Analyst Data Part II: Why is it valuable?

Yesterday we discussed what the Analyst Data offered in the StrataBet Invest subscription is, but today we explain why it is valuable.


This graph from Part I shows the proportionality of chances in the English Premier League from the beginning of 2014/15, in addition to the conversion rates of those chances versus the “Shots on Target” metric found in the wider media.

We highlighted that the relationship of Great Chances, Good Chances and Attempts is almost perfectly linear.

This is important because it demonstrates how chance data can be used to calculate the “fair outcome” of a fixture almost instantaneously.

Take the basic Match Data from a recent Premier League match between Chelsea and Southampton:


The statistics here give the impression of an even contest, which Chelsea possibly shaded. A fair assumption would be that this match ended in a 1-0, 2-0 or 2-1 home win, or that they were “unluckily” held to a draw at the very least.

Crucially, the Analyst Data presents a very different story:


It shows that Chelsea were extremely inefficient and that Southampton executed their game plan perfectly, which is a much truer reflection of the contest.

Knowing the conversion rates of Great Chances, Good Chances and Attempts a fair assumption would be that this match ended 0-2 and that Chelsea would have been “lucky” to have achieved a draw. This is much closer to the truth, as Southampton actually won 1-3.

This is just one significant value-add of the StrataBet Invest data. On Monday we will highlight two more ways to leverage it for trading and modelling in Analyst Data Part III: How do I use it?

Analyst Data Part I: What is it?

The major selling point of StrataBet Invest is the Analyst Data, but what is it?

Every match from our 16 “top” competitions is subject to rigorous data collection by an expert football analyst.

160-200 individual events are typically recorded per match, with the most crucial of these being our measures of chance quality.

At Stratagem we focus on “chances” instead of “shots”, breaking them down into three categories and showing them on our Review pages:

Great Chances
Good Chances

A “Great Chance” is a situation that a player would be expected to score from.

A “Good Chance” is a situation that a payer could score from but would not necessarily be expected to.

An “Attempt” is a situation that a player would not be expected to score from.

*Please note that these situations are not required to result in a shot to be recorded.

Since the beginning of the 2014/15 English Premier League season the proportion of these are as follows:

17% of all chances are Great Chances
28% of all chances are Good Chances
55% of all chances are Attempts

The graph below shows their conversion rates compared to that of shots on target in the same timeframe:


The relationship here is almost perfectly linear: Great Chances result in 13% more goals than Shots on Target, which result in 12% more goals than Good Chances, which result in 13.5% more goals than Attempts.

We will discuss why this is important tomorrow in Analyst Data Part II: Why is it valuable?

Quantitative Research (QR) #2: Stratagem Football Model Performance


A reasonable strategy has been found to choose bets based on odds and model value, this gives a return of +0.80% per unit liability on ~3000 bets across 7 leagues in the top 5 European domestic league systems.


The backtests have been run using the trading system framework. The following match odd were considered:

• Betfair over/under and Asian handicap markets

• Match kick off: 1st October 2014 – 1st October 2015

• Leagues: EngPr, EngCh, GerBL1, GerBL2, SpaPr, FraL1, ItaSA

For each suitable fixture the odds were found 5 minutes before kick-off. The (previously calculated) model fair odds prediction for each bet is calculated and the back and lay values are calculated as (odds/fair odds) − 1.

A bet is placed on the back side if the back value is > 0, a bet is placed on the lay side if the lay value is < 0. Note that this strategy is refined in the post-run analysis.


On a brief analysis of the results for EngPr in the period 1st October 2014 – 1st July 2015 the following strategy was determined:

• Back bet if back value ∈ [0.0, 0.1] and odds ∈ [1.6, 2.0]

• Lay bet if lay value ∈ [−0.1, 0.0] and odds ∈ [2.0, 2.5]

Staking is simply 1 unit of liability per bet.


Back/Lay Odds Range PnL/Unit Number of bets
Lay 2.0-2.5 +0.23% 1886
Back 1.6-2.0 +1.66% 1277

Trade Idea #2: Lay Arsenal for Top 2 @ 2.00 or better

For this week’s long-term trade idea we take a closer look at the Top 2 market in the English Premier League, with primary focus on Arsenal.

Lay Arsenal for Top 2 @ 2.00 or better

At the time of writing the market is implying a ~52% probability of Arsenal finishing Top 2, essentially pricing them at 70% over Manchester United.

Our analysts believe that the market is rewarding Arsenal for their unprecedented chance creation, but failing to account for their inefficiency in converting these into goals.

Quantitative Analysis



Arsenal rank 1st for Great Chances (18) and Good Chances (30) this season, while also ranking 2nd for Attempts (63), Attacking Third Entries (526) and Corners (50). However, they are =5th for Goals (10) and currently sit 4th in the league standings.


Last season they ranked 1st for Good Chances (147) and Attacking Third Entries (2334), 2nd for Corners (255) and 3rd for Goals (71), while they were 4th for Great Chances (54) and Attempts (253). They ended up finishing 3rd in the league.


In 2013/14 they ranked 4th for Goals (68) and Great Chances (57), 7th for Good Chances (124) and Attacking Third Entries (2188), =8th for Corners (210) and =12th for Attempts (236). They finished 4th in the league.



At the start of 2015/16, Arsenal have been by far the most profligate team. They have converted just 0.217 Great Chances and 0.032 Good Chances. By comparison, Manchester United have converted 0.666 Great Chances and 0.0666 Good Chances.


At the start of last season, no team was anywhere near this profligate and Arsenal themselves were converting almost 0.4 of their combined Great and Good Chances. However, at this time they were only averaging ~3.8 per game. 


By the end of last season, Arsenal combined their ruthlessness with more prolific chance creation to help them to 3rd place. However, it is important to note that their conversion dropped to <0.25 of their Great and Good Chances, as their creation rose to ~6.9 per game.

The eventual top four were clearly the most prolific in the league, with winners Chelsea (~0.27 of ~ 6.5 scored) and runners-up Manchester City (~0.28 of ~7.15 scored) having the strongest correlation between conversion and creation.

At the start of 2013/14 Arsenal began very strongly, creating a combined ~7.5 Great and Good Chances per game. Crucially, they scored ~0.215 of them at a time when conversion was down across the league.

By the end of 2013/14 Arsenal had done a good job of maintaining their prolificacy, trailing only winners Manchester City and runners-up Liverpool. However, they could only convert this into a 4th place finish due to Chelsea’s far superior defensive record (26 goals conceded vs. 40 for Arsenal).

As in 2014/15 the eventual top four were clearly the most prolific teams, though this time winners Manchester City (~0.29 of ~8.5 scored) and runners-up Liverpool (~0.29 of ~7.8 scored) were significantly better at converting their many chances than 3rd place Chelsea (~0.215 of ~7.2 scored) and 4th place Arsenal (~0.26 of ~6.5 scored).

Risk Factors

  • Arsenal keep up an unprecedented level of creation and significantly improve conversion.
  • Manchester United fail to maintain their usual rate of creation/conversion.
  • Chelsea do not “bounce back” in the expected manner.