Team News Model Adjustments: A Worked Example

In my earlier blog (“How do you use team news to adjust a model?”) I outlined a very basic technique to adjust model prices for team news using the Stratagem player ratings. I will now put this technique into action and look at a specific upcoming Scottish Premiership match:

Heart of Midlothian v Dundee United
Scottish Premiership
Wednesday 30th December 2015

StrataBet Model Projections: SUP 0.73, TOT 2.89

Our first step is to look at the strongest eleven of each team and separate into the Defensive Unit and Attacking Unit:

TeamNewsExample1.png

Supremacy Adjustments
Using the StrataBet Preview for the match, we can see that Juanma ST3 is suspended and Walker RW/AM4 injured, meaning there will be two negative adjustments on the Attacking Unit.

Reilly ST1 and Pallardo DM2 are the players expected to start, so by taking the difference between the ratings, multiplying by 2 and dividing by 100, we get:

DM2-RW4 = -2 x 2 / 100 = -0.04
ST1-ST3 = -2 x 2 / 100 = -0.04

With Hearts playing a big match against Celtic on Sunday and Dundee United’s game with Motherwell postponed, there should be a small negative adjustment made to Hearts to reflect this.

As is often the case with teams playing in the Europa League on a Thursday and then in their domestic competition on Sunday, it can be struggle to adapt to having only 2 days between fixtures.

Not only are the players likely to be fatigued, the lack of time to plan tactically for the next game can also be an issue. I would deduct an arbitrary -0.03 for each unit, adding -0.06 onto Hearts’ Supremacy:

Hearts Supremacy Adjustment = -0.14

Total Goal Adjustments
With both negative adjustments for Hearts coming from the Attacking Unit, this will have a negative impact on our Total Goal quote. We use the same calculations as the Supremacy Adjustments here:

DM2-RW4 = -2 x 2 / 100 = -0.04
ST1-ST3 = -2 x 2 / 100 = -0.04

Hearts Total Goal Adjustment = -0.08

TeamNewsExample2

Supremacy Adjustments
From the Preview, Sinama-Pongolle SS2 misses the match through injury, but it would have been a manager’s choice decision between him and Erskine AM/RM2 to start, so no adjustment is required.

Morris CB1 is keeping Souttar CB/RB3 out of the team in what is also a manager’s choice decision, and this also does not require an adjustment:

Dundee United Supremacy Adjustment = 0

Total Goal Adjustments
With no player changes of note, there is no Total Goal Adjustment:

Dundee United Total Goal Adjustment = 0

Long-Term Adjustments
Dundee United are still affected by losing three very important players in the summer, and despite the significance of these players decreasing as time goes by, there are still some residual adjustments to be made.

Cierzniak GK4, Fojut CB4 and Ciftci SS4 all departed and it has been evident that these players have not been sufficiently replaced. McKay ST4 has not hit the ground running and is still struggling for form, despite being the only striker of any quality at the club.

I would suggest that McKay’s ST4 rating is due in part to the lack of alternatives rather than form of the player. While in defence and especially in the goalkeeping department, United have had a terrible first half of the season.

Hearts on the other hand arguably improved over the summer, while keeping the nucleus of the team that comfortably won the Championship. They replaced Zeefuik ST3 with Juanma ST3, and Wilson CB4 with Augustyn CB3.

It is also worth noting that the departing players were playing at a level below the Premiership, so in my view these changes do not merit a negative adjustment.

Hearts
As discussed, I do not see any residual Long-Term Adjustments needed for Hearts.

Dundee United
For the sake of this example, and using hindsight to judge the loss of quality on the team this season, I will assume each of the three player departures in the summer have been replaced by a “2” rated player, and adjust for time decay:

Cierzniak: GK2-GK4 = -2 x 2 / 100 = -0.04 (SUP) and +0.04 (TOT)
Fojut: CB2-CB4 = -2 x 2 / 100 = -0.04 (SUP) and +0.04 (TOT)
Ciftci: ST2-SS4 = -2 x 2 / 100 = -0.04 (SUP) and -0.04 (TOT)

With each team having played 19 games so far, we are exactly half way through the Scottish Premiership season, so I will multiply the totals by 50% to get today’s decayed Long-Term Adjustment:

Hearts LTA = 0

Dundee United LTA = -0.12 x 50% = -0.06 (SUP) and +0.04 x 50% = +0.02 (TOT)

Other Adjustments
Although rain is forecast throughout the day in Edinburgh, it appears this will clear up as kick off approaches.

The Tynecastle pitch will no doubt be wet, but it is holding up well in the current weather, and should not deteriorate throughout the match. We will not make an adjustment for a poor pitch or weather conditions in this case.

Adjusted Supremacy and Total Goal quotes
Having now accounted for most things that the model does not include, we can settle on our final Supremacy and Total Goal quotes and compare to the market:

Supremacy: Starting Quote + Hearts Adjustment + Dundee United Adjustment
= 0.73 + (-0.14 + 0.06)
= 0.65

Total Goal: Starting Quote + Hearts Adjustment + Dundee United Adjustment
= 2.89 + (-0.08 + 0.02)
= 2.83

SUP

TOT

Adjusted Model

0.65

2.83

Market Actual

0.65

2.55

Comparing our new figures to the market, we are almost in line. Unless the Overs drift to 5% greater than our own price, then we will not place a bet in this match.

Mark McAfee

What is the secret behind Stoke City’s defence?

In the last ten league matches, Stoke City have tightened their defence significantly, registering seven clean sheets and becoming one of the most difficult sides in the Premier League to break down.

Their miserly defensive unit has helped them to continue picking up results, despite being the joint-lowest scorers in the division with Aston Villa.

For a team now heralded for the attacking quality in their side it is the defensive players currently grabbing the headlines. So, is this through good luck, good organisation, or a combination of the two?

For a side sitting in the bottom half of the table, but with the joint-third best defence, it is pretty clear that their issues are not defensive. They are increasingly reliant on their back four to keep the team in contention by shutting out opponents and one of the key factors in this has been a largely settled unit.

In the early part of the season Stoke were missing Shawcross, the club captain and a key figure. This forced Mark Hughes into playing Cameron at centre-back and although he performed well, the defence, which was unchanged for the first four matches, was giving away chances:

Stoke1.png

Naturally, the more chances you give away, the more likely you are to concede, so it is no surprise to see they failed to keep a clean sheet during this period.

After round four, the defence changed for three successive weeks and they conceded five times, but they did pick up their first win.

Since Shawcross returned to the line up against Newcastle in round eleven, the defence has been unchanged, conceding just four times and registering five clean sheets. Indeed, two of the conceded goals came after the captain had been sent off at Sunderland.

It is hard to argue that the defensive upturn has not been related to his presence in the side, but perhaps an unsung hero in this piece is Wollscheid. The German has been involved in every clean sheet Stoke have kept this season, coming into the team a couple of weeks before his partner.

In the past seven matches, the chance creation of opponents has decreased significantly, going hand-in-hand with the unchanged defence. They have given up just five Great Chances in this time and the highest number of Good Chances conceded in a match is seven. This points to better organisation, with the team better at denying chances in the first place, rather than relying on opponents failing to convert.

Stoke2

It is also worth noting that Stoke have the third most effective offside trap in the league, catching opponents with it 43 times so far, a figure bettered only by Liverpool and Swansea. Their offside trap has maintained its effectiveness since the Wollscheid/Shawcross partnership was reformed, with 19/43 offsides being picked up in the last seven games.

Stoke4

Has this defensive solidity come at the cost of attacking football though?

It certainly could be the case, as in the same seven game period they have found the net just five times. In the opening seven they averaged a goal a game, which in itself still is not clinical, but is significantly better.

Stoke are still doing a reasonable job of creating openings at the other end, averaging 1.59 Great Chances per game. The fact that they are not scoring is another story in itself, but it is testament to the defence that they are continuing to pick up points with a forward line that has not found the net more than twice in a game all season.

Interestingly only four sides have given away more chances than they have overall, with three of those being in the bottom five (Newcastle, Norwich and Sunderland). So, although they have tightened up, the defence is still also relying on opponents’ poor finishing and good goalkeeping to keep the score down.

Stoke4

The amount of Great Chances conceded is significant here. Eight teams have given away more Great Chances, although again all but two are below Stoke in the table.

So, if they are giving away a reasonable amount of chances to opponents, how are they keeping clean sheets?

The emergence of Butland in goal has been an important factor for the side, with the 22-year-old enjoying his first season as undisputed first choice. He ranks third behind only Petr Cech and Joe Hart in saves made from Great Chances and tops the table in saves from Good Chances.

With a goalkeeper in this kind of form giving away chances is often not too harmful, providing the rest of the team with a very solid foundation to build upon.

Stoke5

Lastly, Stoke are a team who are able to prevent pressure being created by keeping the ball effectively in opposition territory.

Even though they are short of goals they still manage to have just below 49% possession per game on average, meaning the defence is not coming under constant pressure. They give themselves time to reorganise by keeping the ball and allowing players time to recover, which in turn results in better defensive shape.

These factors combined with their ability to keep their defenders fit and available suggest that they are unlikely to start shipping goals regularly. If they can get ahead in matches, it is likely that they will manage to stay ahead for the remainder of them.

Whether this is enough to propel them to a top ten finish remains to be seen, but that looks to be their upper-limit based on their current metrics.

Alec Payne

Can Derby County finally get promoted?

As the English Championship completes its 21st match round, things are starting to take shape at both ends of the table. The prize for reaching the Premier League is now astronomical in terms of the money, as well as the prestige of playing in one of the best leagues in the world.

Watford, Bournemouth and Norwich have all shown that promoted teams are capable of holding their own in the top flight. Derby County have been pushing for promotion from the Championship for the past couple of years and been very unlucky not to make the jump.

After dominating the play-off final in Steve McClaren’s first season in charge they were beaten by an 89th minute goal. Last season they topped the table on 1st March, only for a combination of injuries to key players and the uncertainty surrounding McClaren’s position at the club saw them slip to 8th on the final day.

In the summer they appointed Paul Clement, unknown to some but coming with a good reputation following spells at Chelsea, Paris Saint-Germain and Real Madrid as coach/assistant to some of the best managers in the game.

They also spent a lot of money and by the time the transfer window shut they had signed:

  • Bradley Johnson (Norwich City, £6 million)
  • Tom Ince (Hull City, £4.75 million)
  • Jacob Butterfield (Huddersfield Town, £4 million)
  • Jason Shackell (Burnley, £3 million)
  • Andreas Weimann (Aston Villa, £2.75 million)

Chris Baird, Darren Bent, Alex Pearce (all free transfers) and Scott Carson (undisclosed) also joined in small fee/high wage deals.

Expectation was quite rightly very high but things did not go immediately according to plan. They failed to win their first 5 games of the season, drawing 4 and losing 1, and were also knocked out of the League Cup by Portsmouth.

However, since then they have lost only 1/16 games, winning 11, and sit well positioned in 3rd, just 2 points off the top. The top four in the Championship already seem to be pulling away from the pack and just 2 points separate them all, before a gap of 6 points to Burnley in 5th.

But do Derby have what it takes to finally get promoted?

I used StrataBet to investigate how their statistics looked this season, as well as where the turnaround in form had come from.

Some metrics make for obvious reading, such as Derby scoring the second most goals in the division (32) and conceding the second least (13). This looks very healthy, but using the Query Tool I looked deeper into Derby’s chance creation. Many clubs can do well for a short spell due to an unsustainable conversion rate but it’s hard to maintain this over a season.

When looking at the regularly used Shots On/Off Target metric Derby do not stand out. In terms of both Shots On Target For & Shots Off Target For they sit in the middle of the pack as shown in the tables below:

English-Championship-2015-16_Total-Shots-On-Target-(For).pngEnglish-Championship-2015-16_Total-Shots-Off-Target-(For)

This may lead us to believe that they should either be shooting more, or that a dip in the number of goals they score is coming.

However, at Stratagem we don’t just collect statistics on Shots On/Off Target. We collect in-depth data on the quality and situation of each chance, focusing primarily on what we call Great Chances, Good Chances and Attempts.

Derby are at the high end of creating Great Chances (3rd) and Good Chances (4th), while they also concede very few high quality chances – 3rd in Great Chances conceded/6th in Good Chances conceded.

We can work out a rough expected goals measure from all their chances created according to our average conversion rates. We would expect them to have scored around 33 goals and conceded 22, so they are ever so slightly under-scoring but doing much better than expected defensively. Perhaps this is where they will begin to slip?

In terms of style Derby are one of the teams who are most easy on the eye. Clement plays a passing and possession-based game, using pace in the wide areas to get behind the defence. They are versatile and can play direct up to Martin if they require and are also very dangerous on set pieces (scoring from one roughly every 3 games).

Key Entries (possession in the final 18 yards of the pitch) is one metric where Derby are well clear of the rest of the division. They have already achieved almost 10% more Key Entries than the team in 3rd, Brighton.

The table below gives a breakdown and there appears to be correlation between teams at the top end of the table having a higher number of Key Entries, something that makes sense:

English-Championship-2015-16_Total-Key-Entries-(For).png

The key teams to compare Derby to are their direct rivals – Middlesbrough, Brighton and Hull. The Matrix below gives an idea of how they rate compared to these teams:

Derby4

Each team is also given a statistical fair score, which we wrote about in our last blog. This shows that 3 of the top 4 are exactly where they “should” be, with Derby deserving to lead the table. Brighton are the only team in a position much better than could be expected.

Derby are on the right path. They are scoring at what appears to be a sustainable rate and though they are not conceding as many goals as we would expect, this could be down to a high quality back line.

They have already had a rough spell with injuries so will be hoping to avoid a repeat of last season and are hitting form at a good time of year, which should separate the teams with smaller squads.

As Brighton face Middlesbrough this weekend Derby will be hoping for the draw that could allow them to be top of the table at Christmas, provided they can beat Ipswich away. If they manage that, then maybe they can indeed get promoted this season.

Dave Willoughby

Does the league table lie?

“The league table never lies.”

A statement we’ve all heard before, most often towards the back end of a season, but does it hold up?

Of course it depends on the way you look at it, but at Stratagem our firm belief is that the league table does lie and on a more frequent basis than you might expect.

Sometimes these lies can be marginal ones, but on occasion they can be much bigger. These “big lies” almost always lead to trading opportunities.

You might be wondering what we mean by “big lies” and, more importantly, how you can spot them yourself.

To explain, let’s take a look at the English Premier League table after 15 rounds:

FairScoreActual

A quick eye test of the standings throws up a couple things that just appear to be plain “wrong”. Leicester City are top in mid-December, for starters, while champions Chelsea find themselves in mid-table and are far closer to the bottom three than the top four.

However, just because something “looks wrong” does not mean that it is wrong. The table essentially tells us that Leicester City are performing well above expectations and Chelsea well below. Do these teams deserve to be where they are, though?

This is what we, as football analysts and traders, are really interested in.

Our team of expert analysts give “Fair Score” judgements of every game that we review on our Data Collection Platform. They use a variety of factors to come to their decision, weighing things such as chance creation, momentum, pressure, tactical approach and controversial decisions.

Thanks to this we are able to maintain a weekly “Analyst Fair Score Table”, which currently looks like this in the English Premier League:

FairScoreAnalyst

According to this Leicester and Chelsea are largely deserving of their current rankings and the “big lies” that our analysts believe the current table is telling us actually revolve around Liverpool, Southampton and Norwich City.

Jurgen Klopp’s men are currently four places higher in the league than they perhaps deserve to be, while Southampton and Norwich are five places below. Interestingly both the Actual Table and Fair Score Table are aligned on ten teams in total, including the entirety of the bottom four.

It is important to acknowledge there is a fair amount of subjectivity at play here, given that we are accepting a purely human judgement on what a “fair score” should have been in any given match. However, we train our analysts to the highest possible standards and also try to eliminate personal biases along the way, in order to minimise the risk of error.

Also, to add another perspective, our in-house Quantitative Research team have built a “Fair Outcome Model” using our proprietary analyst data.

The difference between “outcome” and “score” looks quite subtle, but is a key thing to highlight. This model does not reward teams in whole goals or even complete points, instead rewarding teams a share of the points available depending on their in-game performance metrics.

These metrics are fairly wide-ranging, but primarily focus on chance creation and concession:

FairScoreModel

Firstly we see that Leicester and Chelsea are again believed to be in their “correct” positions in the table, more-or-less, but the model suggests that the Actual Table is far less trustworthy than the analysts led us to believe.

Stoke City, Norwich and Bournemouth the only teams it is aligned on and the two biggest lies the model thinks are being told revolve around West Ham and Southampton (again). In its view West Ham should be six places below where they actually are, with Southampton six places above.

The interesting thing here is that the analysts and the model both agree that Southampton are deserving of a much higher league ranking based on their performances and the performances of those around them.

Aside from the teams mentioned everyone else is apparently close enough to where they “should” be, though Manchester United have apparently been over-rewarded and West Bromwich Albion under-rewarded by 3 places apiece.

Finally, let’s look at a point-by-point comparison of the teams across all three versions of the table:

FairScoreComparison

This graph gives a very clear representation of how teams have been rewarded, how the analysts believe teams should have been rewarded and how the model believes teams should have been rewarded.

The truth probably lies somewhere between the three lines. As such, from this you can begin to build a more accurate picture of how teams have performed so far this season.

When you compare this to weekly outright market prices you can begin identifying trading opportunities.

Adding further layers of analysis like looking at performances against certain “groups” of teams and assessing line-up strength per game can enhance the accuracy further. We will look at how to do this in a future blog post.

Rich Huggan

How do you use team news to adjust a model?

StrataBet currently provides model prices for matches in 17 competitions around the world, but does not account for team news or other factors that influence a football match (Models FAQ – “What ‘data’ is not included in the model?”).

Using my own experience as a football trader, I will attempt to show how to use the StrataBet player ratings to account for team news on a very basic level.

Strongest XIs
The first part of the process is identifying the strongest XI that each team could field if everyone was available for selection. This can be quickly calculated by viewing player ratings from the most recent Preview, in the “Expected Line-Ups” tab.

If we look at Ross County in the Scottish Premiership, we can list their strongest XI as follows:

TeamNewsModel1

It would not be correct to assume that if Gardyne RW3 and Robertson RB/CB3 were both absent for the next match, you would deduct a value from Ross County’s Expected Goals. Instead, if we think of each team as two units (an attacking one and a defensive one) we can accurately account for the weakening effect of important players being absent (Supremacy), and also the impact on the expected goals in the game (Total Goals).

In general terms, if an important striker is missing for a game, you would expect his team to see a lower goal production than average. Naturally, the opposite is true for a missing goalkeeper or key defender.

Goalkeepers (GK), full-backs (RB/LB), centre-backs (CB) and defensive midfielders (DM) belong to the Defensive Unit. Strikers (ST), support-strikers (SS), attacking midfielders (AM) and wingers (RW/LW) belong to the Attacking Unit. Central midfielders (CM) and wide midfielders (RM/LM) are in neither group, as they are considered to contribute equally in both defence and attack.

If we apply this rationale to Ross County we get the following:

TeamNewsModel2

To give a real-world example here: you have model prices for a match, but on reading the Preview you notice that each team is missing some players from their Strongest XI. What do you do?

Supremacy Adjustments
From the Preview, Ross County are replacing the injured Gardyne RW3 with Franks RW1 and Boyd CB1 is expected to start over Robertson RB/CB3.

On a very basic level, a quick calculation you could make would be to take the difference in the ratings, multiply by 2 and divide by 100. This would give:

RW1-RW3 = -2 * 2 / 100 = -0.04

CB1-CB3 = -2 * 2 / 100 = -0.04

On the face of it Ross County are weakened by 0.08 of a goal for this match.

You would then go through the same process for their opponents and add the net figure onto the Supremacy quote. For example if Ross County were away from home, their opponents were at full strength and the model Supremacy quote was 0, then the new quote is 0.08.

Total Goal Adjustments
Using the same team news as above, we can see that one of the missing players belongs to the Attacking Unit and the other to the Defensive Unit.

We would use the same adjustments as the above example to arrive at 0.04 for each player change, but this time the Defensive change is a +0.04 adjustment onto the Total Goals, and the Attacking change is a -0.04 adjustment onto the Total Goals.

In this case, they cancel each other out. Again, you would complete this task for Ross County’s opponents to arrive at a net figure and then add/subtract onto the Total Goals quote.

Other things to consider
As I have mentioned, this is a very basic way of looking at adjusting model prices for team news.

It does not take into account the match prices. If Barcelona were -2.5 favourites at home to Real Sociedad, then they are expected to attack for most of the match. This would have an impact on how much we would adjust if Messi ST5 was missing, or if Busquets DM4 was out.

It is obviously correct to adjust for important players being suspended or injured, but what if the manager simply decides to start a lower rated player over a higher rated one? If we trust his judgement, considering he sees every player in training each day, then it may not merit a negative adjustment at all.

These are some of the issues that will be explored in further detail in future blog posts.

Mark McAfee

Leicester City: great pretenders or real contenders?

We are 14 games into the 2015/16 Premier League and the table is starting to take shape.

Most of the usual suspects are sitting in the upper reaches, but one team are seemingly determined to crash the party – Leicester City. Having suffered just a solitary defeat they sit joint top with Man City, boasting an attack that is the second most potent in the division.

But are Leicester massively over-achieving? Or are they a genuine competitor for the league title?

In this piece I will take a closer look at the numbers that have got them to where they are today.

Leicester3

Going forward, Leicester are a functional and highly effective attacking unit, with their average of 2.07 goals per game bettered only by Man City on 2.14. Defensively, however, just seven teams average more goals against than Leicester’s 1.5, including all bottom five clubs.

Let’s compare that number to the teams around Leicester in the table at present. Of the others making up the current top six, Liverpool average the highest goals against, but their figure stands at just 1.07.

So how can Leicester sustain their place in the table when they are giving away at least 0.4 more goals than their nearest competitors?

The short answer is that although they score more than they concede, the difference in average is smaller than those around them. By having a poorer defence, there’s more pressure to convert when they do create. Those who are tighter at the back can afford to be a bit more wasteful in front of goal, as they are giving away less to opponents.

It will undoubtedly be hard for Leicester to sustain their current strike rate, although there’s no denying that they are being incredibly clinical at the moment.

Leicester2

Breaking it down a little further into Key Entries and we again see Leicester coming up short compared to those around them. On average, they make 44.43 entries a game; whilst all other teams in the top six are far more active upfield.

However, they actually give away fewer entries per game than three of their rivals, though they are the only side who concede more entries than they produce.

So does this tell us Leicester make the most of their opportunities? Or that their opponents aren’t clinical enough?

The answer is there is truth in both statements…

Leicester1

In terms of chances created per game (Great and Good), Leicester are once again rated highly with 5.5 being made on average, a figure only bettered by Arsenal and Manchester City. Leicester’s chance creation to entry ratio is very strong, with a chance created in every 8.8 entries.

However, once again they fall down in the defensive side of the game, giving away more to their opponents on average over 90 minutes (3.93 chances). This equates to an opponent creating a chance every 12.2 entries.

Of course there are a number of variables that one must consider in each game played, such as form of individuals, game state, injuries and more. However, to give away a relatively high number of chances compared to opponents once again questions the sustainability of their league standing.

Leicester4

A quick look at last year’s top six shows us that if Leicester were to sustain their levels of productivity, they could indeed finish inside this year’s, although a genuine challenge for the title looks much less likely.

They are showing that they are among the best sides going forward, but one thing all those listed above have in common is a much better defensive record than what Leicester have right now.

Overall, it seems that Leicester aren’t a side sitting at the top of the table through luck, but through effective attacking play that sees them capitalise on enough of their chances when they arrive.

However, the chances of sustaining a title challenge or even just a place in the top four look unlikely if they can’t stem the number of chances being conceded at the other end.

Alec Payne