As part of a longer-term attempt to deconstruct expected goals into a variety of different, more granular and perhaps slightly more descriptive models, I’ve knocked together an expected save model today, and I thought I’d highlight some of the more interesting results out of it. Below is the data for this year’s Premier League, containing:

• Total shots on target
• Shots on target saved
• Goals
• Expected saves – the model’s prediction of how many SoT should have been kept out
• Saves above expected – how a keeper’s actual numbers compare to their expected numbers
• Difficulty – the average difficulty of shot the keeper faced (this is calculated as sum(1 - xs) / count(shots))
• Rating – simply saves over expected saves to make it easier to compare keepers

I’ve ordered by saves above expected because it’s a more in your face than the rating.

Season	Keeper	Shots	Saves	Goals	Expected Saves	Saves Above Expected	Average Difficulty	Rating
2015	Jack Butland	47	37	10	32.76	4.24	30.30%	112.94%
2015	Alex McCarthy	34	29	5	26.28	2.72	22.70%	110.34%
2015	Petr Cech	41	33	8	30.49	2.51	25.63%	108.23%
2015	Hugo Lloris	31	26	5	23.80	2.20	23.24%	109.26%
2015	Heurelho Gomes	38	28	10	25.94	2.06	31.74%	107.94%
2015	Adrián San Miguel del Castillo	30	22	8	20.65	1.35	31.17%	106.53%
2015	Tim Howard	43	32	11	30.97	1.03	27.98%	103.32%
2015	David de Gea	23	17	6	16.05	0.95	30.22%	105.92%
2015	Darren Randolph	12	8	4	7.43	0.57	38.08%	107.67%
2015	Sergio Romero	10	7	3	6.47	0.53	35.33%	108.24%
2015	Thibaut Courtois	21	14	7	13.77	0.23	34.45%	101.71%
2015	Michel Vorm	1	1	0	0.86	0.14	14.06%	116.36%
2015	Kelvin Davis	7	5	2	4.87	0.13	30.43%	102.67%
2015	Lukasz Fabianski	36	27	9	26.91	0.09	25.26%	100.35%
2015	Carl Jenkinson	5	3	2	3.01	-0.01	39.79%	99.65%
2015	Joe Hart	16	12	4	12.18	-0.18	23.86%	98.50%
2015	Adam Federici	11	6	5	6.53	-0.53	40.67%	91.93%
2015	Boaz Myhill	42	31	11	31.61	-0.61	24.73%	98.06%
2015	Robert Elliot	5	3	2	3.81	-0.81	23.82%	78.76%
2015	Simon Mignolet	30	20	10	20.95	-0.95	30.17%	95.47%
2015	Wayne Hennessey	8	5	3	6.04	-1.04	24.48%	82.76%
2015	Tim Krul	49	33	16	34.61	-1.61	29.36%	95.34%
2015	Willy Caballero	8	4	4	5.74	-1.74	28.25%	69.69%
2015	Artur Boruc	24	12	12	14.03	-2.03	41.52%	85.50%
2015	John Ruddy	45	25	20	27.10	-2.10	39.79%	92.26%
2015	Asmir Begovic	32	19	13	21.21	-2.21	33.71%	89.57%
2015	Kasper Schmeichel	42	25	17	27.49	-2.49	34.55%	90.94%
2015	Costel Pantilimon	52	33	19	35.76	-2.76	31.22%	92.27%
2015	Maarten Stekelenburg	20	9	11	12.41	-3.41	37.95%	72.53%
2015	Brad Guzan	46	31	15	34.74	-3.74	24.48%	89.24%

Some brief observations:

• It’ll be interesting to see who goes to Euro 2016 for England, Jack Butland and Alex McCarthy are both making a good case early in the season.
• That said, Alex McCarthy has faced the easiest shots on average of any keeper in the league (save Michel Vorm, who has had only one save to make).
• Hugo Lloris is performing above xS, but not so much that Tottenham’s 5 goals conceded is overly flattering. Lloris is another that’s right down there in the difficulty stakes, and it’ll be interesting to analyse over the coming weeks whether this is tame shot-making, or defensive organisation.
• The Brad Guzan vs Marrten Stekelenburg comparison at the bottom is fascinating – imagine if Southampton allowed as many shots as Aston Villa.

It’s early in the season, and saves are easier to make than goals (I’m not saying goalkeepers are the bassists of football, just that they save more than they let in, and strikers miss more than they score), so as you’d expect, the model matches reality fairly well so far. We can see this if plot expected saves versus saves – above the line is good, below is bad, further to the top right are the leakiest defences, bottom left are mostly backup, although Darren Randolph and Sergio Romero seem to have done fine when called upon this year.

I’ll be keeping this updated through the season and I’ll surface anything interesting I find in the historical data or across Europe. In the meantime, please enjoy the consistently inconsistent Tim Howard:

Season	Shots	Saves	Goals	xS	xSdiff	Difficulty	Rating
2010	141	97	44	99.48	-2.48	29.45%	97.51%
2011	133	94	39	93.39	0.61	29.78%	100.65%
2012	128	89	39	85.00	4.00	33.59%	104.71%
2013	152	115	37	109.83	5.17	27.74%	104.70%
2014	109	65	44	72.81	-7.81	33.20%	89.28%
2015	43	32	11	30.97	1.03	27.98%	103.32%