Friday, 30 October 2015

Shooting for the Top.

Ten matches into the Premiership season and the league table is beginning to show some kind of structure.

Arsenal and both Manchester clubs are stretching their legs at the top of the table as they pull slightly away from a host of mid table aspirants, any of which could be drawn into a contest for either survival or Europa League qualification.

Towards the foot of the table recent struggling teams, such as Villa and Sunderland are drifting slowly away from the rump of the table, along with the three promoted teams, although the season is still young enough for a couple of wins to transform despair into mere mediocrity.

Ten matches is an interesting sample size with which to draw conclusions. It is large enough for the points already won or not won to hang heavy on the remainder of the season.

A side which has grabbed points through early goals and a fierce rearguard, such as West Ham or salvaged points from losing positions, such as Leicester, may not repeat these atypical feats. But the points are already banked should winter prove less productive.

A team has also played 10 of their 19 opponents. So the temptation to pass judgement on the well being of a team has to be tempered with the possible inequality in schedules which may currently exist.

Everton share 11th spot with Swansea on 13 points with an identical goal difference and the toffees shading it by scoring one more goal. But Everton's opponents so far have a combined goal difference of +33 compared to -17 for Swansea's 10 opponents to date.

This disparity is also seen if we look at a commonly used precursor to goals, shots or more specifically shots that required a save.

Swansea has a small positive shot on target differential of +5 through ten games, while Everton is -9.

However, Everton's opponents again have a net, positive differential value of 115 shots on target, while Swansea's overall have been less productive in the attacking half and more generous defensively, with a combined negative differential of -61.

It would appear that whether measured in goals or accurate shots or headers, Everton has faced more taxing opposition than Swansea and a mere cursory glance at the respective shot differentials of each team may, misleadingly elevate Swansea (+5) above Everton (-9).

If you run a least squares analysis, which relates a team's shot differential to the shot differentials recorded by their opponents to date, Everton's more difficult schedule now raises them above the Swans.

Using Sunderland, the side with the worst schedule adjusted attempts on target differential as a baseline figure of zero, Everton has a shot differential of nearly four compared to 3.5 for Swansea.

So should Everton meet Swansea at a neutral venue, then based on schedule adjusted attempts on target, the Merseyside team might be expected to out shoot on target their Welsh opponents by an average of half a shot. And given the relationship between attempts on target and goals, should be slightly favoured to win the game.

Attempts on target are more plentiful and form the basis for many varied models with which to predict future performance, ranging from mere counting models to shot location ones, but at various stages of the season strength of schedule will distort the raw data.

Chelsea, I guess they've got what they deserved.
Manchester City are particularly solid whether viewed in terms of SOS adjusted attempts or goals. They rank 2.3 standard deviations above the average in the former and 2.2 above in the latter, ranking currently first overall in both.

West Ham rank a lowly 11th when SOS adjusted attempts are used to define quality, well below their current position of 3rd, so even with points in the bank, their "improvement" may be partly an illusion.

Currently high flying Leicester are more likely a mid table side based on both adjusted shots and goals and WBA won't be the first relegation quality team to have found themselves in the top ten after 10 matches.

Worryingly for Mourinho, Chelsea has so far churned out and allowed attempts and goals that are broadly consistent with their current league position.

Thursday, 29 October 2015

The Penalty Shootout Kings are Stoke City.....and Chelsea.......and Kidderminster.

Stoke City moved imperiously into the quarter finals of the only major trophy they have lifted in their 152 year history.

The Capital One Cup is a butchered version of the prestigious trophy the Potters lifted back in 1972 and with the delayed entry of the more preoccupied Premiership heavyweights and a truncated format, Stoke have reached the last eight by winning just once in open play.

Their two successful penalty shootout victories have not only set up a home tie with Sheffield Wednesday, it has also projected them to a respectable 50% success rate in penalty kick tiebreakers.

Meanwhile defeated visitors, Chelsea now have breathing space to consolidate their position in mid table, while seeing their success in shootouts fall to 40%.

Of course percentages are largely meaningless, obscuring more instructive information, such as sample size. Chelsea has won one more shootout than Stoke, 7 compared to 8, but Stoke has had 6 fewer opportunities, 14 compared to 20 and has in raw terms been more efficient.

Therefore it is understandably that the Daily Mail was spoiled for potential headlines when they presented Opta's breakdown of historic penalty shootouts for England and Wales' club sides. Eventually plumping for Liverpool as the most efficient as well as the most numerically superior of the current Premiership sides. (Another imaginary feather in Klopp's cap).

The article does an admirable job of presenting fascinating trivia, especially following a week where so many matches were decided from 12 yards.

Records are presented as just that, with percentages invariable going hand in hand with sample size and only once does the article over stretch when attempting to equate success rate to raw ability. Kidderminster (0-3) are declared worse than Newcastle (1-9).

Jose decides who will spin the coin for him at Stoke. (I'd perhaps give Kenedy a miss, Jose)
Penalty shootouts are often declared a lottery, although with success rates for an individual kick at around 80% rather than 1 in 14 million, a coin toss would be a better analogy.

An individual penalty kick requires the application of the most basic of footballing talent. To propel the ball forward twelve or more yards with as much power or guile before the keeper (Begovic) can remember which is your favoured side.

Therefore, while open play demands a much more varied array of talents, as well as combined team inter play, penalties are more likely to see a similar level of ability from those taking or attempting to save each attempt.

With luck and talent being ever present contributors to a sporting outcome, when talent is likely to be of a very similar and high level, as it may be in a penalty shootout, then luck will become the major factor in success or failure.

The spread of successes in raw and percentage terms, will range from the bad/unlucky of Newcastle and Kidderminster to the good/lucky of Brentford and Liverpool. Kiddy's 0 from 3 has around a 13% chance of happening by (bad) luck to a side that has taken part in three 50/50 contests. Newcastle's 1 from 9 is around a 2% chance if they too had played out nine evenly matched contests.

However, Opta's penalty shootout roll call contains 133 different sides, which increases the chances of finding a side as "bad" as Newcastle appear just by chance.

If we look at the spread of outcomes for all 133 teams, taking into account number of shootouts and efficiency we can find how much we need to regress the actual record of each team towards the group average success rate of 50%.

And with the data available so far you should regress all the way to 50%. Newcastle's 11% success rate from 9 attempts likely hides an underlying talent level that is very close to 50%. The same as Brentford who has won 14 from 23 at 61% ........ and every other team who has ever taken a penalty shootout in anger.