Premier League 2017-2018

Found in [R , soccer , EPL] by @awhstin on

This is a page dedicated to weekly predictions for English Premier League. I am a fan of the Premier League and I support the Southampton Saints and am a long time Gooner though try as I might no sort of projection I do can make them better. A lot of the data for this comes from the awesome engsoccerdata package available on github. My predictions are under constantly construction, but they are based on Poisson distributions, and you can read a little bit about those here. Though this will appear on my blog feed it can be found in later weeks on the navigation bar via the button.

Results Summary

right wrong total
58% 42% 270


Position   Team MP W D L F A Diff P
1 Manchester City 27 23 3 1 79 20 59 72 L W W D W
2 Manchester United 27 17 5 5 51 19 32 56 W W L W L
3 Liverpool 27 15 9 3 61 31 30 54 W L W D W
4 Chelsea 27 16 5 6 49 23 26 53 D W L L W
5 Tottenham Hotspur 27 15 7 5 52 24 28 52 W D W D W
6 Arsenal 27 13 6 8 51 36 15 45 L W L W L
7 Burnley 27 9 9 9 21 24 -3 36 L L D D L
8 Leicester City 27 9 8 10 39 40 -1 35 D W L D L
9 Everton 27 9 7 11 32 46 -14 34 L D W L W
10 AFC Bournemouth 27 8 7 12 31 41 -10 31 W D W W L
11 Watford 27 8 6 13 37 47 -10 30 D L D W L
12 West Ham United 27 7 9 11 34 46 -12 30 W D D L W
13 Newcastle United 27 7 7 13 25 36 -11 28 D L D D W
14 Brighton & Hove Albion 27 6 10 11 22 36 -14 28 L L D W D
15 Crystal Palace 27 6 9 12 25 42 -17 27 W L D D L
16 Swansea City 27 7 6 14 20 37 -17 27 D W W D W
17 Huddersfield Town 27 7 6 14 23 47 -24 27 L L L L W
18 Southampton 27 5 11 11 28 40 -12 26 D D D W L
19 Stoke City 27 6 7 14 27 53 -26 25 L W D L D
20 West Bromwich Albion 27 3 11 13 21 40 -19 20 W D L L L

Week 28

Week 27 played out about as I expected but to a different degree altogether. Huddersfield clattered Bournemouth 4-1 and the North London Derby ended with the better team on the day winning. I imagine that the Liverpool vs. West Ham, United vs. Chelsea and Palace vs. Tottenham are going to be exciting games to watch.

Home Home Goals Away Goals Visitor
Leicester City 2 1 Stoke City
AFC Bournemouth 1 0 Newcastle United
Brighton & Hove Albion 3 1 Swansea City
Burnley 2 1 Southampton
Liverpool 3 1 West Ham United
West Bromwich Albion 2 1 Huddersfield Town
Watford 1 2 Everton
Manchester United 2 1 Chelsea
Crystal Palace 1 3 Tottenham Hotspur

Week 27

I finally have caught up, streamlined my scoring, and updated this page. We have seen quite a few upsets. Swansea seem to be giant killers now, and now that the transfer window has closed there are new faces to add to the equation. Only big surprise here is that I have Huddersfield beating a pretty solid AFC Bouremouth side.


Week 26

I am so far behind.


Week 25

After a sun and fun filled week I am back at it. Though work has officially started back up I still only have vacation on the brain luckily I have things like soccer predictions to distract myself with. Since I’ve last been here Manchester City’s record setting run finally came to an end with a thrashing from Liverpool 0-4. Then in a song as old as time Liverpool lose to bottom-of-the-table Swansea. This week is pretty uneventful but I think the Everton-Leicester and the Newcastle-Burnley contests will be games to watch… Oh and Tottenham vs. Man. United.


Week 24

On vacation.


Week 23

There are quite a lot of changes this week. I realized a couple weeks ago how unweidly the page has become with results, not to mention how hard to navigate and long the compiling takes, so I decided to do something about it. I removed the calculations from previous weeks as well as the results charts, call it a learning experience. I have kept the changes and the notes which you can scroll down to view. These will help me after the season look at when and why changes were made. I will most likely add some form of the previous results table back in the next couple week when I have time.


Week 22

No time


Week 21

No time


Week 20

No time


Week 19

Where do I start? Manchester City continues their steamroll through all challengers and KDB has started taking games into his own hands (feet). Roy Hodgson’s Crystal Palace move up to 14th with West Ham nipping at their heels with convincing performance away at Stoke. For the coming couples weeks the weighted distributions will stay the same as it has been fairly accurate.


Week 18

The games midweek are always a surprise. Two big draws shuffle up the Premier League table going into this weekend with Arsenal drawing West Ham United away and Liverpool drawing with West Brom at home at Anfield. The weekend could likely see Burnley back up in fourth if the Spurs lose to Man City, Arsenal fail to pick up points again Newcastle and Liverpool, who struggled at home, fail to score away at Bournemouth. Stock up on Guinness this weekend should be good.


Week 17

There is no time.


Week 16

I added more weight to the last five game performance to see how that plays out in the next two weeks. Week 15 showed us that even when a team has 75% possesion, 16 shots on, and 33 shots total they can still come out two goals behind (I am looking at you Arsenal). There is also the little matter of what to do about Tottenham.


Week 15

With Week 14’s mid-week games packed in between two big weekends of fixtures it will be interesting to see how certain managers deal with substitutions, possibly opting to keep certain big players rested for later. This week brought a lot of changes, Everton seemed to find their footing against a decaying West Ham side, could this be Big Sam’s doing? Leicester show they are a side to be reckoned with after shutting down a flat Tottenham. There are no changes to the algorithm this week, mostly because I didn’t have time.


Week 14

Week 13 was an exciting foray into all the best bits of the Premier League. Arsenal getting a win they may have not deserved. Chelsea swiping a point from Liverpool (like a beloved children’s show fox) with what many are calling a *crot from Willian and Tottenham continue to falter as they fail to put away West Brom at home. This week I made mostly aesthetic changes moving the goals and results to the middle of the table which will be how they appear from now on. My prediction model remains unchanged this week.

*a wayward cross that somehow ends up on target, sometimes in the back of the net


Week 13

Last week was one of the most exciting week of games this season. Each game had you on the edge of your seat and anxiously sipping your Guinness. I narrowly lost my fantasy matchup but it was a good week for the prediction results. For a gooner the Arsenal result seemed to look like a return to some confidence in the Arsenal squad. I think that the addition of more weight on the last five game performances may be the only change I make going into Week 13.


Week 12

Week 11 was a prime tribute to the fragility of team positions mid and low in the table. With teams like Bournemouth, Brighton and Burnley all picking up low percentage chance wins. Though I did get the Brighton result right I believe I still have quite a bit of tweaking to do with market values but I left the overall calculations unchanged for this week.


Week 11

Last week shook out okay, though my only disappointment had to be the Watford Stoke result. This week I only use the last five games plus market value with offense weighted differently to see how that works out.


Week 10

The last changes I made to the market cap weight seem to have made a difference. I think I may try and establish a model based on last five game performance for Week 12-14 to see if that does any better. Maybe it could predict things like Manchester United losing to Huddersfield (17/2 odds Huddersfield winning).


Week 9

Week 8 saw the introduction of the ‘market cap’ value to the predictions. This made the bigger teams (Chelsea, Arsenal, etc.) with the larger purses have wider margins, and fall harder. My predictions included some lofty goal scoring and high rated offenses which then fell to the supposedly lower rated teams. Chelsea falling to Crystal Palace, Arsenal falling to Watford, it was all a mess. I made some changes going into this week to lessen the impact of the market values and add a heavire weight to the results from this year.


Week 8

I am getting behind on these but I was travelling again last weekend. This week is the debut of the first market cap feature addition. It seems to be a fairly cut and dry week with fixtures so may have to wait a couple weeks to see if accuracy changes.


Week 7

Sorry for the hiatus… back to the results! It was very interesting to see a wild week like Week 5, which was tough to forecast for many on the web, where I went 4-6 on results. Follow that up with Week 6, where it was positive 8-2. The EPL is a fickle thing. This Week 7 is the last week slated to just be adjusting values within the calculations of performance variables. After this week I will only be making changes and additions of new data such as team market cap, foreign league performance etc.


Week 6

Hiatus week #2 (left results unchanged)


Week 5

Hiatus week #1 (left results unchanged)


Week 4

It was a rough Week 3 for predictions. As I think about what has worked so far this season it seems that lessening the head to head history factor and allowing the current season performance to weigh heavier in the model.


Week 3

Minor improvements to the way goals are calculated and compared mde the goal predictions closer, this week will hopefully improve on that. Another big team gives us an upset for the second week in a row with Arsenal losing 0-1 to Stoke. This week the changes are just incorporating last week’s results, leaving the goal projections raw, and next week I plan on adding a revised head-to-head factor.


Week 2

What an exciting Week 1. There were a couple of upsets, with Chelsea falling to Burnley 2-3, and Liverpool drawing with Watford. The newly promoted teams madea good showing with Crystal Palace falling to Huddersfield 0-3. My adjustments are not extensive this week, just including the first week’s results, and lessening the promoted team handicap.


Week 1

In this first week I left the goal predictions as-is with no adjustments for draws or by any sort of rank strength. I want this first week to start off as simple as possible and go from there.