Data and Modelling in the Work of the Goalkeepers

I developed two models. The pre-shot model (Model 0) was meant to be comparable to Paul’s model and used location (shot xy coordinates) and contextual features. The post-shot model (Model 1) used location, contextual and placement features (Ball height and distance from the goal centre).

The features in each model are shown below:

Feature Notes Pre-shot model features Post-shot model features

Big Chance		Y	Y
Shot Angle	Angle subtended by ball and goal posts	Y	Y
Shot-x	x distance from shot location to goal line	Y	Y
Shot-y	Absolute y distance between shot location and pitch mid-line	Y	Y
Headed		Y	Y
Assist	Intentional assist	Y	Y
Regular Play		Y	Y
Direct Freekick		Y	Y
From Corner		Y	Y
Fast Break		Y	Y
Set Piece		Y	Y
Penalty	All penalties are Big Chances	Y	Y
Ball height	Ball height at goal mouth	N	Y
Ball height squared		N	Y
Goalmouth-y	Absolute y distance between ball and goal centre	N	Y
Ball height*Goalmouth-y	Interaction term	N	Y

I used nine seasons of EPL data (2010 to 2018) with a 70-30 train-test split to develop the pre-shot model. I used an xGBoost classifier and optimised the parameters using a grid search. The MSE for the pre-shot model in the test set was 0.16 and the AUC was 78%. For the post-shot model, the MSE was 0.13 and the AUC was 87%, a considerable improvement. Sometimes I get tired of numbers and betting, when that happens you can play casino games without a Gamstop.

Results

Keeper performance was assessed in the 2017 and 2018 seasons. This was slightly different to Paul, who included data from the 2019 season-to-date, but as I did not have that data to hand, I did not use it. The table below shows the results. The columns labelled “Ratings” are the ratios of the expected to actual goals conceded; a keeper with a ratio above one concedes fewer goals than predicted by the relevant xG model, and a keeper with a ratio below one concedes more.

Rank Preshot ModelRank Postshot ModelRank PRKeeperShots OT FacedGoals ConcededxG Conceded Preshot ModelxG Conceded Postshot ModelRating Preshot ModelRating Postshot ModelRating PR

1	2	1	Alisson	96	22	29	27	1.34	1.24	1.38
2	3	4	David de Gea	310	77	97	92	1.25	1.20	1.17
3	1	11	Nick Pope	145	34	42	43	1.25	1.25	1.05
4	6	7	Bernd Leno	144	41	49	44	1.19	1.07	1.11
5	4	2	Lukasz Fabianski	388	108	128	123	1.19	1.14	1.18
6	5	3	Hugo Lloris	247	65	75	72	1.16	1.10	1.17
7	8	5	Martin Dubravka	178	58	64	61	1.11	1.05	1.13
8	9	8	Jack Butland	201	59	63	61	1.07	1.03	1.06
9	11	13	Tom Heaton	98	28	29	29	1.05	1.02	1.01
10	14	9	Ederson	158	47	49	47	1.04	1.00	1.06
11	16	14	Sergio Rico	165	54	56	53	1.04	0.97	1.00
12	13	18	Jordan Pickford	314	102	105	103	1.03	1.01	0.99
13	7	10	Mat Ryan	324	104	107	110	1.03	1.06	1.05
14	12	17	Wayne Hennessey	194	63	64	64	1.02	1.02	0.99
15	22	12	Petr Cech	172	56	57	52	1.02	0.93	1.01
16	18	15	Ben Foster	331	109	109	105	1.00	0.97	1.00
17	24	16	Rui Patricio	136	42	42	39	1.00	0.92	1.00
18	10	23	Kepa Arrizabalaga	121	39	38	40	0.97	1.02	0.89
19	17	20	Neil Etheridge	207	69	66	67	0.96	0.97	0.93
20	23	19	Adrian	96	28	26	26	0.95	0.93	0.94
21	21	25	Asmir Begovic	275	105	99	99	0.94	0.94	0.88
22	15	–	Thibaut Courtois	106	32	30	32	0.94	1.00	–
23	19	22	Alex McCarthy	187	67	62	64	0.93	0.95	0.92
24	25	24	Jonas Lossl	297	111	101	98	0.91	0.88	0.88
25	26	21	Kasper Schmeichel	273	92	84	79	0.91	0.86	0.92
26	28	–	Fraser Forster	103	33	30	27	0.90	0.82	–
27	20	27	Joe Hart	207	79	69	75	0.87	0.95	0.84
28	27	–	Heurelho Gomes	101	42	36	36	0.86	0.85	–

The correlation between Paul’s ratings and my pre-shot ratings is .92, which is quite high. The differences seem largely due to differences in the data sample, and small sample sizes. For example, Nick Pope turns out to be high in my rankings and only middling in Paul’s. Pope didn’t play in the EPL in 2018 due to a serious injury, so my ranking is based only on the 2017 season which isn’t really enough. In addition since returning from injury, he hasn’t yet been able to replicate his 2017 performance.

However, that is not important for our present purposes. The crucial point is how much my pre-shot and post-shot rankings differ.

We can see that when post-shot information is included, some keepers rise quite a lot in the rankings and some fall. Kepa for example looks a much better shot-stopper when shot placement is taken into account, while Petr Cech looks significantly worse.

In conclusion, when evaluating goalkeepers, I would always include post-shot information and if you need to place a bet and are a self-excluded punter, you can take advantage of offers from betting sites without Gamstop.