Jake Osgathorpe uses Infogol xG data to take a look at the last 20 Premier League seasons and compare the champions by putting them in a 20-team, 38-game season simulation...
"Of the league simulation we have put the 20 champions into, ____ 'win' the league 32% of the time."
*Based on Liverpool being crowned PL champions this season
We used a variety of models to compare teams, recognising that winning a game is the primary aim for a football team, but also acknowledging that the margin of victory can add extra information about a team's underlying quality.
It was a challenge, involving evaluating each performance across a single season and then relating that to the changing competitive balance that exists across two decades.
We could simply rank each team from one to 20, but instead we put them all in a league simulation, with each team playing one another at home and away.
The champion of champions of the last 20 Premier League seasons are... Manchester City's 17/18 team, the side who broke pretty much every record on their way to a never-done-before 100-point season, finishing a record-breaking 19 points clear of second placed Manchester United.
Of the league simulation we have put the 20 champions into, Manchester City 17/18 'win' the league 32% of the time.
As has been the case over the last few years, it was Manchester City and Liverpool battling it out for supremacy in this simulation, with Jurgen Klopp's 19/20 side - who are all-but certain to win the title - finishing second in the champion of champions sim.
The Reds would win the simulation 21% of the time. They really are an elite team and one of the best Premier League sides we have witnessed.
Manchester City sandwich Liverpool in the table, as their 18/19 team comes in as the third best champion from the last 20 Premier League seasons. They win the champion of champions sim 17% of the time.
In fourth and fifth are two Chelsea teams. Jose Mourinho's first season as Chelsea manager in 04/05 was a special one (pardon the pun), as the Blues were unstoppable, conceding just 15 goals in 38 games, keeping 25 clean-sheets. They win the sim 9% of the time.
Chelsea have won five of the last 20 league titles, and their 09/10 team under Carlo Ancelotti ranks higher than Arsenal's invincible team based on our model. That win the simulation 4% of the time.
I'm sure by now that many will be in uproar as to why Arsenal's invincible season hasn't yet featured in this list, and that is because that 03/04 side ranks as the seventh best champion of the last 20 seasons.
In the days before #xG there was "luck stripped shots on target". Here's what #AFC #Arsenal 200304 unbeaten season looks like in retrospective simulations of all games. Fancy #MCFC to have a better 200405 ;-). Go to @12Xpert site to get your shot data. pic.twitter.com/zylz0oraff? mark taylor (@MarkTaylor0) March 28, 2020
We dug deeper into Arsenal's 03/04 campaign using shots on target data (no xG data available), and calculated that the Gunners would have won the 03/04 Premier League title just 41% of the time based on performances, with their record on average reading 22 wins, 9 draws and 7 defeats.
All in all, as champions they weren't as convincing from a data stand-point as the teams above them in the champion of champions simulation, with the Gunners finishing in the top four in less than one in five simulations.
Manchester United have won 35% of the 20 available titles, all under the mastery of Sir Alex Ferguson, with the best Manchester United team on our ratings being the class of 06/07, and that team ranks only ninth in the champion of champions simulation.
The Champions League winning team of 07/08 was a sensational unit with an unplayable front three of Cristiano Ronaldo, Wayne Rooney and Carlos Tevez, but that team ranks in the bottom half of this league, as does the 08/09 team, Sir Alex Ferguson's sides third straight league title.
The side that ranks as the worst title winner of the last 20 seasons is Leicester and their miracle title success of 15/16. During that season, Leicester ranked as only the fourth best team based on expected goals, so from a data perspective, they were extremely fortunate to finish top of the table.
We calculate that Claudio Ranieri's side would finish bottom of this table over a 38-game season in 19% of simulations, finishing in the bottom three 45% of the time, with just a 1% chance of finishing in the top four.
In total, a Manchester City team would win this league 56% of the time from four entries. Both City title winning teams under Pep Guardiola have been sensational teams, with those two combining for a 49% chance of winning the simulation.
Liverpool have only one entry into this simulation (19/20) compared to seven from Manchester United, but the Reds have a better chance of winning the 20-team league (21%) than all seven United sides combined (4%).
In fact, Liverpool's 19/20 side have a better chance of winning the simulation from one entry than Manchester United and Chelsea do combined from 12 entries.
I think that shows just how good this current Liverpool team is, and how they should be lauded as one of the best ever Premier League teams.
Four of the bottom five are Manchester United teams (00/01, 02/03, 10/11 and 12/13), with all having at least a 25% chance of finishing in the bottom three.
Head to Head
Given this simulation was run as 38-game season, we have home and away probabilities for all match-ups, so I've condensed it down to five teams for the H2H section: Manchester City 17/18 centurions, Liverpool 19/20 first Premier League title, Manchester United 07/08 league and Champions League winning team, Chelsea 04/05 Specials and Arsenal 03/04 invincibles.
I'll run down the match-ups in alphabetical order.
Arsenal 03/04 vs Chelsea 04/05
We make the Gunners narrow 37% (2.77/4) favourites to win the meeting at Highbury, with Chelsea avoiding defeat 63% (1.6) of the time.
At Stamford Bridge, Chelsea are stronger favourites, as we give them a 45% (2.26/5) chance of getting the win, though the goal expectancy in both games is very low (63%, 66% under 2.5 goals) due to both sides staunch defences.
Arsenal 03/04 vs Liverpool 19/20
As can be read above, the current 19/20 Liverpool side are rated as the second-best team in this overall simulation, and our model would make them favourites both home and away.
At Highbury, we give Klopp's side a 38% (2.6313/8) chance of winning the game, and we make them nearly even money favourites at Anfield (49% - 2.0421/20).
Arsenal 03/04 vs Manchester City 17/18
Pep Guardiola's City team of 17/18 ranks as the best Premier League title winner in the past 20 seasons, so like with Liverpool above, our model makes them favourites both home and away.
We give them a 39% (2.568/5) chance of beating the invincible team at Highbury, and make them odds-on to beat the Gunners at the Etihad (51% - 1.9620/21). Again though, the goal expectancy is ow in both matches, with the most likely scoreline at the Etihad a 1-0 City win (16%).
Arsenal 03/04 vs Manchester United 07/08
A much closer contest than the previous three match-ups, as Arsenal are strong favourites to beat Sir Alex Ferguson's Champions League winning side at Highbury (44% - 2.27).
United are favourites at Old Trafford, but not we aren't as strong on their chances as Arsenal's home game, giving them a 40% (2.56/4) chance of getting the win.
Chelsea 04/05 vs Liverpool 19/20
An intriguing and closely matched game at Stamford Bridge, with Jose Mourinho's Chelsea Specials edging favouritism thanks to home field advantage (36% - 2.89/5).
We make the current Liverpool team stronger favourites at Anfield though, with Klopp's side winning the game 45% (2.226/5) of the time, though it would be no blow-out success (1-0 - 16%, 2-0 - 12%).
Chelsea 04/05 vs Manchester City 17/18
Based on our model, Manchester City narrowly edge favouritism in this match-up at Stamford Bridge, with Pep's side having a 35% (2.8615/8) chance of beating Mourinho's side on their own patch.
At the Etihad, City are stronger favourites (46% - 2.17).
Chelsea 04/05 vs Manchester United 07/08
Jose Mourinho's side would be strong favourites at Stamford Bridge to beat the team that would go on to beat them in the Champions League final, as our model gives a 47% (2.13) chance of a Chelsea win.
At Old Trafford, our model has the game as a pick'em, giving both sides a 35% (2.8615/8) chance of coming out on top.
Liverpool 19/20 vs Manchester City 17/18
The battle between the two best champions of the last 20 seasons. Liverpool, with home advantage, are 37% (2.77/4) favourites to get the better of Pep's side. This is only one of two matches from the 38-game simulation that City 17/18 are second favourites on our model (the other is away at City 18/19).
At the Etihad, City are marginally stronger favourites to get the win (41% - 2.447/5), but these two games would be titanic tussles between peak Liverpool and peak City.
Liverpool 19/20 vs Manchester United 07/08
It is fair to say that, based on our model and calculations, that this current Liverpool 19/20 side would have their way with Manchester United's 07/08 double winners.
At Anfield we make Liverpool odds-on favourites (51% - 1.9620/21), and at Old Trafford they are favourites too (39% - 2.568/5), though it would be great to see Liverpool's famed front three and United's front three of Cristiano Ronaldo, Wayne Rooney and Carlos Tevez on the pitch at the same time.
Manchester City 17/18 vs Manchester United 07/08
City are strong favourites in both of these games. Odds-on 1.855/6 (54%) favourites at the Etihad, and 2.47/5 (41%) favourites at Old Trafford.
It would be an intriguing battle for a number of reasons, another meeting of Sir Alec Ferguson and Pep Guardiola, but also the players on show. But based on our data, the City centurions were on a different level to any Manchester United team from 2000 onwards.