Nothing beats what your eyes see themselves when it comes to picking winning selections however the right stats and ratings can be a useful tool in every punter’s armoury. Here Darren Davies, launches his new look at ratings with a debut contribution to Progambler.

**Onside**

‘It’s better with a bet on’, and with Ladbrokes posting pre-tax profits of £235m for 2006, it just gets better and better. And it’s not like the Magic Sign are on their own, Gala Coral is the UK’s biggest private firm, Bet365 have increased their sales from £112m in 2002 to £986m last year, whilst Betfred have doubled their turnover from £387m to £874m in three years. Earlier in April, the eighth annual Profit Track 100, which ranks Britain’s fastest-growing private companies by average profit growth, listed spread betting firm Spreadex, boosting profits 76% a year from £1.2m in 2003 to £6.4m in 2006, bookie Stan James, increasing its profit 70% a year from £1.1m in 2002 to £5.5m in 2005, and Betfair, where an average of 6m of bets matched every day helped profits surge 61% a year from £8.3m in 2003 to £34.6m in 2006. ’It’s better with a bet on’, as far as the bookmakers are concerned never were truer words spoken.

This article intends to guide the reader through the complex task of denying the bookmakers of some of their hard-earned prosperity. By concentrating on the fixed-odds markets for football betting I will show how it is possible not only to win, but to win consistently and ultimately to make an income. However, before reading on you need to ask yourself this question, ’why do I bet on football?’ If the answer is, ’to make it more interesting’, or ’cause I always back the Hammers’, or worse, ’that bleedin’ scorecast is bound to come in if I don’t’, then stop here and think nothing of it, this article is not for you. If, however, your answer is indeed, ’to make an income’, then read on.

# Value Betting

It may well be that I am preaching to the choir, however the concept of value is so central to any winning punter’s thinking that I would like to reiterate a few basic principles and dispel some commonly held, and widely purported, myths regarding value betting.

*Principle #1*

In simple terms a punter has identified value when a bookmaker prices the outcome of an event at odds greater than the true odds of that event occurring. The assumption is that a skilled punter can identify such situations, a matter to which we shall return.

*Principle #2*

Where value has been identified the outcome has to be backed to occur, regardless of whether the punter thinks it will win or not, conversely, where the absence of value has been identified the outcome must not be backed, even when the punter in question believes it to be a stone cold certainty to occur, due to *Principle #3*.

*Principle #3*

Backing outcomes at odds greater than their true odds of occurring will guarantee a long-term profit, backing outcomes at odds less than their true of odds of occurring will guarantee a loss and will break your bank in the long run. Now let’s dispel some commonly held myths.

Myth #1’Value betting means always backing outsiders’. Referring to *Principle #1* it should be clear that this is the case only when the outsider in question is priced up by the bookmaker at odds greater than the true odds of it winning, otherwise *Principle #2* applies and no bet.

Myth #2’There is no value in backing odds-on shots’. Again, *Principle #1* applies, if the selection is overpriced, even at odds-on, value exists and *Principle #2* demands a bet. In fact, Joseph Buchdahl, in his book Fixed Odds Sports Betting, argues that there is a such a thing as favourite-longshot bias, which he provides evidence for and that states bookmakers overprice odds-on favourites and underprice longshots, ‘in order to insure themselves against insider information and subsequent excessive liability’.

Myth #3’Value betting removes the element of skill in picking a winner’. Any fool can get lucky, it takes real skill to identify value, bookmakers are nobody’s fools, quite the contrary, and to figure out when they have made a mistake is the consummate punting skill. To summarise, and to quote Buchdahl, whose approach outlined in his aforementioned book provides the basis for this analysis of football betting,’ Value-bet or don’t bet at all’, and finally and most pertinently, ’Find the value and the winners will take care of themselves’.

### Finding the value

We have all experienced that feeling when we look down at the coupon and something just jumps off the paper screaming to be backed, that team, that opposition, and at that price. We know their going to win and we know the price is too big, in these cases follow your instincts every time. For ’feel’ is an important weapon in the successful punters arsenal, that gut feeling, the hunch you get is more often right than wrong and you should follow it every time. However, these experiences are usually few and far between and it can’t be forced, as soon as you start to look for it, by definition, it isn’t there. Give it up. Often we ’fancy’ a particular team to win (or lose) and start to look for reasons to back (lay) them, usually they are a form team, running particularly hot or cold which is what alerts us and qualifies them as a ’fancy’. We start to look for team news, maybe we look into their home/away form more closely, even look at recent head-to head results, look up the football forums on the internet to see if there is any wisdom to be found in cyberspace. Based upon the results of this analysis we decide to invest some of our hard-earned money, and again, such endeavour is often rewarded. Both these approaches can be described as a qualitative analysis. Making a judgment based upon available information. Few punters take the next step and quantify these factors, assign them a numerical value, and total them to provide a quantitative measure of the superiority of one football team over their opposition. This is essentially a Ratings System, the sophistication of which is dependant upon the amount of information used to determine the numbers, ranging from the very simple, say league positions and goals scored, to the very complex, incorporating league positions, goals scored and conceded, head-to-head record, match statistics (shots, corners, possession), home/away performance etc etc. Once the number crunching is done, the rating (the number for the away side subtracted from the number for the home side) for the match needs to be translated into the relative chances of a home win, a draw, and an away win, from which a judgment about available value can be made.

**Developing a Ratings System** I have developed and perfected my own ratings system over the last three years. The first step involves identifying a potentially profitable market, the second to develop a theory as to the factors that influence the outcome and to then quantify them into a ratings system. My first relatively unsuccessful attempts centred around correct score betting and the total goals markets. The former suffered from the huge overounds the bookmakers enjoy on these markets and the latter didn’t provide enough value with regard to the matches that the system was throwing up. The point to note here is that I tested a theory about the outcomes of football matches, what I learnt was not to test it in real time, rather back-test the theory against previous seasons results (much cheaper when the theory is wrong), and to choose markets that favour the punter rather than the bookmaker.Up until recently, pre-internet, it was impossible to bet singles on football matches, it was minimum trebles with the high-street bookmakers. The arrival of the internet and the subsequent competition it introduced to the market meant that betting singles on football matches soon became a reality. This is a punters market for sure; only three possible outcomes, low overound, widely available information, reliable form, lots of competition. Take a look at the table below:

**Table 1.** **English Football % 1X2 2006/07**

League |
1 |
X |
2 |

Premiership | 47.89 | 25.79 | 26.32 |

Championship | 48.19 | 22.28 | 29.53 |

League 1 | 45.11 | 25.18 | 29.71 |

League 2 | 44.75 | 25.72 | 29.53 |

The results for this season are consistent with previous seasons. What struck me is that nearly half of all football matches end as a home win. This pattern is repeated year after year and across the football world. This seemed a promising market to investigate further. I theorized as to the factors that influence the outcome of a football match, listed and prioritized them, assigning them relative values and began to back test. A laborious task involving calculating a rating for each team for each match, converting it to a percentage chance of a home win and from there to ‘fair’ odds.The website www.football-data.co.uk provides access to results and statistics from previous seasons for the English, Scottish and major European leagues. Additionally, and crucially they also provide data on the odds available thus allowing an available value judgment to be made once the ratings system has provided the ‘fair’ odds.A good ratings system will model reality; it will predict the outcome of a football match. If, historically, this can be demonstrated it can be used with confidence in the future. My own system, developed through three years of back testing, the final year of which went ‘live’, so to speak, with real money, models

reality like this:

**Table 2.** **Rating System results distribution**

Rating |
% home wins |
# Matches |

< -60 | 7 | 28 |

> -60 < -56 | 0 | 7 |

> -56 < -51 | 25 | 16 |

> -51 < -46 | 41 | 17 |

> -46 < -41 | 13 | 23 |

> -41 < -36 | 20 | 20 |

> -36 < -31 | 15 | 27 |

> -31 < -26 | 31 | 32 |

> -26 < -21 | 32 | 34 |

> -21 < -16 | 31 | 36 |

> -16 < -11 | 23 | 71 |

> -11 < -6 | 29 | 70 |

> -6 < 6 | 37 | 303 |

> 6 < 11 | 44 | 178 |

> 11 < 16 | 41 | 180 |

> 16 < 21 | 54 | 191 |

> 21 < 26 | 49 | 152 |

> 26 < 31 | 44 | 151 |

> 31 < 36 | 50 | 125 |

> 36 < 41 | 46 | 94 |

> 41 < 46 | 50 | 92 |

> 46 < 51 | 52 | 52 |

> 51 < 56 | 60 | 52 |

> 56 < 60 | 60 | 43 |

> 60 | 73 | 225 |

For the purposes of Table 2 the ratings from seasons 2004/05, 2005/06 and 2006/07 have been grouped by intervals of 5 and the corresponding percentage of home wins for each interval are shown, together with the total number of matches for each interval. If the ratings system is any good we should see a low percentage of home wins where there is a large negative rating, i.e., the home side has a rating much lower than the away side, whereas we should large positive ratings have a high percentage of home wins, where the home side has a superior rating to that of the away side. ** **

** **

** **

**Doing the Math!** A decent secondary education and you are more than capable of doing the maths required for this type of analysis. Even if you’re not, Excel is, though a word of caution, Excel only does what you tell it do, it has yet to develop the function that tells you you’re stupid and that you should really be doing it like this! Anyway, looking at Table 2 it seems that the ratings system is doing a decent job of describing reality. Again though we need to move from a qualitative description to a quantative analysis. In order to do this we need to consider the degree of correlation between the ratings and the number of home wins. Excel does this very well in the shape of a scatter graph and the resultant correlation coefficient.

**Graph 1. ****Scatter**** ****Graph for Ratings and % of Home wins**

The x-axis is the ratings, the y-axis the percentage of home wins and the points the plot for each interval. The regression line passes through these points as a line of best fit and R^{2} is the correlation coefficient of the regression line. The value of R^{2} can fall in the range of –1 to +1, where –1 would indicate perfect negative correlation and +1 perfect positive correlation. R^{2} describes how much of the distribution of the ‘real’ data can be described by the ratings system; the closer to +1 the more of the distribution is described by the system. The R^{2} value of 0.8101 indicates the system is a good description, a good model, of reality. Oftentimes R^{2} can be expressed as a percentage, in this case 81%, however this does not mean that it predicts 81% of the results correctly (no such luck!), rather that it predicts 81% of the variance observed in the results, the remaining 19% of variance is not described by the system and is due to factors outside of the computations involved.The equation of the line is a mathematical description of any point on that line. The equation can be used to predict the percentage chance of a home win when we know the rating of any given match. Let’s look at an example. Home team rating = 65Away team rating = 37 Match rating = 65 – 37 = 28 Equation of the line: y = 0.428x + 37.063 Where x = match rating and y = percentage chance of a home win. Therefore y, the percentage chance of a home win, is calculated as follows y = 0.428(28) + 37.063 y = 11.984 + 37.063 y = 49.047 We now know that for this match there is a 49.047% chance of a home win, close to 1 in 2, which we commonly know as 50-50 or evens. To convert this percentage to decimal odds we need to do the following calculation, 100/49.047 = 2.04 So, now we know that, according to the ratings system, the ‘fair’ odds for the home win are 2.04. Using decimal odds as opposed to fractions allows easier comparisons to the odds available and subsequent value judgments. Let’s say that our favourite bookmaker has priced the home win up at 2.25, well then we have identified value and *Principle #2* demands we have a bet. If however the odds are a miserly 1.87 then there is no value to be found and therefore no bet. Of course if we shop around we might get 2.1, and we’ll probably be able to get 2.24 on Betfair, in each case a bet is warranted.For my own purposes I compare my ‘fair’ odds to the average odds for the home win as shown on BetBrain and take it from there. Conveniently football-data collate this information for ready analysis.

# Results

Looking at three seasons results, backing each match where value is identified on Betfair, the results are as follows:

Season | 2004/05 | 2005/06 | 2006/07 |

Bets | 124 | 355 | 443 |

Profit/Loss | +18.45 | +34.57 | +12.27 |

% Yield | 14.88 | 9.74 | 2.77 |

Three consecutive seasons of profit, yield is defined as the percentage profit on turnover, and certainly for 2004/05 and 2005/06 this is more than acceptable, for 2006/07 less so. In 2004/05 I back tested only against the Premiership, in 2005/06 the Premiership and the Championship whereas for 2006/07 the data comes form the Premiership, the Bundesliga, Serie A and the Spanish Primera. During this time it has been clear to me that certain sub sets of ratings are performing better than others. This is illustrated by the following table:

**Table 3a. ****Ratings sub sets**

Season |
2005/06 |
2006/07 |
||

Sub set |
Bets |
Profit/Loss |
Bets |
Profit/Loss |

<-10 | 10 | -10 | 33 | +25,9 |

>-10<10 | 20 | +18,85 | 43 | +6,10 |

>10 < 20 | 38 | +1,93 | 52 | -12,77 |

>20 < 30 | 65 | +10,95 | 70 | +23 |

>30 < 40 | 65 | +6,14 | 58 | -18,85 |

>40 < 50 | 59 | -4,2 | 30 | -2,72 |

>50 | 93 | +11,9 | 157 | -8,4 |

Grouping the sub sets into less than or equal to 30 and greater than 30 reveals a clear pattern. That being that returns are maximized by only betting on matches with a rating below 30. **Table 3b. ****Sub set groupings**

Season |
2005/06 |
2006/07 |
||

Sub set |
Bets |
Profit/Loss |
Bets |
Profit/Loss |

< 30 | 133 | +21.73 (16.3%) | 198 | +43.03(21.7%) |

> 30 | 217 | +13.84(6.4%) | 245 | – 29.97(-12.2%) |

Note. Detailed results from 2004/05 unavailable Why is this the case? I have no definitive answer, a few theories, that these are for the most case evenly matched teams where home advantage is decisive for example, or that in extreme cases with low or even negative match ratings we are looking at the ‘minnow effect’, where unfashionable teams at the bottom of the leagues play host to one of the top four or five teams in the league and raise their game, or maybe the top teams are somewhat complacent and lower their performance, or a combination of both. Another possible explanation is that these unfashionable teams are under-estimated by the bookmakers; or rather their more illustrious visitors are under-priced based upon reputation, players, coach etc. thereby overpricing the hosts in relation to their actual chances, or ‘fair’ odds. Whatever the explanation there seems to be a niche to be exploited here.

# Conclusions

The ratings system produces an average yield of 19%. Any self respecting tipster or system would be more than satisfied with such a return given that it has the potential to generate over 200 bets per season. With three years of supporting data the system can be said to be both reliable and valid, that is the results are repeatable given similar conditions, and that it measures accurately what it is designed to, i.e. the superiority of one team over another, or as one of my favourite adverts on TV used to say about it’s product, ‘it does exactly what it says on the tin..’. Personally speaking, the motivation to develop such a system is not so much about making money but about how to beat the bookmakers at their own game. Of course, it would be disingenuous of me to say that making money isn’t a motivating factor, it is, simple as that, but the satisfaction comes from knowing that I can better predict the correct odds for a football match than the combined forces of the bookmaking fraternity, and given my Dad’s cautionary wisdom, ‘that you never see a bookie on a bike’, then that is no mean feat indeed. In order to help you develop your own system I intend to explore in greater detail the factors that influence the outcome of a football match, what constitutes form, how long form is valid over, how to quantify home advantage and how to quantify the quality of opposition. Additionally I intend to show how to use Excel to both help collate all this information and to do all the number crunching subsequently. Unfortunately a ratings system, a good one at that, isn’t a promise of financial success without the understanding of the importance of a bankroll, of a staking plan or of risk management. Otherwise it all can be a waste of time, effort and of course, money. On a positive note such an understanding can maximize your profits, for example, taking the 2006/07 season and backing all matches with a rating under 30 to level stakes produces a finishing bankroll of 921.4 points from a starting bankroll of 500. However, using a particular staking plan, virtually risk-free, the bankroll swells to an impressive 3407.2 points. Well worth the effort of time and energy I think you will agree.

## Speak Your Mind

You must be logged in to post a comment.