In order to value betting on draws, I needed to buy overpriced odds to get value from the misjudgments of Mr. Market. I had to compile my own odds to bet on much higher odds than I estimated to take advantage of bargain prices with a good margin of safety.
To get started, I just needed a KISS.
A “Keep It Simple, Stupid!” approach was my starting point to compile the odds from an unemotional and straightforward perspective.
At the time, I was in love with Moneyball and received its KISS from the first pivotal scene, in which Peter Brandt said to Billy Bean:
People who run ball clubs, they think in terms of buying players. Your goal shouldn’t be to buy players. Your goal should be to buy Wins. And in order to buy wins, you need to buy Runs.
Replace the word “Runs” with “Goals”, and that’s it.
In order to buy Wins, Football clubs need to buy Goals.
Goals scored and prevented, of course.
If you want to get Wins, you have to score one goal more than the opponent and get the three points to advance in the ranking.
Where do most of the goals come from?
85% of the time they come from shots in the box.
Thus, a football team is focused on taking as many shots in the box as possible to score goals and on avoiding shots in the box against to prevent goals.
So, to compile my own odds and assess the team performance, I used the Total Shot in the Box Ratio (TSBR), by dividing the number of shots in the box taken by a team by the total shots in the box overall.
TSBR = ShotsintheBoxFor / (ShotsintheBoxFor + ShotsintheBoxAgainst).
I started doing this collecting data from the 2011/2012 season of the Italian Serie A, the league I knew best, calculating for each team the Home and Away TSBR.
I just had to accurately group the teams according to their TSBR rating to know how the results usually play out when teams play one another. In this way, I had the number of wins, draws, losses, and the average expected goals for each fixture type.
With the average expected goals, I used the Poisson distribution to verify the correspondence of the percentages of wins, draws and losses with those I calculated with the TSBR rating method.
During the process, I realized that the Poisson distribution underestimates the draws. Thanks to the complete picture that I had available for each fixtures types, I could start to backtest and build my value betting strategies verifying the hypotheses that came out by observing the expected number of draws.
Looking back for moving forward
The advantage of betting on draws comes not only from the possible misjudgments of their odds by the bookmakers when using the Poisson distribution.
Thanks to some typical behaviour patterns I mentioned in the first part of my investment thesis on Football draws, even amateur bettors, big spenders and pro tipsters can make football draws a value investment.
Their recurring behaviours have survived to these days and there are no changes on the horizon.
They keep moving in their comfort zone.
The nature of their behaviours does not change over time.
As a result, I think that those behavioural patterns will continue to shape the future.
Therefore, all that remains is to pay attention to the past for observing what has survived.
The past is the best indicator to develop long term value betting strategies on draws, just like it happens in the business world as Warren Buffett points out:
The rearview mirror is always clearer than the windshield.
How did I build my Value Betting Strategies?
Before building my betting strategies, I had several hypotheses to test, thanks to the analysis I performed after compiling my odds. I also had a large sample size, which included the last eight seasons of the Italian Serie A. Finally, I had compiled my odds on the TSBR, an objective metric for evaluating the performance of a team.
It was a good starting point for avoiding the risk of Overfitting when building betting strategies.
It occurs when a strategy is so complex and filled with too many parameters and variables, that you risk adapting your data in a personalized way. Overfitting also occurs when you create rules which are specific to the dataset you use in the attempt to increase the profitability of the system.
To avoid overfitting, it is necessary to take two more steps by keeping the strategies simple and through cross-validation.
In the first case, you take into account a few rules to avoid the noise in the training data. Secondly, you take random subsamples from the larger sample size throughout the development of the strategies, by performing observations with different subsets.
It was useful, for example, excluding some of the most recent seasons from the entire sample size set and verify how the strategies have worked in the far past. Then, the strategies must have also worked in the recent past to prove survival to these days, and it was useful to cut the oldest seasons too.
To make these observations more accurately, I enlarged my data sample from the Italian Serie A and used a database with more than 100,000 matches played in the top 50 leagues from 2012 to 2019.
Also, my other goal was not to fall into the Cherry-picking trap.
Cherry-picking occurs when figuring out specific spots to confirm a particular hypothesis while discarding a significant portion of related data that may contradict that hypothesis.
So, I tested on all football leagues to prove large-scale dissemination of my ideas and not selecting only the leagues useful to validate a strategy.
I did the same, involving all the football teams for each league.
Finally, when building a betting system beyond an underlying theory and having a large sample size, you need also consistent results over time, guaranteed by a high number of bets per year.
In this regard, I would like to show the graphs of some strategies I have developed so far, with the most representative leagues.
High Volume Strategy
Low Maximum Drawdown
High Yield Strategy
Where do the results come from?
The results of my strategies come from the Pinnacle closing odds.
When you have results calculated on those closing odds, it’s like having a hallmark on the true final prices of a football game, because they represent the most efficient odds on the market; since they are shaped by the weight of the money wagered by the sharpest bettors before the game kick-off.
In fact, Pinnacle attracts the highest liquidity and allows sharp bettors to wager a higher amount of money. It is a bookmaker with a low margin operative model with increasing betting limits until the start of the game. Therefore, it attracts high-level bettors with large bankroll whose bets lead to efficient odds on the market closing, due to this smart volume of money that moves the odds line towards the real prices.
Conclusions
Once you have underlying hypotheses, a large sample size and consistent results years-by-years, your strategies are more likely to succeed for a long time into the future.
I have to admit it was a funny and exhausting process at the same time.
It was like taking pictures in a long photography session.
I observed the draws from different angles with new eyes for every shoot.
It was like taking snapshots of a landscape at different times of the day.