in freefall since the season began.
For what it’s worth, there are 3 matches in GW9 with insufficient predicted xG difference between the teams to credit one of them with a goal advantage, but I will not be repeating last week’s mistake, and the 2 goals that CHE, MCI and LIV are predicted to score will be allowed to stand (see table below).
My spreadsheet’s predicted players points table last week had Fernandes (17), KDB (5), Werner (7), James (2), Kane (9) and Chilwell (8) occupying the top 6 places, and averaging 8 points between them, so which players are ranked highest for GW9?
The table below shows how many points players are expected to get on average (excluding bonus points, and yellow/red cards). The red ‘P’s indicate where predicted scores have been inflated by the effects of recent penalty kicks, and the numbers in the column to the right of expected points show what these scores would have been if the distorting effects of penalties are removed. In other words, these are the average expected points assuming there are no opportunities to score from the penalty spot.
Last week’s table was dominated by players from big clubs, whereas this week’s sees greater representation from smaller clubs. I’d have been bullish about Antonio‘s captaincy credentials but for his recent hamstring injury, while Wood‘s elevated position is suspect given the overestimate of Burnley’s home attack strength alluded to earlier.
Sandwiched in between that pair come Golden Boot contenders, Vardy and Salah. The latter’s prominence in the table owes far less to penalty kicks than the former’s, so Mo might have been the one to beat, but for a positive COVID test result putting the spanner in the works, when analysing My Stats Tables in Fantasy Football Scout’s Members Area. Grealish is a new name for us to consider in this section and, judging from my predicted players points table for the next 4 gameweeks (see further down), he might be set to feature regularly here.
The 8 teams with representation this time are WHU, LEI, LIV, BUR, AVL, CRY, TOT, and EVE, and 6 of these are inside the top 8 for taking fewest minutes per xG accrued in my Team Offence table below:
Because of the dramatic increase in penalties this season, however, I checked how it would affect the rankings if teams were sorted by non-penalty xG instead, and the table below shows how doing so changed things. Unsurprisingly, given the preposterous number of penalties they have already benefitted from, LEI are the main casualty of this measure, dropping from third best to eighth.
The only teams not showing up well in the tables above are BUR and CRY, which counts against Wood and Zaha. These two players are set to face each other at Turf Moor in GW9, as were Salah and Vardy at Anfield. The 4 teams our remaining captaincy contenders face are BHA, FUL, MCI, and SHU, so bear them in mind as we turn our attention next to my team defence table.
At first glance, the table above arguably puts Wood back into the picture, but note CRY have conceded 3 penalties (including the outrageous one given against Ward in the 89th minute, which deprived my FPL team of his clean sheet points!). So I checked again how sorting teams by a metric that excluded penalties would affect the rankings, and the result of doing so (see table below) sees Palace go from second worst to a much less damning ninth worst. Either way, SHU remain in the bottom 3, and FUL in the bottom 5, which bodes well for Antonio (if fit) and James Rodriguez, and by extension, Calvert–Lewin.
MUN and LIV drop out in the table above showing the worst 9 teams for non-penalty xG conceded during the last 6 gameweeks, while LEI and AVL come in. The ranking of fourth worst for the last named will come as a shock to Martinez owners.
More pertinently, you will note that the defences of BHA, MCI, and BUR were conspicuous by their absence in either table, and that is because they rank in the top 6 teams sorted by non-penalty xG conceded (first, fifth and sixth best respectively), which is unwelcome news for any would-be captainers of Grealish/Watkins, Zaha, and Kane/Son.
I turn now to my xG and xA tables to see if they can revive/solidify interest in any of the aforementioned players. Usually I sort my Players xG table by Minutes Per xG but again, because of the distorting effects of penalties, I’ve elected to sort by non-penalty xG instead this week, and the captaincy case for Kane, Watkins, Vardy and Salah is strengthened by their inclusion in the 7 player shortlist below. I’d much prefer this metric be ratioed for minutes played, but that option isn’t available currently.
The case for Captain Jack rests more on his creative output than his goal threat, and he is top of a very short shortlist of players meeting the criteria for a place in my expected assists table. Rodriguez strengthens his case for differential captain consideration too.
Honestly, I’m no clearer now about the best captaincy option this week than I was before repeating the processs I’ve found so helpful the previous 3 gameweeks. Players from MCI and CHE were unusually absent from my predicted players points table, and understandably so in the case of the former, but with CHE ranked 5th best for non-penalty xG in my team offence table, and NEW 6th worst for non-penalty xG conceded in my team defence table, a case could certainly be made for either Werner or Ziyech, albeit neither feature yet in my players expected goals or assists tables.
There were 3 defenders in the predicted players points table last week, but only Tarkowski this. Yet another Burnley player whose lofty position ought to be treated with caution though, given the previously discussed probable overvaluation of their team’s home attack strength.
That said, BUR are the only other team alongside WOL deemed more likely to concede zero goals than one (see table below). There were 4 such teams last week, with 3 of them (WHU, LEI and BHA) obliging with clean sheets. Unlike last week, however, there are no teams deemed more likely than not to keep a clean sheet in GW9.
Longer term, my Clean Sheet Probability table (see below) has been signaling for some time the arrival of MCI as the best defence to be invested in from GW10 onwards, and I have been banging the drum for Cancelo since before the season began. Not only was he singled out preseason in my backline bonus magnet blog as the best Sky Blues defender to own, but he been one of the Citizen’s most attacking outlets in all of his appearances so far. Against him is the availability of Zinchenko, and imminent return from injury of Mendy. There’s no question in my mind that the Portuguese is first choice, but rotation remains an ever present risk in the game of ‘Pep Roulette‘.
LEI were singled out last week having moved up into 3rd best in my model’s projections for the next 6 gameweeks, courtesy of 2 good clean sheet prospects in their next 3 games (GW8 & 10). With the first of those clean sheet banked, however, they slip back down a couple of places.
Belatedly, BUR make their first appearance of the season in the table above with 2 good clean sheet prospects in their next 3 games (3 in the next 6), but these only raise them up into 8th place, due to how unlikely the chances are in the other games, but that matters not if we are benching the likes of Taylor for those.
After this gameweek, it will be the turn of ARS to have 2 good clean sheet prospects in their next 3 games.
Looking further ahead with regards to players predicted points, the table below shows the expected top dozen points scorers over the next 4 gameweeks. The attacking triumvirate of Watkins, Grealish and Barkley really catch the eye here with their debut appearances in this table coinciding nicely with a favourable fixture swing for AVL.
That’s all for now, folks. May the GW9 flop be with you.
Coley a.k.a. FPL P0ker PlAyer
Follow @barCOLEYna
@FPL_GFCT 