With just days to go before the English Premier League (EPL) starts its 19th season, punditry is going on record with their predictions, so it’s only fair that I also give you ammunition which, come next May, you can use to shoot me down from whatever pedestal onto which I’ve moved. And I can guarantee you, I will have moved on, as preseason predictions are something I usually loathe. After all, they’re just glorified (if often highly-informed) lists.
Premier League 2010-11 Preview: Using SCIENCE To Project The Final Standings
With the English Premier League’s start just days away, SB Nation Soccer editor Richard Farley decides to move away from a traditional predictions piece. All hail the power of the multi-core processor.
This year, the SB Nation Soccer desk is taking a different approach, employing a method I’ve been using over the last couple of years. The basic process goes something like this: Look at each player’s performance from the preceding season and try to determine his contributions to the team’s goals for and goals allowed. Regress that performance as needed, and if a player is no longer with the same club, try to make the evaluation as team-neutral as possible. Add in considerations for improvement or decline. Then take the player and plug him into his team’s depth chart for the 2010-11 season and make an assessment regarding the playing time. With that, you can estimate how much a player will effect his team’s goals for and allowed for this season.
If you read that with a combination of “Huh” and “Okay, but how, exactly,” don’t worry. That’s half-intended because this isn’t even half a science, but there are some huge benefits to this approach. With it, you can do things like this:
| Rk | Club | Avg | W | D | L | GF | GA | 1st | Top 4 | Top 7 | Relegated | Best | Worst | Range* |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Manchester United | 1.7 | 24.6 | 7.4 | 6.0 | 79.0 | 23.5 | 57.7% | 98.1% | 100.0% | 0.0% | 1 | 7 | 1-4 |
| 2 | Chelsea | 2.8 | 23.1 | 6.0 | 8.9 | 95.0 | 41.9 | 19.0% | 88.4% | 99.6% | 0.0% | 1 | 10 | 1-6 |
| 3 | Arsenal | 2.8 | 22.9 | 6.5 | 8.6 | 88.1 | 38.0 | 18.5% | 88.2% | 99.5% | 0.0% | 1 | 10 | 1-6 |
| 4 | Tottenham Hotspur | 5.3 | 19.5 | 7.4 | 11.1 | 68.4 | 41.9 | 1.4% | 36.0% | 89.5% | 0.0% | 1 | 17 | 2-9 |
| 5 | Liverpool | 5.4 | 19.2 | 8.0 | 10.8 | 62.2 | 37.9 | 1.3% | 33.6% | 88.0% | 0.0% | 1 | 17 | 2-9 |
| 6 | Manchester City | 5.5 | 18.7 | 9.1 | 10.2 | 54.2 | 31.7 | 1.4% | 31.8% | 87.2% | 0.0% | 1 | 18 | 2-9 |
| 7 | Everton | 6.1 | 18.5 | 7.6 | 11.9 | 63.7 | 43.1 | 0.7% | 21.1% | 79.6% | 0.0% | 1 | 17 | 2-10 |
| 8 | Stoke City | 10.0 | 14.6 | 8.7 | 14.7 | 46.1 | 46.2 | 0.0% | 1.1% | 18.1% | 1.7% | 2 | 20 | 5-16 |
| 9 | Aston Villa | 10.1 | 15.1 | 7.1 | 15.8 | 57.3 | 59.8 | 0.0% | 0.9% | 16.5% | 1.8% | 1 | 20 | 6-16 |
| 10 | Fulham | 12.5 | 13.2 | 7.6 | 17.2 | 48.2 | 60.2 | 0.0% | 0.2% | 4.6% | 8.2% | 3 | 20 | 7-19 |
| 11 | Sunderland | 12.6 | 13.1 | 7.9 | 17.0 | 45.7 | 57.0 | 0.0% | 0.2% | 4.5% | 8.6% | 3 | 20 | 7-19 |
| 12 | Birmingham City | 13.0 | 13.3 | 6.6 | 18.1 | 54.9 | 71.8 | 0.0% | 0.1% | 3.2% | 10.5% | 3 | 20 | 7-19 |
| 13 | Bolton Wanderers | 13.5 | 12.7 | 7.2 | 18.0 | 48.8 | 65.7 | 0.0% | 0.1% | 2.5% | 13.4% | 3 | 20 | 8-19 |
| 14 | Newcastle United | 13.7 | 12.1 | 8.6 | 17.3 | 39.4 | 53.5 | 0.0% | 0.0% | 2.2% | 15.3% | 4 | 20 | 8-19 |
| 15 | West Ham United | 14.2 | 12.4 | 6.7 | 18.9 | 51.3 | 73.1 | 0.0% | 0.1% | 1.8% | 20.0% | 2 | 20 | 8-19 |
| 16 | Blackburn Rovers | 14.6 | 11.9 | 7.5 | 18.6 | 44.4 | 65.0 | 0.0% | 0.0% | 1.2% | 22.3% | 4 | 20 | 8-19 |
| 17 | West Bromwich Albion | 14.9 | 11.6 | 7.8 | 18.6 | 41.5 | 62.2 | 0.0% | 0.1% | 1.0% | 26.0% | 4 | 20 | 9-20 |
| 18 | Wigan Athletic | 15.7 | 11.5 | 6.5 | 20.0 | 49.2 | 78.8 | 0.0% | 0.0% | 0.5% | 35.0% | 5 | 20 | 9-20 |
| 19 | Wolverhampton Wanderers | 16.0 | 11.0 | 7.1 | 19.8 | 43.2 | 71.6 | 0.0% | 0.0% | 0.6% | 40.3% | 5 | 20 | 10-20 |
| 20 | Blackpool | 19.7 | 6.7 | 6.7 | 24.6 | 28.9 | 86.5 | 0.0% | 0.0% | 0.0% | 96.9% | 8 | 20 | 18-20 |
* - the Range represents two standard deviations in either direction of the clubs’ mean finish.
Those are the projections you get when you take the player evaluations, aggregate them into team performance, and then write software to simulate 10,000 seasons. Every match is played adjusting for home-and-road conditions, the goal-scoring level of the league, and quality of opposition. And just as each match in real life has an inherent variability to it, so do the simulated matches, which is why we’ve run 10,000 seasons - to tease out anomalous results.
Thanks to this approach, we can provide more than just an ordered list. Scoring tons of goals or too many allowed? We get an idea of why a team may finish in a certain position. How likely is a team to make Europe, Champions League or win the league? We can make an informed guess. Could they get relegated? If so, how often? And even considering the anomalous result, what’s a team’s best and worst possible finish? Extremes aside, what’s a realistic range of performance? This method allows us to assess all those things.
But let’s not get too carried away. At its heart, the tool is just people making qualitative assessments about players. Those assessments are solidified (if you will), aggregated and then processed by a number of scripts, but the core is still looking at a player and asking “how is he going to do this year?”
I would like to think this process forces us to take a more detailed look at each player - incorporate more data, more rigorously than others might - but all the matters is the end product. You get to be the judge of that.
Over the next three days, we’ll incorporate these end results into our club profiles, taking those questions marks in the table, filling them in, and giving you our outlook on the 2010-11 Premier League season.
See More:
