Chelsea and West Brom played out a three-all draw to the excitement of many neutrals, but here’s how the game looked through an analytical lens.
The Blues played West Bromwich Albion on Sunday in a six-goal thriller that ended 3-3. Chelsea controlled possession (75 percent) whilst West Brom mainly sat back in a low defensive block, looking to play on the counter-attack. The three goals Chelsea conceded came from turnovers in transition resulting from sloppy mistakes.
The Blues did well to fight back in the second half after going into halftime 3-0 down, managing to level the scores in the dying embers of the game. Frank Lampard and Chelsea fans are surely relieved to not lose to the newly-promoted side languishing at the bottom of the league. The product on the pitch was exciting, but here’s a dive into the data:
Assessing match performance based on expected goals (xG) shot-based expected goals
| Metrics | Chelsea | West Brom |
|---|---|---|
| Adjusted goals | 3 | 3 |
| Shot based xG | 2.54 | 1.04 |
| Adjusted shot based xG | 2.3 | 1 |
| Non-shot xG | 2.6 | 0.6 |
The table above shows Chelsea and West Brom’s shot-based xG. This tells us how many goals the teams were expected to score based on the quality of shots throughout the match. While the game ended 3-3, a deeper dig into the numbers tells a different story. Based on shot-adjusted xG, West Brom should have one goal. The Blues, on the other hand, could have had about 2.5 goals.
| Player | Shot based xG for the match | Adjusted for conversion rate |
|---|---|---|
| Mount | 0.001 | 0.0 |
| Havertz | 0.001 | 0.001 |
| Alonso | 0.03 | 0.0 |
| Abraham | 0.36 | 0.3 |
| Werner | 0.14 | 0.2 |
| Mount | 0.19 | 0.2 |
| Mount | 0.04 | 0.0 |
| Mount | 0.001 | 0.0 |
| Abraham | 0.12 | 0.1 |
| Alonso | 0.01 | 0.0 |
| Kovacic | 0.001 | 0.0 |
| Mount | 0.06 | 0.1 |
| Havertz | 0.27 | 0.34 |
| Mount | 0.06 | 0.1 |
| Hudson-Odoi | 0.13 | 0.1 |
| Werner | 0.02 | 0.03 |
| Mount | 0.13 | 0.1 |
| Abraham | 0.84 | 0.6 |
| Total | 2.54 | 2.26 |
The shot-based xG calculation is the summation of the xG from the shots taken in the game by the players. The xG in the table above is adjusted for the conversion rates of the players. The shot-based xG for the game against West Brom can then be adjusted by each individual player’s conversion rate to get an adjusted shot based xG. This is 2.26 versus the original non-adjusted 2.54.
The conversion rate (calculation in the table below) measures the number of goals scored based on the quality of chances. The conversion rate is calculated by dividing the number of goals each player has scored by the number of goals they were expected to score (xG) last season.
| Player | Goals | xG | Conversion rate |
|---|---|---|---|
| Azpilicueta | 4 | 4.03 | 0.99 |
| Mount | 7 | 7.99 | 0.88 |
| Werner | 29 | 23.1 | 1.26 |
| Havertz | 19 | 15.18 | 1.25 |
| Arrizabalaga | 0 | 0 | 0 |
| Jorginho | 9 | 7.81 | 1.15 |
| Zouma | 1 | 1.46 | 0.68 |
| Kovacic | 2 | 1.89 | 1.06 |
| Chilwell | 3 | 1.41 | 2.13 |
| Abraham | 14 | 19.13 | 0.73 |
| James | 2 | 1.76 | 1.14 |
| Kante | 2 | 1.44 | 1.39 |
| Rudiger | 2 | 0.8 | 2.5 |
| Christensen | 0 | 0.96 | 0 |
| Alonso | 4 | 3.38 | 1.18 |
| Silva | 1 | 1.15 | 0.87 |
| Pulisic | 12 | 10.33 | 1.16 |
| Ziyech | 4 | 5.55 | 0.72 |
| Clarke-Salter | 1 | 0.41 | 2.44 |
| Bakayoko | 1 | 2.69 | 0.37 |
| Miazga | 2 | 1.13 | 1.77 |
| Hudson-Odoi | 3 | 3.33 | 0.9 |
| Giroud | 14 | 11.22 | 1.25 |
| Tomori | 1 | 1.74 | 0.57 |
| Barkley | 7 | 5.53 | 1.27 |
| Caballero | 0 | 0 | 0 |
| Sarr | 1 | 0.46 | 2.17 |
| Emerson | 0 | 0.86 | 0 |
| Moses | 0 | 0.92 | 0 |
| Gilmour | 0 | 0.28 | 0 |
| Player | Shot based xG for the match | Adjusted for conversion rate |
|---|---|---|
| Robinson | 0.3 | 0.3 |
| Bartley | 0.41 | 0.27 |
| Townsend | 0.06 | 0.06 |
| Diangana | 0.12 | 0.19 |
| Sawyers | 0.13 | 0.13 |
| Philips | 0.02 | 0.02 |
| Total | 1.04 | 0.98 |
The same was done for the Baggies to get their adjusted xG of 0.98. But wait there’s more… Non-Shot expected goals
Non-shot expected goals
Finally, we have non-shot expected goals, which measure’s how many goals a team should score based on data—as opposed to shots. One such factor is the number of passes into the opposition’s penalty area. According to a model by Joyce Boice, a computational journalist, passes into the opposition’s six-yard box/penalty area on average leads to a goal 14 per cent of the time.
| Player | Passes to penalty area per 90 | Passes to penalty area accuracy percentage | Accurate |
|---|---|---|---|
| Kante | 0.92 | 100 | 0.92 |
| James | 11.02 | 50 | 5.51 |
| Havertz | 2.76 | 66.67 | 1.84 |
| Mount | 4.59 | 40 | 1.84 |
| Silva | 1.22 | 100 | 1.22 |
| Hudson-Odoi | 15.88 | 44.44 | 1.22 |
| Total | 40 | 18 |
| Actions in opposing team box | Chelsea | Percentage of time leads to goal | Goal |
|---|---|---|---|
| Accurate passes in six yard box for Chelsea | 18 | 14% | 2.6 |
| Accurate passes in six yard box for West Brom | 4 | 14% | 0.6 |
The first row of the table above shows the total number of successful passes into the penalty area (per 90 times the success rate) for each player and sums it up. That represents 18 successful passes into the penalty area by the Blues (calculated in the table above this one). If we multiply that by 14 per cent (i.e the number of times on average a pass into the penalty leads to a goal) we get 2.6. Based on this metric, we can say the Blues should have scored around two to three more goals.
A similar computation is done for West Brom to show that an additional 0.6 goals should have been scored by the Baggies based on this metric and assumption.
Conclusion
Adding up Adjusted Shot based xG and Non-Shot xG the game could have potentially been 5-3 based on this simple calculation
Chelsea did create the better chances and dominated the game, even in the first half. It is the Blues’ defensive mistakes that cost the team the game. Marcos Alonso and Thiago Silva made poor passing choices by passing the ball straight into the path of the opposition that lead to the first two goals. The third goal was a result of something we have seen far too often: poor defending from set-pieces.
Other than these three errors, West Brom did not have much to show in terms of chance creation and good quality shots. Chelsea did have the better chances but failed to finish some good quality looks. For instance, Timo Werner’s shot over the bar and Abraham’s miss from a Reece James cross early in the first half.
As the season goes on, expect Chelsea to be more clinical in front of goal. However, the key to the team’s success this season lies in eliminating defensive mistakes.
All data is from Wyscout. The model referenced is FiveThirtyEights’s prediction model.