Manchester City's title winners last season are the best ever Premier League team with Liverpool fourth, according to one Reddit user's detailed algorithm.
Pep Guardiola's 2018-19 side are the greatest side the league has ever seen ahead of Chelsea in 2004-05, Man City in 2017-18 and Liverpool's runners-up last season.
That means the best Manchester United team come in 5th when they won the league in 2008 with Cristiano Ronaldo and Wayne Rooney.
Advert
Whats worse for Red Devils fans is despite 13 Premier League titles, they only feature in the top 10 of this list once.
Tottenham's runners-up in 2016-17 rank in sixth place, but the Chelsea side that won the title that season rank 14th.
Advert
Reddit user Tsubasa_sama took the 19 best ever Premier League teams according to transfermarkt.com and added in Derby's awful side from 2007-08 that picked up just 11 points.
He then simulated 10,000 seasons worth of results from the sides playing each other and came up with these surprising results.
Explaining his methods on Reddit, he wrote: "A season in a 20-team league is composed of 380 matches, each of which is made up of a home team and an away team.
"The fundamental assumption I will be making is that the number of goals scored by a team follows a Poisson distribution with mean dependent on the teams' attacking strength and the oppositions' defensive strength.
"This is a reasonable assumption since the Poisson distribution is typically skewed towards lower numbers when the mean is small (and football is a low-scoring game).
Advert
"However it is not perfect - for example the occurrence of goals in a match is not independent from when the last goal occurred, also a goal being scored is a rare event in a football match and so you will run into sample size issues.
"A better compromise would be to use a shot-based metric like xG since shots are much more frequent events, though xG statistics were not available for some of the older teams in this list so that had to be abandoned.
"Since this is just for fun I decided not to look too far for the perfect model and stuck with what I had, which was simple to code."
He added: "The Home Field Advantage is a well-established phenomenon in football and so for each of the 20 teams I want to look at how they performed at home and away, treating each separately.
Advert
"Therefore for each team I need to find out four things: their home attack, home defence, away attack and away defence."
Using Manchester City from last season as an example, he wrote: "In the 2018-19 season Man City scored 57 goals at home in 19 matches. This is an average of 3.000 per match. So against an "average team" in the league (which may not exist) they are expected to score 3 goals at home. This number is Man City 2018-19's home attack.
Advert
"Man City conceded 12 goals at home in 19 matches for an average of 0.632 per match. Relative to the hypothetical "average team" this is a factor of 0.533 times as many goals compared to what the average team in the league conceded at home in the 2018-19 season per match (1.186). So Man City's home defence is 0.533.
"Repeating this for all the other teams in the league home and away will give us the numbers needed to find the Poisson mean for both teams in a match. Let's have a look at an example to see how these numbers are used to predict a match outcome."
Using Manchester City's team from last season against Derby in 2007-08, he came up with the following.
"The mean number of goals I expect Man City to score in this match is given by the formula: MCI2018-19 home attack * DER2007-08 away defense = 3.000 * 1.583 = 4.749.
"Similarly the mean number of goals I expect Derby to score in this match is given by: DER2007-08 away attack * MCI2018-19 home defense = 0.421 * 0.533 = 0.224.
"So the expected scoreline in this match will be a 4.749 - 0.224 win in favour of Man City, i.e. roughly 5-0 on average."
Did you get all of that?
Featured Image Credit:Topics: Premier League