I thought it was important to go into some methodological details on my predictions, so if you’re interested here are the technical details on my method.
Week 30 predictions and results are in: my model predicted 8/10 games correctly with an asterisk on Crystal Palace’s road win over Stoke City because the model predicted 35% win probability for both teams.
The two errors were West Ham’s win over Sunderland, where the model predicted a 43% likelihood of a draw, although West Ham winning was the second most likely result at 33%. An acceptable error.
The bigger error was missing Manchester United’s road win over Liverpool, which my model had as the least likely outcome at 19%. Not sure what went wrong there, although it largely agreed with the betting lines on the game. so I’m not the only one who got it wrong.
Overall a 9/11 success rate in a system with three outcomes is strong. There might need to be some tweaks, but overall a good week.
|Game||Home Team Win %||Draw %||Visiting Team Win %||Actual Outcome||Correct Prediction|
|Manchester City vs. West Bromwich Albion||60%||28%||11%||Man City 3 - WBA 1||Yes|
|Stoke City vs. Crystal Palace||35%||28%||35%||Stoke City 1 - Crystal Palace 2||Yes*|
|Tottenham vs. Leicester City||75%||12%||12%||Tottenham 4 - Leicester City 3||Yes|
|Southampton vs. Burnley FC||75%||6%||18%||Southampton 2 - Burnley 0||Yes|
|Aston Villa v. Swansea City||26%||12%||61%||Aston Villa 0 - Swansea City 1||Yes|
|Newcastle vs. Arsenal||11%||24%||64%||Newcastle United 1 - Arsenal 2||Yes|
|West Ham United vs. Sunderland||33%||43%||22%||West Ham 1 - Sunderland 0||No*|
|Liverpool vs. Manchester United||32%||49%||19%||Liverpool 1 - Manchester United 2||No*|
|Hull City vs. Chelsea||70%||27%||3%||Hull City 2 - Chelsea 3||Yes|
|Queens Park Rangers vs. Everton||23%||34%||42%||QPR 1 - Everton 2||Yes|