clock menu more-arrow no yes mobile

Filed under:

xGunners: Week 26 Premier League simulated match results

Odds for Arsenal vs Everton, plus the rest of the matches

Everton v Leicester City - Premier League
Theo Walcott returns
Photo by Mark Robinson/Getty Images

The home stretch of the Premier League season is here. With just 13 matches left to play there is a lot left to be decided.

Except for those things like the title (Manchester City would have to have a historic collapse to lose it) or where Arsenal will finish (they have 6th pretty much locked up after that loss away to Swansea).

Still, the top 4 race is very tight with four teams fighting for three spots and Arsenal still not mathematically eliminated and stalking waiting for a couple teams to stumble. The relegation fight is even crazier with ten teams in real danger of possibly going down.

Week 26 Simulated odds

Arsenal this weekend have the return of Theo Walcott to look forward to with Everton coming to the Emirates. New signing Henrikh Mkhitaryan will likely make his first start for the Gunners but new striker Pierre-Emerick Aubameyang’s debut for the club is in serious jeopardy with an illness potentially keeping him out.

Arsenal are the favorites for this match but I would expect Everton to try to replicate what Swansea did to Arsenal on Tuesday with a low block and quick counters. This could really trouble Arsenal if they don’t sort out their midfield to be able to cut out those attacks quicker.

The big match of the weekend is Liverpool vs Tottenham. This match has massive top four implications not just for these two but for Arsenal’s slim chances. The best result for Arsenal is a draw combined with an Arsenal win against Everton.

Simulated finish and points

Using the same match odds that I simulated, I also ran 10,000 simulations of the season as whole. I put a few tables together to show the results in an easily understood (hopefully) visual format.

This table shows the chances of a team’s finishing the season in each of the 20 places in the table:

This chart, a “box and whisker” plot, shows the range of probable points for each team in the league based on the simulations. Gray represents the lower half of the midrange of possible points (25-50%), yellow the upper half (50-75%):

This last table is pretty straightforward. It’s the odds of a team winning the title, finishing in the top four, or being relegated, and the average number of points earned in the 10,000 simulations run for this data:

Simulated Finish

Team Title Top 4 Relegation Avg Sim Pts
Team Title Top 4 Relegation Avg Sim Pts
Manchester City 99.80% 100.00% 0.00% 94.67
Liverpool 0.10% 82.20% 0.00% 74.83
Manchester United 0.10% 79.50% 0.00% 74.25
Tottenham Hotspur 0.10% 76.70% 0.00% 73.8
Chelsea 0.00% 68.00% 0.00% 72.7
Arsenal 0.00% 5.30% 0.00% 63.38
Leicester City 0.00% 0.10% 0.00% 54.04
Bournemouth 0.00% 0.00% 0.10% 50.67
Burnley 0.00% 0.00% 0.00% 49.38
Everton 0.00% 0.00% 0.60% 47.6
Crystal Palace 0.00% 0.00% 4.40% 43.8
Brighton & Hove Albion 0.00% 0.00% 11.20% 41.32
Watford 0.00% 0.00% 19.20% 39.73
Southampton 0.00% 0.00% 20.20% 39.63
West Ham United 0.00% 0.00% 26.80% 38.72
Stoke City 0.00% 0.00% 27.10% 38.61
Newcastle United 0.00% 0.00% 31.00% 38.14
Swansea City 0.00% 0.00% 39.10% 37.12
Huddersfield Town 0.00% 0.00% 41.70% 36.89
West Bromwich Albion 0.00% 0.00% 47.10% 36.21

Team ratings and expected points

Next up: team ratings, which are derived from the inputs that go into the model.

These team rankings use data from the 2015-16 season through the current season. Data that is more recent is weighted more heavily. Rankings are compared to league average, and scaled so that 100 is average. Each point above or below 100 represents 1 percent better or worse than the average team.

Last but not least is expected points for each team this season.

There are two expected points totals shown here. The first is based on the pre-match odds that I do before every match, as I have done above, and the second is based on the post-match xG that each team produces. More information on xPoints can be found here.

If you are interested in the methodology of the model, or for any of the work I do here, you can find that on my personal blog.