Alternative Premier League: How the Table Looks After Gameweek 1

With one game postponed and more draws than a furniture shop, the opening week of Premier League football has perhaps been something of an anti-climax, unless you were a travelling Trotter at Loftus Road on Saturday.

But never fear – there is still order to be found! So sit back, get yourself a slice of cake and pour yourself a glass of fizzy pop as we find out who really did the best from the weekend’s football.

Before we get cracking with the ins and outs of the weekend scores, this is a good chance to outline how I determine the strength of opponent factor. The strength of the opponent is based on the league position of the team at the start of the weekend’s play on the Saturday, and each league position is assigned a difficulty rating. So if the team in second beats the team in sixth 2-0, then they would get a higher point score than if they had beaten the team in fourteenth by the same score, since logic would dictate that the higher your opponents are in the league, the tougher the game will be. (Note that the league positions for Week 1 are worked out using the final Premier League table from last season with the promoted teams taking the place of the bottom three respectively).

It was those Wanderers from Bolton who were this week’s big winners both on the pitch and in the Alternative Premier League (APL) with their 0-4 demolition job on new boys QPR. Their impressive opening day tally of 8.82 points was a result not only of their large winning margin, but also the fact that they did this away from home.

The reason that Manchester City scored a comparatively smaller total for the same winning margin comes down to the league position of themselves and Swansea and the fact that Man City were playing at home. Manchester City, who took third place, were playing Swansea who took twentieth place – almost as big a positional difference as you can get. Strength of opponent (Man City being placed higher than Bolton) is also the reason that Swansea are currently above QPR after Gameweek 1.

Wolves sneak ahead of Manchester United by 0.04 points after both teams recorded 1-2 wins away from home. As Wolves finished below Blackburn last year their result was comparatively better then United’s who were placed top and faced a West Brom side in eleventh position. The away bonus attached to both scores place both Wolves and United above Man City.

With five instances of teams cancelling each other out, scores for draws are calculated purely on league position as goal difference is clearly not a factor. So with this in mind, it is Stoke and their robust defence who emerge as the top drawing side. Stoke were placed eleven places below Chelsea, so a draw was a better result for the Potters than the Blues, and the points; Stoke 4.12 and Chelsea 3.28, are reflective of this.

Everton and Spurs take the bottom two positions as a result of their match being postponed.

Here’s how the table looks after gameweek 1:

What are your thoughts about the table after gameweek 1? Share your opinions in the comments section below.

12 thoughts on “Alternative Premier League: How the Table Looks After Gameweek 1”

  1. Some food for thought. Why is the position of Team A used in calculating the score? Shouldn’t what matters is the position of the Team B that they are playing?

    Your post noted penalizing Man City for being higher than Swansea, more so than Bolton was of QPR. This seems to be flawed because it penalizes a team for having a high standing, ie Man United would get punished this year for being in the top four every game this year, where as a team in the relegation zone would get a bonus for being so low, despite not being a “good” team. Rating a team based on the strength of their schedule, instead of including their own position on the table, makes much more sense. So having Stoke higher because they tied Chelsea instead of Wolves works, but Chelsea shouldn’t be negatively affected because they were in the top 4, and tied Stoke, rather because they tied a team that finished 13th last season.

    Man City and Bolton should gain comparative scores for the difficulty of the teams they played, both being recently promoted. Having a difference between home and away makes perfect sense.

    1. I completely agree with this also. You should be rewarded/punished for the team that you play not the team that you are.

  2. Agree with the point above – it makes no sense to penalize a team for a higher standing.

    And there should be a strong factor for Home vs. Away.

    Arsenal looked terrible against Newcastle because we expect better from Arsenal, but think of them simply as Team A vs Team B: The match was a 1-1 draw, with Team A (Arsenal) playing at Team B (Newcastle). Yet you ranked Team B much higher.

    Do you really think a home draw is more impressive than a road draw?

    I think you need to rethink your scoring system.

  3. I question the logic behind using league tables to determine team strength. It’s extremely messy data. For one thing, the entire idea of the alternative table seems to be based on the assumption that there are things the league table doesn’t tell us about how teams do, so having the league table be such a fundamental constraint is problematic. Second, because team strength does change dramatically in the off-season and because the table takes a few weeks to sort, you’ve got a lot of noise in the data. Right now, your model tells us that Bolton and Wolves are as strong teams as Manchester United and Manchester City, and that they’re considerably stronger than Arsenal, Chelsea, Tottenham, or Liverpool. This is obviously untrue, and any calculations you make next week founded on the assumption that it is true are just going to be ridiculous. Early in the season, the numbers are going to respond violently to any new data, and almost always overcorrect because of it. Right now, in your model, the Bolton/Man City game is going to be a clash of the titans that simultaneously gives a massive award for the winner for beating a top tier team and smashes the loser halfway down the table. That’s obviously rubbish.

    Once you get to the middle or end of the season it’ll start to shake out, but it’s eventually going to pick up the opposite problem – it will be too beholden to old data. In the latter half of last season Liverpool began acting like one of the strongest teams of the league, but a model like yours would have had no hope of picking that up because of the damage done by the first half of the season.

    It also fails to respond to specific situations well at all. Yeah, Stoke is a mid table team. But more accurately, Stoke is a brutal team at home and a lightweight away. A stronger model would treat Chelsea playing at Stoke as a reasonably touch match for Chelsea, whereas Stoke playing at Chelsea as a presumed walk in the park. Stoke, after all, got points off of Arsenal, Liverpool, Man City, and Chelsea at home last season, whereas they dropped points to all three relegated teams away last season. Compare to a team like Sunderland, who had considerably more consistency, and you get a very different picture.

    There’s a much better approach to be had by using head to head records and comparative records against other teams, with some hedges to deal with the newly promoted teams (using records from the Championship against recently relegated teams to get comparative records with Premier teams, and historical performance data of newly promoted teams at first, then valuing new data very highly for those teams at first until the information gap has closed a bit).

    1. I don’t think the poster is assuming that their Table in the first week is very indicative of the team rankings in the EPL, just as the EPL Table after Week One doesn’t either.

      Honestly, the problem with strength of schedule is that it can’t possibly account for teams having good or bad games (rather as an average) or screwjobs where a team is robbed of a goal or get Arsenal’d like with the absurd penalty rewarded to Liverpool last season or when Man. United was give a couple bonus minutes to tie Man City, though it at least does trend with how a team is currently doing.

      Honestly, trying to computerize any system for ratings will be rife with “issues”. It just can’t be done because some things aren’t quantifiable. For instance, late in the season when Arsenal falls apart, teams playing them will get about as much benefit from playing them as they would a tougher full season team like Man United. You just can’t quantify such things. :^P

      I do think it is an interesting experiment.

  4. A Manchester City team that beat Swansea 4-0 is lower than a Man United team that beat West Brom 2-1?
    That doesn’t seem right.
    How many goals would City have to beat Swansea by to equal a 2-1 win over West Brom??
    6 or 7?

  5. Scores can be erratic for the first few weeks, this is a good observation, but things order themselves out over the season.
    A team’s own league position is taken into account to keep perspective of comparative achievements. So if Team A in 20th beat Team B in 1st 1-0, they would get a higher score than if Team C in 2nd beat Team B by the same score.

    1. The place where I could see differential placement in the table mattering would be in accounting for the score in the game. A 20th place team beating the 1st place team 1-0 is a huge feat, where as a 1st place team beating a 20th place team 1-0 is a lesser feat. So that would make sense via wins or losses, but still not draws because there is no goal differential to “modify”.

      I agree with Dave C about sharing the formula used. I’m geek enough to delve much deeper into this than any normal person should want to. :^P

  6. Evan, Yespage and Phil Sandfer all make good points.

    Karl – I mentioned it in the previous thread on this subject, but you should just post the actual methods you use rather than just describe the criteria you are looking at (i.e. what values are the multiplier factors for away wins, how do you calculate a multiplier based on team-strenght/league position, etc). Why not do this?

    Without such info, we’re really just stabbing in the dark by discussing the mysterious logic that has given Bolton (for example) “8.82 points”.

  7. By including a ‘strength of schedule’ measure in your calculations I think you are trying to make the table reflect the ‘Have they played the hard matches yet?” question.

    Starting with last year’s table seems like a good choice, but immediately switching to this year’s table after the first week doesn’t make sense. You might consider smoothing together the results from last season into this. As time passes, the older data counts less and the newer data counts more. See exponential smoothing here: You can also tune how strongly weighted the newer results are.

Leave a Reply

Your email address will not be published. Required fields are marked *