QPR's Poor Shot At Staying In The Premiership.

Despite the best efforts of seasoned Premiership managers, Mark Hughes and Harry Redknapp, QPR returned to the Championship at the end of the 2012/13 season following a brief, two season long return to the top flight. The seeds of their likely downfall were sown in the season's very first game, as an inability to score was compounded by a porous defence in a 5-0 home defeat to Swansea. Not the best of combinations and from those unpromising beginnings, relegation had become almost an inevitability by the time of their first league win in mid December.

QPR's season ending goal difference per game reached an eye watering -0.79 of a goal, sandwiching the side's performance between the historical averages of -0.75 for the 19th placed Premiership team and -0.95 for the bottom team. The -0.95 average is distorted by Derby's nightmare year, so in the absence of any similarly abject rival, QPR finished the campaign in last position.

Their relegation season can be better understood if we look at the type of shots they were attempting and allowing, based on the shooting location and the likely goal expectation of each attempt during 2012/13. In the table below, I've examined every shot from all of QPR's 38 league matches and calculated the likelihood that an average league side would score, miss the target or see their efforts blocked or saved. Shots have then been sorted into goal expectancy ranges and plotted for frequency. Alternate conversion rate bands have been omitted from the labeling for greater clarity.

Rangers, unsurprisingly were out shot by opponents in their relegation year by a ratio of around 1.2 to one. Despite shooting less often than their immediate opponents, they comfortably dominate the region of the plot which has the smallest expectation of resulting in a goal. Over 15% of the Hoops' shots had between a 1 and 2% chance of a resulting score. By comparison, their opponents attempted just 10% of their total attempts in this low probability, usually long range region of the shooting landscape.

The cumulative goal expectation for the 78 QPR shots with individual expected success rates between 1 and 2% was a paltry return of just over one goal and that is just the return they achieved. Low reward, long shots do not appear to be worth the almost inevitable loss of possession that usually occurs. Just 16 of the 78 shots required a save, 42 missed the target completely, 19 were blocked and only WHU conceded a goal to Taarabt, although they still won the match.

There may be a slight residual value from the blocked efforts. Blocked shots usually manage to travel around a quarter of the distance to the goal before they are blocked, but fighting for possession around the penalty spot appears a poor return from conceding possession in an attacking position outside the opponent's box.

The true scale of how Rangers were out shot in 2012/13 becomes apparent as we move across the plot from left to right towards attempts that carry a higher chance of finding the net. Higher quality chances created in QPR matches with a 20% or more probability of success quickly become almost exclusive to their opponents. This carries on a trend that became apparent once we moved away from the area of speculative long distance shooting that Rangers appeared to specialize, if not excel in.

Overall, comparing every chance allowed or attempted, an average side would have expected to score 35 goals from the opportunities attempted by QPR. Rangers managed just 29 (they were gifted an own goal by Everton). Defensively they allowed 60 compared to an average expectation of just 56. So their raw attacking and defensive qualities may also appear to be below par or alternatively, they were both ambitious with their speculative shots and unlucky at both ends of the pitch.

The only real bright spot was that they managed to numerically outdo opponents in creating 78% conversion rated penalty kicks, but they then proceeded to miss three of the four they were awarded.

This type of analysis can be further developed. For example, it is unlikely that even if the shooting from distance obsessed Rangers had been particularly fortunate and their opponents much less so, they would have managed to avoid the drop. In almost every simulation their goal differential lies within the range of a bottom three Premiership side. The shooting profile, either forced on them or voluntarily undertaken, or more likely a combination of the two, gave them virtually no chance of survival.

We may also attempt to spot any tactical change with changing lineup or leadership, although this may be a sample slice too far.

Wide Right. How The Lions Hung On In Brisbane.

It is always tempting to isolate individual incidents and speculate on how a more favourable outcome would have changed the outcome of a sporting contest. A debatable disallowed goal or a goal prevented by a miraculous last gasp intervention can be retrospectively added to the ledger in football, but the conclusions are often unsatisfactory. Goals are relatively rare events in football and their arrival changes the game dynamic, so post game insertion of "what could have been", even if we can accurately assess how often a player might have successfully scored from an enticing position will inevitably leave us mired in subjective speculation.

The conundrum is less severe in rugby. Scoring is more commonplace, 20 points per side comprising a variety of tries, conversions, drop goals and penalty kicks, each worth different amounts of points is around the average expectation for a single side. And kicking, be it via a two point conversion or a three point penalty, provides an ideal modelling subject. In the midst of one of the ultimate games of team cooperation, kicking provides an isolated, repeatable challenge, where a single player pits his talent to a task while the opposition stands quietly in the wings.

Weather and ground conditions (or choice of footwear) are significant minor contributing factors, but the likely success rate at kicking a ball through two uprights can be relatively easily calculated with sufficient data. Such a model is described here.

The first Lions test took place on Saturday morning. A close match was expected and the British and Irish Lions were narrow three point favourites, making them a six point superior side to Australia if the game was taken to a neutral venue and superficially the pre-game opinion was accurate. The Lions won by two points.

The sides shared four tries and with try scoring models in rugby a long way in the future, we can speculate how a different distribution of successful and unsuccessful kicks might have changed the final outcome.

The Lions had Leigh Halfpenny as their kicker, statistically the world's best current kicker, both from distance and from more conventional areas of the pitch and in reserve they also had an equally above average talent in Jonathan Sexton. By contrast, Australia had riches in quantity, but not quality. Christian Leali'ifano was stretchered off after 37 seconds, James O'Connor kicked poorly until he was relieved of the kicking duties by Kurtley Beale, who came off the bench to attempt the last four of Australia's nine kicks.

In recent years, neither of Australia's preferred options have broken par for kicking expectation. While Halfpenny has kicked around 25% more kicks that an average kicker would achieve once kick position is accounted for, Beale, O'Connor and Leali'ifano are each about 10% below the return expected from an average kicker.

On the day, with tries shared, kicking was to decide the result and Australia's deficiency in kicking ability should have been more than compensated for by a combination of winning more total kicks, a number of which were relatively straightforward. Above I've simulated the net points outcome for each side for each of the 15 kicks that were awarded on Saturday (nine to the Wallabies and six to the Lions).

The most likely outcome was a nine point win in favour of the host side.

If we assume unrealistically that, regardless of kick outcome, 15 kicks would always have been awarded, we can see how "lucky" the Lions were to win, given the pitch position of the kicks and the relative qualities of the kickers. Hardly any of the kicking simulations see the Lions gaining more points than the Wallabies.

Beale, quite naturally attracted most of the headlines by missing two relatively simple kicks inside the final five minutes. Especially the penultimate attempt that he pushed wide right, when conversion rates of well in excess of 90% could be expected. But his kicking partner, O'Connor was equally profligate.

The Lions' chances of avoiding defeat in such a numerically lopsided kicking contest barley reaches 2% and as many have pointed out, it is to rugby's credit that supporters of both sides were sympathetic to Beale's misfortune.

A lucky Lions win ? Possibly, although as in football, it is easy to imagine a concerted Lions onslaught should they have found themselves trailing even as late as the 76th minute. Scores change every sporting contest. Beale's "wide right" moment and unfortunate slip at the death when attempting a 40+ yards attempt from a central area, made for a dramatic finale to a hugely entertaining match, but for the ultimate, potentially game winning, missed kick, revisit the unfortunate Don Fox in the last minute of a Wembley Challenge Cup final.

https://www.youtube.com/watch?v=eer30sfqgkk

The second test is on Saturday, 11-05 am UK time.

Scoring Efficiency and Game State.

An old post from a year ago that is available on the OptaPro site, but I forgot to ever link it here.

It uses Opta data to relate game state, expressed as a percentage of a side's initial points expectation to shooting type, efficiency and outcomes. It ties in well with some of the recent posts here based around shots, blocks and goals.

It can be found here.

http://www.optasportspro.com/en/about/optapro-blog/posts/2012/guest-blog-scoring-efficiency-and-current-score-by-mark-taylor.aspx

Evolution or Relegation.

It was hard to appreciate the unconventional potency of Rory Delap's long throw unless you witnessed it in the flesh and particularly early in Stoke's Premiership years. Low and flat, with little hang time, it presented a unique challenge to defences that were used to a more leisurely approach to repelling crosses. Keepers were especially hesitant and many found themselves rooted to their lines, attempting to keep out powerfully deflected headers that often originated from little more than the edge of their six yard box. Few teams enjoyed the attacking experience served up by Stoke, none more so than Arsene Wenger, who chose to call for an amendment to the throw in  law, rather than attempt to develop an effective method for defending against Delap's slingshots.

Of course, Delap was just a part of the Stoke experience, where possession was shunned, opponents harried without a loss of defensive shape and when Delap retreated to the bench, wide deliveries and set plays were employed to fill the void. In short, Stoke presented a unique challenge for established Premiership sides that were used to pitting their footballing skills against opponents with similar aims and ambitions. Some made heavy weather of adapting to the challenge.

Delap himself has gone on record as saying that although sides did struggle to cope with Stoke's style, especially his throws, adapt they did and the relatively successful strategy gradually became less so. However, Tony Pulis, the mastermind behind the Potters' survival strategy was a reluctant innovator and a considerable rump of his tactical creation remained a prominent part of Stoke's approach until club and manager finally cut the ties after five eventful seasons.

When Stoke slightly surprised themselves and many of their supporters by claiming automatic promotion in May 2008, they did so with a positive goal difference of just 13 or under three tenths of a goal a game. Sides that progress from the Championship to the Premiership, overall see their average goal difference per game slide by around a goal. So Stoke's projected goal difference in their inaugural Premiership campaign was a paltry minus 0.7 of a goal per game and that has traditionally been only good enough for penultimate place and relegation come the end of the season.

That they comfortably over achieved on that projected total by finishing 12th with an actual goal difference per game of -0.44 of a goal, in the process beating all market predictions, was in no small part down to the difficulty opponents experienced in trying to cope with Stoke's wildcard, the extreme long ball tactic, coupled with an aerial, set play bombardment.

Over the period of their Premiership tenure, Stoke have claimed 225 league points against bookmaker odds based model predictions of 216 and if we take out the first season, model predicted actual outcome accurately to within a couple of points. But does this mean that Delap's assertion that sides were gradually coming to terms with Stoke's style is undermined ?

The bookmakers odds for Stoke matches over the period offer a good proxy for a robust predictive model.
Small sample noise can often make a side appear to decline or improve when in fact little has changed, but all good odds setters realise this and so longterm reality usually sits well with longterm, actuality. However, it is also possibly that errors on one side of the ledger can balance errors at the other extreme to create an illusion of accuracy where rather the opposite exists.

In the graph above, I've plotted Stoke's under or over-performance against a bookmaker's model expressed in league points per game based on the number of times their opponents had gone through the unique experience provided by Pulis' side over the 190 EPL games played by the Potters. The general trend appears to support the realization of one player, if not the manager, that sides had increasingly worked Stoke out.

A group of teams meeting Stoke for the second or third time were allowing Stoke an average of three tenths of a point per game more than a robust, but general bookmaking model would predict. But by the time the mutual meetings had risen to ten Stoke were now the under performing outfit to the tune of nearly half a point a match. Familiarity, in the absence of much, if any tactical evolution from a Pulis led Stoke had bred success for their opponents.

Fortunately for Stoke's survival, promoted sides were constantly refilling the left hand side of the plot and continued over performance in that area was keeping Stoke afloat, although the margins for error were bound to become even tighter as old Premiership friends, such as West Ham, WBA and Newcastle began to rejoin mid plot.

Stoke include a rare innovation in their Premiership starting 11.

Anecdotal evidence from a player appears to be backed up by the data and hints at the need for teams to constantly evolve or decline. Many Stoke fans and one chairman realized during this season that the only two possible outcomes from a stagnating Stoke style was relegation or managerial change. Sides can sometimes poach a decisive advantage by utilizing such simple measures as better fitness levels, but they quickly drop back into the pack as others catch up and novel tactics carrying an element of surprise soon become stale as opposing coaches work out a solution.

The best managers never let their sides stand still for long.

Shot Blocking From Corner Kicks.

Corner kicks are often the life blood for those technically limited sides which use them to provide a rare chance to deliver quality balls into an admittedly crowded penalty box. Space and the ability to create it is a major factor in enhancing the likelihood that an opportunity will be converted and so the congested attacking area that predominates during a corner is less likely to result in clear cut chances being created compared to normal, open play, unless you've specifically built a side to exploit this niche. The largely unpunished defensive grappling of recent seasons during a corner kick only goes to reinforce the difficulties faced by would be goal scorers from this particular set play.

In previous recent posts I've looked at the often quite large differences in the conversion rates of chances that are created by various routes. Shots created by way of a quick counter attack appear to be easier to convert, once shooting position is accounted for, presumably because a stretched and under manned defence is less likely to be able to effectively close down both shooters and passers.

So if shots from counter attacks represent one end of the spectrum for defensive pressure, those attempts made in a crowded penalty area immediately following a corner kick are likely to be consistently attempted under the severest of opponent pressure.

Individual shots can of course be atypical of the general situation. For instance a player may through either luck or good instincts find himself totally unmarked even in the most crowded of areas, but in a reasonably large sample we should expect to see a conversion rate effect due to tighter marking when defenders have time to organize themselves.

As in previous posts I've used shots generated during usual open play as a benchmark figure and incorporated whether attempts were headers or by the boot before adding a variable to represent whether the chance came immediately following a corner.

As with most aspects of football analysis, there is an amount of subjectivity in defining goals scored from a corner. Andy Carroll's towering injury time winner for Liverpool at Blackburn in 2011/12, for example was scored just eight seconds after a Liverpool corner kick and much of the defensive chaos was caused by the presence of a handful of visiting defenders in the Blackburn box. But between the ball leaving the corner quadrant and hitting the net it was touched by a Blackburn player and even found it's way back to the halfway line. A goal from open play or a goal from  corner?

As with shots attempted on the counter, chances from corners do appear statistically different from those made in open play once position is factored in. A shot from the edge of the six yard box is over twice as likely to be converted in open play than during a corner kick. This discrepancy is maintained  as we move further out towards the penalty spot, a slightly greater than 20% conversion rate in open play drops to 10% in the rugby scrum following a corner kick.

Jostling for space and glory.

Similar figures are present for headed attempts. A player again can double his usual chances of heading in a corner kick if he or his team can engineer the type of unhindered aerial effort that is more typical of an attempt made from open play. A limpet like defender, more often with a handful of shirt, typical of the excesses allowed at corners, on average drops a would be scorer's rate from over 20% in the open to just above 10% from the sweet spot of the edge of the six yard box following a corner.

We can begin to see how defensive pressure at corners materially alters the anatomy of shot by looking at the frequency of blocks made in our two different scenarios. An average of around a quarter of all goal attempts are snuffed at nearly at source by defensive blocks. So blocks are a major component of shooting analysis. And as with conversion rates, blocking rates appear to be dependent upon the amount and proximity of bodies in  the box.

A player shooting from the edge of the six yard box can expect to suffer the disappointment of seeing 2 out of every 9 shots blocked by desperate opponents, but this rockets to 2 out of 5 following a corner. Headers follow the same pattern, 1 block in 7 for corners compared to 1 in 14 during open play from the same six yard area, displaying the general difficulty experienced by defenders when trying to block headed attempts.

The excitement often generated by corners is not because of an enhanced chance of converting any opportunity that is created, because on average it is a lower quality chance. But in the opportunity for defenders to showcase their game saving interventions by throwing themselves in front of what, from the viewpoint of shooting position would be considered an excellent scoring possibility in more familiar open play.

Pinball meets football.

Power and Placement.

Two exceptional goals graced the opening weekend of the Confederations Cup. First up was Barcelona bound, Neymar's stunning right foot strike from the edge of the box. The shot's upward slice taking the ball well out of the reach of the Japanese keeper, yet still retaining enough raw power to reach the target with the minimum of fuss. Pirlo, Italy's bearded midfield maestro then emulated Neymar's execution by dispatching a free kick from a similar distance into the Mexican keeper's top right hand corner of the net.

A striker friendly ball may have contributed to the keeper's discomfort, but both players also demonstrated the kind of technical excellence required to score in the top class, world arena.

Although the former goal came from open play and the latter from a set piece, both strikes had many shared characteristics. Both were shots originating from roughly the same position on the pitch, were struck with reasonable power and crossed the goal line at similar heights and positions relative the the bar and post. If the pace of either shot had been reduced or the placement was at a more manageable height for the keeper, then the chances of the keeper pulling off a save would most probably have increased.

The origin of a shot is undoubtedly a major factor in predicting the likely outcome of the opportunity, but factors such as placement and power are also going to be significant predictors of success. In this recent post I looked at the increased likelihood of scoring from identical field positions in general open play and then from a counter attack and similarly at home and on the road. Both these cases can be used as a proxy for differing levels of defensive pressure, but we can equally try to model for more obvious differences, such as the pace and placement so noteworthy in Neymar and Pirlo's most recent efforts.

Shot placement is increasingly available from various data resellers and shot strength is at the moment defined by a subjective assessment. By including power and placement in a rudimentary shooting model we my appear to be gaining more information about the shooter. However, initially such analysis may be more applicable to the keeper because a striker may appear lethally effective when shooting at the top corner, but attempts that miss high may outweight those rarer and more memorable successes.

To illustrate the possible impact of power and placement along with shooting origin on expected scoring rates
of individual shots I looked at the effect these three parameters had on actual outcomes, along with whether the attempt was by way of shot or header.

Power and Placement Effects for Attempts from Around the Penalty Spot.

Vertical Yards from Goal.	Horizontal Yards from Centre of Goal.	Pace of Shot/Header.	Type of Attempt.	Shot Placement.	Goal Expectancy.
12	0	Weak.	Header.	High.	8%
12	0	Weak.	Header.	Low.	5%
12	0	Normal.	Header.	High.	35%
12	0	Normal.	Header.	Low.	27%
12	0	Weak.	Shot.	High.	17%
12	0	Weak.	Shot.	Low.	12%
12	0	Normal.	Shot.	High.	55%
12	0	Normal.	Shot.	Low.	45%

The penalty spot is a useful constant to use as a universally recognised x,y coordinate and the table above shows how the chances of an on target effort being successfully converted alters with changing shot characteristics. Each of the five parameters were statistically significant indicators in my data set.

Although the usual proviso regarding relatively small sample sizes again applies, the wide variety of overall shot quality from identical x,y coordinates is well illustrated. A weakly hit, scuffed shot is much less likely to result in a goal compared to the type of well executed strikes we witnessed at the weekend. A shot placement to the upper area of the goal appears more difficult for a keeper to save, all other things held constant, but again a striker my waste opportunities by missing more frequently if attempting to grab an extra couple of % of goal expectancy by aiming high and missing completely.

As it was struck, Pirlo's effort would likely have scored twice every 13 attempts, if he'd merely floated it over the wall and on a similar target, the keeper would have probably saved over thirty attempts in between each concession.

Counter Culture.

One of the biggest problems associated with analyzing footballing performance is the overall lack of available data, both in quantity and quality. Goals are of course universal, but more granular data such as shots on target and conversion rates have only been available for recent seasons and if we move up a level to include x, y data, both collection and availability problems become even more severe.

The level of understanding that comes with increased detail in the data can be demonstrated by the usual case of Stoke City. Low frequency, but high conversion rates for their chances created over their initial tenure in the Premiership, would seem to be a perilous, luck driven and unsustainable way of staying in the top flight. But by looking at the positions close to the goal from where they converted a large proportion of their chances, especially in the early seasons, the Potters' continued survival just the right side of the 40 point barrier becomes slightly easier to rationalize. Without knowing that Stoke scored many of their goals in or around the six yard box, they would appear as an extremely lucky outlier or a team with outstanding finishers (if indeed such a player commonly exists).

However, paying too much heed to the unusual can obscure the perfectly acceptable work that can be done with the more basic data. Goals scored on the counter attack are assumed to fall into a similar niche category as the type of set piece strikes that have kept Stoke afloat over recent Premiership seasons. Teams of quality are widely perceived to cherish possession, leaving the lesser sides to contest their own glass ceiling, where goals are scrimped from less conventional means.

Goals from Counter Attacks for EPL Teams with 100 or more Open Play Goals. 2009-13.

Team.	Total Goals from Open Play.	Goals from Counter Attacks.	% of Open Play Goals from Counters.
Manchester City.	196	24	12.2
Arsenal.	230	28	12.2
Aston Villa.	124	15	12.1
WBA.	102	11	10.8
Tottenham.	188	20	10.6
Wigan.	106	11	10.4
Chelsea.	208	17	8.2
Everton.	140	11	7.9
Liverpool.	160	11	6.8
Newcastle United.	104	7	6.7
Manchester United.	230	15	6.5
Fulham.	116	7	6.0

Above I've outlined the percentage of open play goals that teams score on the counter attack over the previous four completed seasons. Using this basic analysis we can try to draw some conclusions that may be developed further. Firstly, there appears to be little correlation to finishing position. High achieving sides, such as Manchester City and Arsenal score a league high proportion of their open play goals from counter attacks, yet similarly talented Manchester United recorded a near league low percentage.

The spread of percentages ranging from a high of 12% to a low of 6% would imply that there is a clear difference in the ability of Premiership sides to convert on the counter. However, the disparity is unlikely to be as large as these raw numbers imply. Sample sizes are relatively small and a heavily regressed figure ranging from over 50% regression towards the mean for the larger samples to nearly 80% for the smaller ones, would seem in order. A more predictive figure would likely be highs of around 10% of total open play goals and a low of just over 8%.

Goals scored on the counter are by their very nature partly opportunistic, so match situation will likely play a considerable role in the regularity of their appearances. However, there does appear to be a cautious connection between squad age and the ability to create rapid fire counter attack scores. Fulham, currently have the Premiership's oldest outfield squad and perhaps significantly, they are the least reliant on counter attack scores from open play.

If we move from the basic counting stats for counter attack scores, we can begin to put more flesh onto the characteristics of such goals and also add weight or otherwise to the tentative conclusion from the basic data.

By cross referencing the category of goals scored to the x,y data we can try to see if the quality of chances created on the counter, when defensive set ups may be both less organised and less well manned, differ from "normal" goals scored during open play.

In the data set I have around 5% of open play chances are created on the counter. However, counter attack goals accounted for 10% of the goals scored, so the conversion rates of counter attacking chances appear greater than those for normal non set play opportunities. Conversion rates were 15% for the former and around half that for the latter.

If we use the x,y shot data matched to actual outcome, the ultimate origin of the chance does appear to be significant, with  opportunities created by way of counter attacks appearing statistically significantly easier to convert. To illustrate this, consider a penalty kick where the conversion rates in the EPL approaches 80%. If a shot is taken from the same spot from a counter attack, in my sample the expected conversion rate drops to just over 30% and if we project the expected success rate from none counter attacking open play scoring opportunities, the rate drops again to below 20% .

A quick Swindon counter catches out the Stoke defence.

So how you create a chance would appear to be a major factor in converting them successfully. Equally stark, at least in my sample, is the mode of execution. Around 10% of normal open play chances are executed with the head, yet chances taken from counter attacks were virtually exclusively shots with the feet. Similarly, defenders were rarely on the end of a counter, but they were on the end of nearly 10% of all other chances created from open play.

Small sample sizes often lead to extremes of results, but hopefully the overall trend remains and the size of the disparity between individual results can be tempered by regressing aggressively towards the mean. Counter attack goals we may tentatively conclude are slightly more likely to originate from young and presumably quicker players throughout the outfield team, they present clearer cut chances than other open play scoring opportunities from the same shot location and they seem to be more controlled moves because chances inevitably ended with a shot rather than a header.

10% of open play goals would seem to be the natural upper limit for sides to aspire to on the counter. So despite a high expected conversion rate, it would appear that attempting to build a side to exploit opponents on the break is the poor relation in niche tactical approaches, such as one based around set pieces, where the opportunities and therefore, goals are more numerous.

Is Shooting Easier On The Road?

There is a perception that a side adopts a more defensive outlook when playing away from home compared to games played at their home stadium. Certainly one slightly unfair barb aimed at Stoke's former boss, Tony Pulis was that he tended to setup with a packed defence, where the primary aim and duty of the majority of the team was not to concede a goal and very little team resources were committed to attack......and he was even more defensively minded on the road.

The general defensive outlook shown by visiting teams can be rationalized if we look at the average game state for home and away teams at the kick off. The average home side takes about 1.6 points from a match under three points for a win, compared to a shade over one point for the visitors. Therefore, on average at the start of the match, the one point currently owned by the visitors is relatively close to the longterm expectancy in terms of league points gained, but the hosts often start the match expecting more. In short, the visitors are happy with the current game state, the hosts, less so.

If away sides on the whole are more content to hold station for longer periods of a match, we should be able to see some of the effects of this more defensive approach in the more granular data, most notably shooting statistics.

X, y data that incorporates the field position from where the shot was taken, currently doesn't include any detail regarding the defensive pressure faced by the shooter. However, we should be able to spot the effects of this defensive pressure if we look at shots where this pressure is more likely to be present. If our assumption that visiting teams, at least initially set up in a defensive way and the average time for the first home goal to arrive is someway into the second half, we should expect visiting teams to be defensively orientation for longer than their hosts.

So I incorporated a variable for venue in my x,y coordinates, shooting model to see if the term was a significant factor in determining how likely a shot was to result in a shot on target, a block or a goal. Venue, it appears, does alter the likelihood of the outcome of a shot. Away from home compared to at home, shots from identical field positions are slightly more likely to result in goals, more likely to be on target and less likely to be blocked. Below I've listed the probabilities for each of the three outomes for a variety of shots from different positions on the pitch.

Venue Specific Shot Probabilities Based on Shot Location.

Venue.	Vertical Distance from Goal. (yards)	Horizontal Distance from Centre of Goal.	Chance of a Goal.%	Chance of a Block.%	Chance of Shot on Target.%
Home.	28	10	1.5	40	19
Away	28	10	1.8	35.5	23
Home.	12	12	5.9	24	31.5
Away.	12	12	6.8	21	36.5
Home.	8	0	21.4	20.2	38.7
Away.	8	0	24.3	17.3	44.2

The model is constructed around thousands of actual shot outcomes using the x,y data of each shot position and adding a venue variable, at least in this dataset, is a statistically significant addition, implying that scoring is easier from identical pitch positions for the travelling side. It is not much of a leap to consider the amount of defensive pressure, both in the build up and the execution is less onerous away from home.

Of course, shooting efficiency is just one factor that contributes towards actual scoring records. Home sides score more goals on average than away sides because circumstances may make shooting away from home slightly easier, but the home sides have more attempts. 55% of shots are made by the hosts.

Hitting The Target.

It is quite natural than the defensive aspects of football are always going to be eclipsed by the attacking side of the game. A goal represents a certainty and is, in the short term at least, mission accomplished. By contrast defensive actions, outside of the obvious saves made by goalkeepers and the much less common goal line clearances by defenders, impressive as they may be, merely prevent "what might have been". Even if the last ditch tackle fails, there is still the possibility that the striker will still miss the goal.

If we backtrack even further, a chance denied by rapid closing down or a well timed intervention well away from goal is even harder to record and quantify. As other sports have discovered, offensive production is quantified by events that occurred, but defensive ability is often a mixture of unseen and unrecorded actions that lead to things not happening.

Attacking prowess at the most basic level is recorded in goals scored, which in turn can be more elaborately investigated by more numerous connected events such as shots, attacking passes and more recently final third entries and gaining of possession in the opponents half.

Goals and shots allowed are the defensive equivalent of the attacking side of the ball's mainstream statistics and we can begin to unravel defensive metrics by thoroughly examining these readily available numbers. If we can begin to identify teams which are limiting their opponents scoring attempts we can perhaps then begin to try to see the actions they are taking to accomplish this.

I'll use shooting accuracy to try to illustrate the point. Shooting accuracy may represent a collective or individual ability to create "better" goal attempts, since for a goal to be scored a shot must first be on target. Therefore, if we see defenses which consistently cause opponents to shoot inaccurately, we may also be seeing evidence, which can then be identified, of effective defensive actions.

We can superficially see if shooting accuracy lies more with skilled sides by plotting raw shooting accuracy against seasonal success rates and the correlation is both reasonably strong and positive.

Football is of course a game requiring both attacking and defensive abilities. So the ability to shoot accurately (if it exists) is just part of the total package needed to consistently win matches. We shouldn't be surprised if the correlation with seasonal success isn't cast iron, because we are merely examining one side of the ball. Manchester City once finished a division one season as top scorers, yet were relegated, but in the modern era most teams recognise the merit of not neglecting one aspect of play to the detriment of another. Therefore, it seems reasonable to suggest that better all round teams are also on average able to fashion better shooting accuracy, either via better strikers, better creators or a combination of the two.

If we accept that shooting accuracy may be a skill, we can expand our investigation by looking at the ability of defences to cause opponents to shoot inaccurately. All shots aren't created equally at source, although in large enough sample this problem may be reduced. Therefore, I used my goal probability model, which uses x,y data to estimate the likelihood that a shot will strike the target, miss altogether or be blocked, based on the pitch position from where the shot originated.

Taken over all of the recent members of the EPL, the model produces a benchmark by which an individual shot can be measured. However, by introducing variables to represent individual sides, we can also see if the ability to make sides miss is a significant and consistent ability possessed by certain defences during recent Premiership history. Conversely, we can also identify sides which may be significantly poorer defensively and allow opponents to hit the target more frequently than is the general case.

Based on my dataset, Fulham appear as a side which consistently made hitting the target difficult for opponents compared to the league average. Using a position on the edge of the box and 5 yards to the left of centre as a bench mark figure, (roughly where Crouch is in the photo below) the generic average for the model was a 30% chance of hitting the target and the predicted average outcome for opponents of Fulham, based on the recent record of the London side was for them to be successful just 25% of the time. The recent Arsenal defence fared less well, consistently under performing by allowing slightly more on target shots than the average on all ranges of shots and, as an illustration, allowing opponents 36% accuracy for our 30% test shot.

Stoke demonstrate their below average  shooting accuracy.

To redress the balance with Gunners fans, their side hit the target consistently and significantly better than par at the other end of the pitch, jumping to 33% for the test shot, 8% ahead of underperforming Stoke. Scoring and ultimately providing a contribution towards season long success, it would seem may be a mixture of individual or team based accuracy in shooting, frequency of attempts and taking you shots from as advantageous a position as you can fashion.

From the plot above,shooting accuracy appears to be a trait of successful sides. Therefore preventing a side from hitting the target would seem to be a sensible aim, although correlation does not imply causation. Some sides, in this case Fulham appear to be statistically different in this skill than the average team. They made opponents miss more frequently in my dataset. So it might be instructive to examine game footage to see if they appear to be doing something either better or differently to other Premiership sides.

Hopefully,as models become more complex with the addition of extra layers of information (multi player position is the obvious next step), we can begin to reveal more accurate descriptions of the actions that go towards making teams successful, especially on the defensive side of the ball. Or at the very least we may know what we are looking for and which grounds we are likely to find it.

For other shot models, check out Different Game's wonderfully named SPAM model, a massive and on going project and the recent shot mapping from 11tegen11.

Leigh Halfpenny's Perfect Ten (+1).

The British and Irish Lions' tour of Australia is gradually grinding through the gears and they will hope for a more formidable test from their host's Super Rugby sides when they play the Queensland Reds on Saturday. The opening bout of the tour in Hong Kong against the Barbarians showed little more than Owen Farrell's ability to take a punch and still be able to deliver an improvised judo throw, while Western Force barely lived up to the second part of their name, even as 40 point underdogs.

Aside from highlighting the potential for serious injury that lurks in every professional game of rugby, Wednesday's first outing on actual Australian soil did provide a master class of kicking from arguably, the world game's current number one points kicker, Leigh Halfpenny.

Leigh Halfpenny, in Cardiff blue at the historic Ricoh Arena.

Halfpenny was presented with eleven kicking opportunities on Wednesday and duly kicked all of them. Six of the kicks were relatively straight forward, around 20 yards out and more or less central to the posts. Even allowing for the slight drizzle, these kicks are virtually automatic for a top class, international kicker. The real testers were the five conversions sandwiched in between these gimmes. Wider kicks for obvious reasons become much more testing, as the narrowing of the angles and the increased distance quickly takes average conversion rates for professional kickers steadily down towards coin toss territory. In addition a right footed kicker kicking from the right touchline finds landing such conversions very slightly more difficult than does a similarly talented lefty. Right footed Halfpenny was presented with three kicks from wide on the right hand touchline.

Leigh Halfpenny's 11 Kicks Against Western Force.

Kick Type.	Pitch Area.	Distance to Posts.	Average Probability of Success.
Penalty.	Central.	23 m.	94%
Conversion (Sexton try).	Central.	14 m.	97%
Conversion (O'Driscoll try).	Touchline.	42 m.	53%
Penalty.	Central.	17 m.	96%
Conversion (Croft try).	Touchline.	39 m.	61%
Conversion (Heaslip try).	Touchline.	39 m.	64%
Conversion (Vunipola try).	Touchline.	40 m.	59%
Conversion (Bowe try).	Touchline.	41 m.	55%
Conversion (O'Driscoll try).	Central.	15 m.	97%
Conversion (Farrell try).	Central.	14 m.	97%
Conversion (Parling try).	Central.	15 m.	95%

In the table above, I've listed each of Leigh's kicks, along with the likelihood that an average, first choice club and international kicker would be successful with each individual attempt. The probabilities are derived from the actual outcomes of 1,000's of such kicks over the last couple of seasons throughout world rugby.

I argued here that Halfpenny could be considered the world's best current kicker, but it was his excellence at distances in excess of 40 meters where he outshone all other current kickers. At distances of 40 meters or less he is still above average, but only just. Therefore, it is perhaps ironic that in a show of outstanding kicking ability, Halfpenny didn't get the chance to demonstrate that area of his talent where he really excels, namely from extreme distance. Halfpenny's figures indicate that he would have had around a 55% chance of converting BO'D first try compared to the league average 53% and each of the other ten attempts also show a small improvement against par.

 
The table above shows the likely frequency of the number of successful kicks made by firstly an average kicker and then by Leigh Halfpenny. The most common outcome, around a third of the time, would be for nine of the eleven kicks to be successfully kicked. A clean sweep from the eleven attempts is less likely with Halfpenny making such a sequence once every 18 attempts, whereas an average player would make all eleven once every 21 attempted sequences. So 11 from 11 is a reasonably unusual event, but it would have been much more newsworthy if he'd only kicked 5 successfully.

The kicking profiles of the Jonathan Sexton and Owen Farrell, two kickers who were also in the side on Wednesday are similar to Halfpenny's from 40 meters or less, so the Lions are well served with similar, high quality, normal range kickers. But if the requirement is there to kick points from halfway, good as Sexton is at this extended range, the Welshman will be the unanimous choice. Good as Wednesday's display was, Halfpenny hasn't yet had the opportunity to show just how good he really is as a kicker.

The Length of a Contest and the Favourite's Chances.

In this guest post from last week, I look at the impact of the length of a sporting contest on the chances of the favourite progressing or the under dog pulling off a shock. It is particularly relevant as the tennis season swings into full stride and match length increases from a maximum of three sets to five.

Roger gets a court named after him at the French Open.

photo courtesy of Stuart Smith.

Where The Talent Shone Through In 2012/13.

The close season is always a busy time of the year even if very little action takes place on the field of play. Managerial change, although often planned months before can often drag on for longer than anticipated and transfer deals that seem certain to take place and enhance next years title challenge can also flounder as other sides take an interest.

The end of the season is also the perfect time to assemble team and individual statistics. Team makeup has remained relatively stable and the ageing patterns that can impact on individual players over multiple seasons are largely absent.

This post isn't going to be a data dump, in deference to the many sources which have sprung up to provide a multitude of different types of mainly Premiership data. Nor will it try to make concrete connections between the numbers and various significant match outcomes. Instead I will try to firstly look at how efficiently EPL teams carried out certain tasks during 2012/13. Variation in efficiency numbers will inevitably occur in the relatively short space of a single season, even where there is very little difference in talent between all the sides. Random variation ensures that everything we attempt to measure is a combination of skill and luck, with the former beginning to overwhelm the latter only in much larger datasets.

The wider the actually distribution in efficiency stats compared to that expected from the league average, then the more likely that there are factors involved in the execution of these stats that differ between teams. In short some, teams are more adept at carrying out these actions than others.

I'll then list how strongly these actions correlate to the broad measure of success enjoyed by each side last season. Correlation doesn't naturally lead to causation (leading may cause teams to carry out tasks more efficiently), but knowing there is a correlation is a good starting point for further investigation.

Accurate passes in the attacking half.

Arsenal top the efficiency stats with success rates of nearly 78%, closely followed by both Manchester clubs. Reading prop up the table with 58% and the range of conversion rates on an average of over 10,000 passes per side strongly indicates different levels of team ability in this area of the pitch. Correlation to success in the final league placings is also strong, although Wigan are a notable outlier with top five efficiency recorded from above average frequency of attempts. Similar results are seen for the final third of the pitch, but with even stronger end of season correlation. Wigan slip to near mid table, with below average frequency when we move further up the pitch.

Accurate passes in own half. 

Efficiency now ranges from 93% success rate for Manchester United to 84% for relegated Reading and the frequency of attempted passes for United is twice that of Reading. Again there is a likely significant difference in the ability of teams to carry out such passes and correlation to league position is also strong. Wigan are again the best of the relegated sides, but now reside in the bottom half of the ranking.

"We pass better than you do".

Accurate Long Balls.

Last of the stats where there is likely to be a very large, real difference in ability to carry out such on field actions. Given the lack of love for long balls, there is a surprisingly strong correlation between the ability to play them well and league success. The top four in efficiency are the two Manchester clubs, Arsenal and Liverpool, Tottenham and Chelsea are 7th and 8th respectively. Two of the bottom four were relegated, QPR and Reading. Completion rates range from 67% to 43%, but lower overall attempts averaging just below 2,000 for the season, mean a little more regression towards the mean is needed to improve the likely accuracy of these raw figures.

Unsuccessful touches of the ball.

Ball control matters, or perhaps a team which gets ahead can make easier passes and bump up their efficiency numbers. The Manchester clubs and Arsenal have the least unsuccessful proportion of touches, just over 1%, Stoke have the highest percentage of miss controls at nearly double that rate. Correlation with league position is again strong and once again Reading and QPR occupy places in the bottom four and Wigan found themselves relegated, despite another mid table berth.

Successful Aerial Duels.

Still a relatively high chance that one team can be significantly better or worse than another in winning balls in the air, although the correlation to league success is by far the weakest on show so far. Stoke are both the most efficient (59%) and the most frequent of aerial combatants. Reading contested almost as many high balls as the Potters, but succeeded in a league low of 42% and were relegated. Hoofing it, but to no effect and then in turn getting hoofed on? Ground hogging Arsenal contested 800 less aerial challenges than Stoke, but were third in league efficiency.

Changes of possession. Winning possession in the attacking third.

The top six are all in the top eight of teams ranked by their efficiency at winning possession in the final third and the three relegated sides are all in the bottom six. The elite also have frequency rates in excess of three times those at the bottom of the list. Possession alone tells you little, but where and how often you win it seems to add a much needed context to the beleaguered raw measure.

Accurate through balls.

There was a significant difference in the ability shown by teams when picking a through ball, but that talent differential didn't help them win or more accurately, it didn't correlate to winning in 2012/13. Wigan and Liverpool were among the most efficient and Manchester United and Stoke shared time with the worst, leading to a virtually zero correlation between league position and through ball efficiency. It was raw accurate through ball totals where the strongest correlation with league success was seen. Persistence, rather than efficiency appeared to yield rewards.

Percentage of chances created from set plays.

League leaders, Stoke created almost 20% of their chances from set plays, nearly twice the rate even the most set piece reliant of the top four teams, Manchester City. There was a reasonably strong correlation between set play chance percentage and success in the league. Unfortunately for Stoke it was a negative one. Team's which overly relied on set plays to create a goodly proportion of their chances were competing in a mini league where survival was the primary reward and 40 points the major aim. Cloth was being cut accordingly?

Percentage of chances that are clear cut chances.

What constitutes a clear cut chance will naturally be subjective, but in general more successful sides turn chances into gilt edged ones at higher rates than less successful teams. Manchester United were the kings of the tap in, followed by most of the rest of the top six. Norwich were notable interlopers and while numerically their total chances created were no real match for the title contenders, their efficiency certainly was. (although the advantage may be scouted out of them in subsequent campaigns). The relegated sides were among the bottom eight for efficiently turning chances into clear cut opportunities and that usual cluster of poor sides were joined by Gareth "shoot from distance" Bale's Spurs.

Successful set play crosses.

Another category where the spread in efficiency percentages implied that teams possessed different skill levels. The correlation with end of season success was reasonable and negative, the less efficiently a side found a teammate with a set play cross, the higher up the table they tended to finish. Of course deliberately over hitting a free kick doesn't automatically gain you more points. The likely causation is that players who are better in the air tend to be less adept in other areas of the game and success in these areas possibly lead to more wins for the majority of teams. Football, as Cloughie said, should be predominately played on the ground?

Goal scoring efficiency per shot.

We are now getting into more traditional territory and conversion rates with a regressed 13% for Manchester United at the peak and 7% QPR at the bottom strongly suggests a real talent gap from best to worst. This is reinforced by an efficiency ranking that virtually follows the final league table. Stoke fell to second last in a category that they excelled at in their formative Premiership years, as transition to a more pleasing style proved difficult and only a strong defensive showing, especially pre January kept them afloat.

Percentage of tackles won.

Category topping Manchester City won 78% of their tackles compared to a Premiership low of 74% for Newcastle. Not a huge discrepancy, but with average team tackle attempts topping 2,700 still sufficient evidence to assume that tackling talent does exist in differing degrees for EPL teams. A weak, positive correlation with end of season finishing position possibly indicates that England's premier league is still partly paying due respect to one of it's traditional strengths. Arsenal were the second most efficient tackling side last term, but the presence of relegated QPR and Wigan in the top seven weakened the correlation with finishing position.

Categories where 2012/13's numbers didn't really provide evidence for a real difference in talent levels, included the rate at which clear cut chances were converted (most teams can hit a barn door at ten yards with similar competency) and, to a lesser degree and in a similar vein, shooting accuracy.

The above list isn't exhaustive, but it is sorted in order of the probable biggest disparity in talent levels. It is certain that some teams were much more adept at passing the ball in the attacking half than others in 2012/13 and those more talented sides also tended to be more successful at the end of the season. Tackling ability was much less diverse, that's not to say tackling isn't a talent, but the levels of talent are likely to be broadly similar across the league from a team perspective.

The correlations are presented to describe the broad attributes shown by successful and less successful sides, at worst they provide a crude descriptive measure of the kind of actions successful and unsuccessful teams were efficiently or inefficiently engaged in during 2012/13.

Causation isn't assumed and neither is the direction and teams which perform efficiently, actions which negatively correlate to success may find those talents are essential to their continued survival in the EPL, but the majority of other sides choose to take a different course. Some teams chose to hone a niche approach, to satisfy their limited ambitions?

Equality of opportunity also isn't guaranteed, especially where individual events for each team fail to reach 4 figures, but the efficiency distributions spread around the league average can still be used to reasonably assume the scale of the different talent levels between sides.