Statistics for Managers: A Continuum of AVB

Most fans did not know what to expect when Tottenham Hotspur sacked Andre Villas-Boas in December and replaced him with Tim Sherwood. What we know now is that the two are radically different managers, tactically. Villas-Boas' high-pressing system has been replaced by a more open defensive system, albeit one that still works a high defensive line at times. Sherwood has changed personnel, bringing Emmanuel Adebayor and Christian Eriksen back in from the cold. He has established his own version of the 4-4-2 formation which, while it should be distinguished from the "caveman" 4-4-2 it is often assocated with, nonetheless differs significantly from Villas-Boas' preferred 4-2-3-1.

My question is, does this show up in the statistics? How does it show up, statistically?

The statistics I'm most interested in are shooting statistics. Obviously managers can have effects that are felt statistically in various ways, from possession percentage to pass completion to offsides calls won. But since I've been working with shooting data and collected a nice database on shots taken and conceded, I'm interesting particularly in how the tactical and personnel choices of managers affect the number and type of shots their clubs attempt.

I introduced these statistics in my piece on manager sackings last week. The numbers I'm looking at are pretty simple. First, I'm considering the simple question of shot location. How many shots is the club taking from the "danger zone" in the center of the eighteen-yard box (zones 1-3 in the map on the right)? How many shots is the club taking from the wide areas of the box? How many from outside the box? I have found that almost all of the highest-quality chances are attempts from the danger zone, so I'm also looking not just as shot totals but also at the percentage of total shots taken from the danger zone.

I consider these numbers for both attack and defense.

Further, how a shot is assisted makes a major difference in its expected rate of conversion. Shots assisted by crosses are generally much more difficult to place on goal than shots assisted by other kinds of passes. Shots assisted by through-balls, on the other hand, are far better than normal shots taken from the same area. I mentioned that "almost all" of the best shots are taken from the danger zone—the exceptions are shots off through-balls from the wide areas of the box (zones 4 and 5), which are roughly as good as normal shots from the center of the box.

So I add two more stats, Cross% and Through-Ball%. Since the vast majority of shots off crosses are taken from the danger zone, Cross% is the percentage of shots from the danger zone assisted by crosses. TB% is the percentage of shots from zones 3-5 assisted by through-balls.

Finally, I have average shot quality. This is a rating of the average expected goals per shot.

Chelsea Before and After AVB

So here's the data. I list all the totals in per-match rates, to even out managers who have had more or less time as manager at Stamford Bridge. Here's the raw numbers for both attack and defense.

Attack	xG	DZ SiB	DZ SiBoT	W SiB	W SiBoT	SoB	SoBoT		%Cross	%TB	%DZ	Shot Qual
Ancelotti	1.79	7.58	2.92	3.29	1.13	8.53	2.37		49.0%	9.3%	39.1%	0.093
Villas-Boas	1.51	5.78	2.59	3.44	1.33	8.85	2.44		57.1%	9.7%	32.0%	0.078
Di Matteo	1.42	5.87	2.57	2.17	0.78	7.35	1.83		44.4%	8.9%	38.1%	0.089
Benítez	1.46	6.27	2.50	2.54	0.96	8.08	1.77		33.1%	5.2%	37.1%	0.087
Mourinho	1.47	6.00	2.30	2.45	0.90	7.10	2.20		47.5%	3.5%	38.6%	0.092

The first thing that stands out, I think, is just how superior the 2010-2011 Chelsea attack under Carlo Ancelotti was to those that  have come after it. Obviously personnel makes a big difference here, and Villas-Boas was originally brought in to manage the transition to a new, younger squad. But that is a huge drop in attacking production in just a single season.

You also can note the rate of shots taken from the danger zone and shot quality. AVB stands out significantly here. This will be a continued theme with Villas-Boas as we look at his time with Spurs. He does not care much for maximizing the number of shots taken from dangerous areas, and he likes his teams to shoot freely from deep areas. This leads to his Chelsea taking more shots from outside the box than Ancelotti's, even while average two fewer danger zone shots per match.

AVB's very high rate of shots off crosses is peculiar to me, and it did not continue at Spurs. The more striking number here is Rafael Benítez' extremely low percentage of shots assisted by crosses. He had his team playing through the middle to a much greater degree than did Di Matteo before him or Mourinho after him.

To give some context to these numbers, I've also got league-adjusted numbers. These are in the style of the Team+, Att+ and Def- numbers I use in my power rankings. If I have a manager with 110 under DZ SiB, that means his club took about 10% more shots from the danger zone than the average side in the league. If I have a a manager with 76 %Cross, that means his club took about 24% fewer of their shots from the danger zone off crosses than the usual side.

Attack	xG	DZ SiB	DZ SiBoT	W SiB	W SiBoT	SoB	SoBoT		%Cross	%TB	%DZ	Shot Qual
Ancelotti	142	136	139	131	137	134	143		103	122	101	104
Villas-Boas	122	108	123	138	154	139	146		121	142	85	90
Di Matteo	115	110	120	89	90	118	114		97	132	100	102
Benitez	119	118	116	107	109	133	114		74	80	96	99
Mourinho	123	116	115	110	105	109	124		100	63	104	108

These numbers emphasize again just how much better Ancelotti's attack was compared to those who followed, as well as AVB's preference for high volumes of longer, lower-quality shots. The huge drop in shots off through-balls with Benitez' take-over and continuing through Mourinho, that I do not have an explanation for. It's interesting.

And now here are the same numbers, first raw and then league-adjusted, for defense under these five managers.

Defense	xG	DZ SiB	DZ SiBoT	W SiB	W SiBoT	SoB	SoBoT		%Cross	%TB	%DZ	Shot Qual
Ancelotti	0.83	3.87	1.42	1.82	0.55	5.87	1.55		52.4%	6.2%	33.5%	0.073
Villas-Boas	0.93	3.81	1.56	2.26	0.78	4.81	1.44		52.4%	5.6%	35.0%	0.082
Di Matteo	1.11	4.43	1.91	2.65	0.70	5.96	1.65		52.0%	5.0%	34.0%	0.083
Benitez	0.97	4.08	1.65	2.58	1.19	5.31	1.62		50.0%	11.2%	34.1%	0.079
Mourinho	0.90	4.10	1.70	1.45	0.75	4.30	0.70		46.3%	5.1%	41.6%	0.090

Since managers can't directly tell opposition players when, where and how to shoot, you find fewer clear effects in the defensive numbers. It is again clear how much better Chelsea's defense was under Ancelotti. Di Matteo's reign stands out, in the league at least, as the weakest period of Chelsea's recent run. Benítez maintained a similar attacking quality while shoring up the defense, and these numbers suggest his interventions were critical in Chelsea achieving their third-place finish last year.

Defense	xG	DZ SiB	DZ SiBoT	W SiB	W SiBoT	SoB	SoBoT		%Cross	%TB	%DZ	Shot Qual
Ancelotti	66	69	68	72	67	92	94		110	81	87	82
Villas-Boas	75	71	74	91	90	76	86		111	82	93	94
Di Matteo	90	83	90	109	80	96	103		113	75	89	95
Benitez	79	77	77	108	135	87	104		112	170	88	89
Mourinho	76	80	85	65	87	66	39		97	91	112	105

The other thing that stands out, here, is how Benítez and Mourinho have achieved very similar defensive results through very different methods. Benítez' clubs allowed opponents to take more shots from low-quality positions, but left themselves open to crosses and through-balls. Mourinho's defense has limited shots from outside the area to exceptionally low rates, but allows relatively more good shots from central areas.

I had expected to find AVB's Chelsea allowing tons of shots off through-balls, given the complaints about his high line. We do see such an effect with Spurs. I don't know if the issue here is simply a sample sample size (shots off through-balls are pretty rare) or if AVB was making adjustments to prevent through-balls.

Tottenham Hotspur Managers and Statistics

I want to make one thing clear before presenting these tables. Tim Sherwood has only been manager for four matches. Just about anything can happen in four matches. At one point last season Reading took ten points from four matches.

But, that said, Spurs have not been good overall under Sherwood's reign. They played Manchester United and Southampton roughly even away, but took six points. They deserved to lose to West Bromwich Albion and hung on for  a point. The Stoke City match rates as a high-quality stomping, but that's just one out of four. If Sherwood's Spurs are going to continue taking points at anything near this rate, they will need to play a lot better.

But starting with attacking numbers. I'll just post both tables, raw and league-adjusted, back-to-back. For Harry Redknapp, I'm just listing his club's 2011-2012 numbers, not his numbers from the previous seasons.

Attack	xG	DZ SiB	DZ SiBoT	W SiB	W SiBoT	SoB	SoBoT		%Cross	%TB	%DZ	Shot Qual
Redknapp	1.58	6.50	2.45	3.11	1.00	8.71	2.55		39.7%	9.6%	35.5%	0.086
Villas-Boas	1.43	5.28	2.06	2.83	1.17	9.61	3.04		42.8%	8.7%	29.8%	0.078
Sherwood	1.19	5.25	1.75	3.50	0.75	6.50	1.50		52.4%	3.2%	34.4%	0.089

Attack	xG	DZ SiB	DZ SiBoT	W SiB	W SiBoT	SoB	SoBoT		%Cross	%TB	%DZ	Shot Qual
Redknapp	127	121	116	125	116	137	153		84	140	94	99
Villas-Boas	117	100	97	121	133	155	188		93	143	78	89
Sherwood	99	102	87	157	87	100	84		110	58	93	105

You can see here again AVB's predilection for long shots. Even taking over a club which had generally not focused on emphasizing shot quality, he still cut the club's percentage on shots from the danger zone from 5% below league average to 20% below. Shot quality dropped from about average to 10% below. You can see also some of the progressivity of Redknapp's system. Running around a bit led to relatively few shots off crosses and quite a few through-balls.

So far, with Sherwood, we see a somewhat different profile. He's got the club taking more shots from better areas, though they can't seem to put any on target. The "cross" number may be inflated by the insane stats of the West Bromwich Albion match, and could go down over time. Really, all of Sherwood's numbers here should be taken as a snapshot rather than as a projection. The more interesting data is on Redknapp and AVB, for whom we have something approaching a useful sample size.

And defense:

Defense	xG	DZ SiB	DZ SiBoT	W SiB	W SiBoT	SoB	SoBoT		%Cross	%TB	%DZ	Shot Qual
Redknapp	1.12	4.89	1.84	1.82	0.58	5.47	1.74		46.2%	6.3%	40.2%	0.093
Villas-Boas	0.95	4.09	1.85	1.80	0.76	4.07	0.96		44.3%	10.8%	41.1%	0.092
Sherwood	1.21	4.25	2.00	1.50	1.00	3.25	1.00		41.2%	16.7%	47.2%	0.126

Defense	xG	DZ SiB	DZ SiBoT	W SiB	W SiBoT	SoB	SoBoT		%Cross	%TB	%DZ	Shot Qual
Redknapp	90	91	88	73	67	86	104		98	92	107	107
Villas-Boas	78	78	88	77	87	66	60		96	177	108	105
Sherwood	101	82	100	67	116	50	56		86	299	127	148

I think it's also useful to compare the xG numbers for AVB and Redknapp. They come out with nearly identical xG differences (+0.46 and +0.48), but Redknapp's numbers came from excellent attacking and good defending, while AVB's reversed that formula.

You can also see the effect of Villas-Boas' high pressing and high line in great relief here. AVB's clubs prevented shots of all types with their defensive pressure, but they could be cut open with a well-timed through-pass much more often than the typical club. The balance added up to an excellent defensive performance, but one shaped quite differently from those seen in the above Chelsea stats.

I argued after AVB's sacking that the breakdown in the club's defense in the last few months may have presaged his fall. So far with Sherwood we've seen only worse defense, and one even more open to giving up high-quality shots in the danger zone. Again, one hopes that this is improving, that it's only been four matches, but the early returns are less than encouraging.