Turning Opinions Into Odds For Premier League Matches


How can we evaluate a team's true match winning potential? Today on the blog Mark Taylor delivers part two of his series on the task of setting odds for football matches.

Evaluating a footballer or team, be it as future opponents or as a potential purchase on the professional side of the sport or in trying to spot a wrongly priced set of odds on the weekend coupon, can take many forms. Traditionally, scouting a team in the flesh, using experience and accumulated knowledge held sway. But increasingly the use of figures is creeping into the mainstream, with illustrious names as Arsenal, Liverpool, Manchester City and Chelsea at the forefront of utilising high level statistical analysis to provide improved opinion and sides such as Everton appearing keen to dip their toe into the water.

Neither approach in isolation is likely to provide the complete answer. The pitfall of over rating, and therefore overvaluing recent performance seen in the raw, especially at high profile events, such as World Cups and the Euros, is now unlikely in the traditional approach, where extensive video archive can quickly identify fluke, atypical performances.

Similarly, an appreciation of the role of random variation, especially in small sample sizes, can help to curb any over optimistic reading of the exciting patterns seen in the data, where what you see isn’t a wholly accurate assessment of what you may get in the future.

So if an integrated approach, encompassing both disciplines, is the sensible way forward for the sport, it makes sense to try to use both methods before placing a bet. At the very least, using numbers to validate gut instinct or watching a side in the flesh or on screen to see how they are producing those spread sheet shredding numbers can only add to the sense of involvement in your betting.

In the opening post, we identified the expected goal supremacy between two sides in a contest as a powerful source of information. In short, it is an expression of the long term quality differential between the teams playing at a particular venue.

The most obvious expression of whom is better than whom, is the league position at the end of the season. Upsets are common in the timeframe of a single match and occasionally a side can benefit from an unusually fortuitous clumping of goals over a 38 game campaign (Newcastle finished 5th in 2011/12, despite scoring just 5 more goals than they conceded), but a season’s worth of 380 matches more often sees better sides finishing above inferior opposition.

A gut instinct for where a side will finish, therefore is a crude evaluation of underlying team quality. The identity, if not the order of the top four or five doesn’t really create a problem. The remainder of the table can be broken down into groups, such as mid table perennials, with ambitions to finish in the top half, mid table sides whom are likely to flirt with relegation and teams that are in for a season of constant struggle.

Considering Southampton 2013

The present case of Southampton shows how recent evidence and your own eye witness impressions can be balanced with detailed knowledge of a side’s less recent past, as well as the inclusion of more general league trends.

Recently promoted Southampton is currently an impressive 3rd. They finished last season, their first back in the Premiership, in an ultimately comfortable 14th, slightly above the average finishing spot for a recently promoted team. However, they struggled to escape from the bottom three in the first half of their season and results only picked up once Nigel Adkins was controversially replaced by Mauricio Pochettino in January 2013.

In the flesh, the current side appear much better than a typical side that has recently been playing in the Championship and the appointment of their Argentine manager is a clear and obvious reason for why this improvement has taken place.

However, it also isn’t uncommon for unlikely sides to occupy such heady heights after less than a dozen games and ultimately fall well back into the pack. Newcastle’s 2011/12 effort apart, the likes of WBA in 2012/13 have joined Hull, Blackburn, Bolton and Aston Villa as recent early season pacesetters, before falling well back into the pack in May. Placed 4th after a dozen matches, last season, the Baggies ultimately finished 8th, 12 points off 7th spot.

In addition, part of the Saints’ current position may be in some part due to a soft early schedule and it is certainly as a result of a mean defence. Six of their Premiership matches have seen them keep a clean sheet. The correlation between defensive output in the past and that in the future is weaker than the corresponding attacking stats, partly due to the variety of challenges a defence has to face. So a side that relies on defensive excellence should be less confident about it continuing in future matches. The Saints may also be 3rd partly as a result of the short term inadequacies or bad luck of other, more talented rivals.

So we have good reason to suppose that Southampton are a much better than average Premiership side, but also there are compelling historical reasons why they are currently flying above their true levels. A finishing position of their current 3rd placing would be more surprising than if they followed WBA’s example a fell to back to earth…although perhaps not all the way down to 8th.

The previous, far from comprehensive evaluation has provided a subjective estimation of Southampton’s worth expressed as a likely finishing position. We next need to join together the two different approaches by quantify that worth.

League Position Expressed In Terms of Goal Supremacy

There is a strong relationship between a side’s average goal difference per game and where they ultimately finish and we can use this correlation to convert our informed, but subjective opinion about where in the league pecking order a team should be found to the much more useful currency of goals scored and allowed.

 Team Finishing Position Average Goal Difference Per Match
1st +1.25
2nd +1.06
3rd +0.82
4th +0.67
5th +0.36
6th +0.33
7th +0.17
8th +0.07
9th -0.09
10th -0.11
11th -0.14
12th -0.19
13th -0.24
14th -0.30
15th -0.36
16th -0.44
17th -0.53
18th -0.61
19th -0.75
20th -0.95

Above is the average goal difference per match for each league position over a representative series of Premiership seasons. The average bottom side in the Premiership is over two goals inferior to an average title winning team. Southampton’s current position of 3rd has typically been accompanied by a positive goal difference of just over 8 tenths of a goal per match, falling all the way down to nearly zero if we drop 5 places to 8th. So within the bounds of our optimism or pessimism concerning Southampton’s fundamentals, we can propose an absolute, goal based rating for the Saints.

Now we just need an opponent to provide a game, again with an assessment of their true worth, balanced by pros and cons and expressed numerically.

An Example: Manchester United vs Southampton

Manchester United very recently entertained Southampton. United rarely finish outside of the top two. It is nearly a decade since two sides, Arsenal and Chelsea, managed to finish ahead of Sir Alex’s side in the same season. In recent times, they have persistently confounded the more numerous secondary football stats, such as shots, by consistently finishing higher than this data predicts they should. They have also scored more decisive, late goals than a side even of their ability should, regardless of the disputed length of “Fergie time”. Therefore, from a fundamentally numerical stand point, without the intangibles of a Sir Alex and his backroom staff, United have been due to regress, in the opinion of some, for the best part of a decade.

Their current “lowly” 5th spot (they were even lower when the actual Southampton game took place) may be a manifestation of them finally “playing to their numbers” or it may be the result of a more obvious change in the manager. Moyes may simply be slower than a Potchettino at getting the best from his charges or he may be unluckier in the short term of just 11 matches.

If you consider United under Moyes are of similar talent compared to previous recent sides, you need to pitch their absolute goal difference between 1.25 and 1.06 goals per game, but if their current position of 5th appeals as their true current level of ability, a much less impressive value of around 4 tenths of a goal is appropriate.

However we chose to judge subjective opinion before converting it to numerical format, one final piece of information is needed. League games are venue specific. Sides, in the long term, perform better at home than away. Traditionally home field advantage has hovered around 4 tenths of a goal in the Premiership and although teams can produce widely differing values over a single set of 19 home and away contests, these elevated or depressed levels aren’t generally repeatable over longer stretches of time.

Therefore, without compelling evidence as to why a side’s performance should be much better or worse at home compared to their efforts on the road than is normal for the league, a catch all figure of 0.4 of a goal per game is always preferred.

The final step in turning this integrated approach into match odds involves another plot and some simple subtraction. If we assume (possibly naively) that the current table is an accurate assessment of both Southampton and United, Southampton’s 3rd place is habitually occupied by a side capable of producing season-long performances worthy of a goal difference per match centred about 0.82 of a goal. United’s 5th spot is good for 0.36 of a goal, which jumps to 0.76 of a goal once we add 4 tenths for United’s home advantage in this particular case.

So under these constraints, Manchester United had a venue corrected, absolute expected goal supremacy of 0.76 of a goal per game compared to 0.82 for their visitors. Overall, if we subtract one from the other, Southampton was superior by 0.06 of a goal.

It should come as no surprise that expected supremacy for a match is correlated with a side’s projected chance of winning that match and that relationship is plotted above. Southampton’s match supremacy of 0.06 along the horizontal axis intersects with the plotted line at about 38% on the vertical axis and likewise United’s anticipated long term supremacy for this game of minus 0.06 implies a winning chance of around 35%. The remaining 27% can be chalked off as a draw.

The outcome of the game was a 1-1 draw courtesy of a late Southampton equaliser. The use of odds derived from the current table and very recent form, endorsed Southampton as the team to side with, in the handicap markets, at least and the draw as a more likely outcome than the quoted odds implied.

Against The Odds

In contrast, available quoted pre-game odds universally favoured United, giving them a greater than 60% chance of winning the game. These actual odds are much more consistent with the chances we would have produced had we pushed United into a more customary top two spot and the Saints just outside the semi-permanent elite. One match does not validate one interpretation over another and the continued existence of the bookmakers strongly suggests that a long term view of team quality is consistently more predictive than the shorter term view that made Southampton marginal favourites. Although the latter prevailed on this occasion.

Setting odds is neither wholly an art nor wholly a science. It is a bit of both. Having reached a point whereby match odds can be derived from a combination of numerical and opinion based methods, the next post will look at how we can extend the process to produce odds for all of the secondary markets, such as double results and correct scores. I’ll include a comprehensive, condensed summary of all methods at the conclusion of this series.



Read more of Mark's work on his The Power Of Goals blog

And follow Mark of Twitter: @MarkTaylor0

NFL and football fan. I've seen my two favourite sides, Stoke and the San Diego Chargers play at the new Wembley....and both lost.