A few weeks back, I created a model to try and predict how the USL Western Conference would finish. With the regular season ending this weekend, how accurate was my model?
As a reminder, here’s the original final table that my model predicted:
| W | D | L | PTS | ||
| 1 | Real Monarchs | 20 | 6 | 6 | 66 |
| 2 | Reno 1868 | 19 | 7 | 6 | 64 |
| 3 | Swope Park Rangers* | 18 | 7 | 7 | 61 |
| 4 | Phoenix Rising* | 18 | 7 | 7 | 61 |
| 5 | San Antonio* | 17 | 10 | 5 | 61 |
| 6 | Tulsa Roughnecks | 16 | 4 | 12 | 52 |
| 7 | OKC Energy | 13 | 8 | 11 | 47 |
| 8 | Orange County | 11 | 10 | 11 | 43 |
| 9 | Sacramento Republic | 12 | 6 | 14 | 42 |
| 10 | Colorado Springs Switchbacks | 11 | 8 | 13 | 41 |
| 11 | Rio Grande Valley Toros | 9 | 7 | 16 | 34 |
| 12 | Seattle Sounders 2 | 9 | 4 | 19 | 31 |
| 13 | Los Angeles Galaxy II | 8 | 5 | 19 | 29 |
| 14 | Vancouver Whitecaps 2 | 6 | 8 | 18 | 26 |
| 15 | Portland Timbers 2 | 2 | 6 | 24 | 12 |
*Swope Park and Phoenix finish above San Antonio on wins (the first tiebreaker); but Swope Park will probably beat Phoenix on goal difference (the second tiebreaker).
With one game left for every team except Swope Park (who completed their regular season in midweek), here’s the current league table:
| W | D | L | PTS | ||
| 1 | Real Monarchs | 19 | 7 | 5 | 64 |
| 2 | San Antonio | 16 | 11 | 4 | 59 |
| 3 | Reno 1868 | 17 | 7 | 7 | 58 |
| 4 | Swope Park Rangers | 17 | 7 | 8 | 58 |
| 5 | Phoenix Rising | 16 | 7 | 8 | 55 |
| 6 | Tulsa Roughnecks | 14 | 4 | 13 | 46 |
| 7 | OKC Energy | 13 | 7 | 11 | 46 |
| 8 | Sacramento Republic | 13 | 6 | 12 | 45 |
| 9 | Colorado Springs Switchbacks | 11 | 8 | 12 | 41 |
| 10 | Orange County | 10 | 10 | 11 | 40 |
| 11 | Rio Grande Valley Toros | 9 | 8 | 14 | 35 |
| 12 | Seattle Sounders 2 | 9 | 4 | 18 | 31 |
| 13 | Los Angeles Galaxy II | 8 | 5 | 18 | 29 |
| 14 | Vancouver Whitecaps 2 | 5 | 9 | 17 | 29 |
| 15 | Portland Timbers 2 | 3 | 6 | 22 | 15 |
The real table is definitely moving towards my model’s final table prediction. The biggest exception is Sacramento making the playoffs, thanks in part to their win against Rising (my model had slightly favored Rising winning that encounter).
The model had also predicted that San Antonio would ultimately drop out of the top 4, but they are now guaranteed to finish in 2nd or 3rd place.
So where did the model go wrong?
San Antonio’s form has been very similar to the model’s predictions. They are currently only 1 point ahead of schedule; and if they fail to beat Rio Grande Valley this weekend, they’ll finish with either 1 or 2 points less than my model predicted. The primary reason why they haven’t dropped out of the top 4 is that their closest rivals performed slightly worse than my model had predicted.
Is that a surprise? Not really. A model based on past performance will almost always predict the top clubs to win. In reality, though, slip ups can and do happen.
The model had predicted a three-way tie between Swope Park, Rising, and San Antonio, with the teams ranked according to the most wins, then goal difference in the final table. It didn’t take much to swing those positions. Swope Park, for example, have finished the regular season with 3 points less than the model predicted; and Rising to date are also 3 points down.
With one weekend of games to go, though, the model’s predictions are not that far away from reality; and it’ll be interesting to repeat the modeling again from the middle of next season.
Sadly, and contrary to my September prediction, it looks like there’ll be no top 4 finish for Rising…unless, of course, they manage a 9 goal victory on Saturday!