As a City fan, I am of course thrilled that ESPN thinks that City have the edge on Chelsea in the title race and will win approximately 2/3 of the time. However, as a statistician, I am far from convinced that is indeed the case. ESPN is still relying on its Soccer Power Index, an index with flaws so severe it inspired me to start blogging about soccer stats 5 years ago (on livejournal for Christ's sake) and predicted Brazil had a 65% chance of winning the World Cup at one point last summer.
Let's start by saying that this is the best time of the season to compare teams performance. Since each team plays each other in the first half of the year and plays an equal amount of home and road games, their strength of schedule is exactly the same for the first half. Therefore, you don't really have to adjust a team's level of performance at this point, you can just say both City and Chelsea have played equivalent schedules over the first 19 games. Somehow, ESPN doesn't think this is the case. In their description of SPI, they state "By considering how many goals
a team scores and allows, and the opposition and location of those
games, SPI generates an offensive and defensive rating for each club." Given that Chelsea and City have exactly identical scoring records, it stands to reason that SPI thinks City's schedule has been easier, despite it being exactly the same as Chelsea's in terms of teams played and number of away games. It's not considering away games at the top teams more damaging either, as Chelsea have already played the other top 4 teams away and City have only played 1 of them. And although both teams have played one more game since the midway point, Chelsea played Spurs away and City played Sunderland at home. I don't see how anyone could think City have played a tougher schedule, and I'm skeptical of SPI's projections as a result.
One of the other main problems with SPI is that it continues to look just at goals, and not shots on goal, to inform how well a team has played. In individual games, there are cases where teams put a ton of shots on goal and just run into a great keeper (say, Liverpool against Man United). Then there are games where a team plays terribly but manages to score on their few chances (say, United against Liverpool). Looking at shots on goal lets you focus on the overall team performance more effectively and not just if they got lucky in the box. If we look at shots on goal differential, Chelsea are performing ahead of City, primarily because they allow fewer shots on goal per game. City have benefited by Joe Hart being in great form and stopping a higher percentage of shots than his Chelsea counterpart Courtois, but as we saw last year that may not hold up forever.
Perhaps SPI is looking at momentum instead. After all, City have caught up to Chelsea only recently, maybe they are weighting the most recent results more heavily? Well City have averaged 2.67 points over their past 6 games compared to Chelsea's 1.67, but Chelsea have also played a much tougher schedule. And once we start looking at the shots on goal differential for those matches, Chelsea has actually had a better differential than City over those games, it's just that 54% of Chelsea's opponents shots on goal have resulted in goals, a clearly unsustainable rate. If we adjust it to league average finishing rates, Chelsea would average a goal differential of .96, better than City's .77 over the same span despite playing a tougher schedule.
There's also the fact that City will be more affected by the African Cup of Nations in January, that management places a higher value on the Champions League than the Premier League, and that Kompany and Aguero are still facing injuries. Of course, Chelsea lacks the depth of City's squad, and it's not unlikely that they come back down to earth after their fantastic start. It's quite possible to make a case that City should indeed be the favorites, it's just not possible to make a statistical one.