Below is a Q&A with our new NRL Expert, David Barrett.
Tell us a bit about yourself and your background.
My name is David Barrett. I am an engineer with a university educated mathematical background, and a very keen watcher of sports. I am what most people would refer to as a numbers person. The main aspect that I’m drawn to and really enjoy is the mental challenge of building a successful model that proves to be highly profitable. I take a very analytical approach to what I do, and I perform a lot of research to ensure success. In terms of gambling, I have been a positive expected value bettor for three years now utilising this NRL model. I like to pursue a mathematical edge through statistics, backed up by logic on my bets, as opposed to betting off a gut feel or opinion.
How did you turn into a successful long-term punter?
After many years fooling around with basic modelling, a few years ago took the leap and purchased a lot of literature and devoted a lot of time to it. There has been a lot of learning on my part and I have relished every opportunity that I have had to grow myself. I’ve found that whilst something new may not entirely change a particular betting approach, it could change one small component enough that it could really revolutionise a model or system.
How did you get into NRL? Have you always been a successful punter on NRL? If not, how did you turn into a successful long-term punter on NRL?
I have always been a fan of the sport, I grew up playing footy for many years, but had to stop after many injuries. It is one of the major sporting markets in the southern hemisphere. I became successful with this particular model through applying some concepts that I had read about in papers and other research, and adapting them to how I thought they could be suitable in the NRL. After a few attempts and countless hours, I finally got to the very strong model I have now.
How do you identify your bets? What are the key areas or statistics you analyse? Where do you find the biggest edges?
I have a model that solely looks for value. Simply put, I look for games where I disagree with the line and try to take advantage of that. Over the past 3 years, results have consistently proven my modelling has been a better indicator than the market.
Why do you believe you do so well on NRL?
I use the game data as opposed to the final score to assess the true performance of each team in each game. As any avid sports watcher would know, in no way does the best team on the field always win. Often the impact of a lucky bounce of the ball, referees decisions, or 100 other factors can result in a team winning a match despite playing worse than their opposition. Using the game data rather than the final score to assess performance is effective as there are a lot of indicators of team performance that the NRL “experts” are simply not talking about or considering as they are not aware of them. I come up with my own line for each game and compare that to the market line to find value. I find that I have quite a contrarian approach to betting, and in my opinion this works very well. I back a lot of underdogs as I feel that these are mispriced. The market overrates favourites and underestimates underdogs. Many who watch the NRL would iteratively know this to be true. Regularly the no hope team beats the hot favourite. A mixture of the salary cap, questionable refereeing quality, regular injuries and a number of other factors make the games a lot more variable than the market expects.
In terms of what information I use for my analysis, I use a wide range of offensive and defensive metrics and use an algorithm to determine accurate team ratings on the back of these numbers. I am of the opinion that the market needs to regress its ratings towards the mean (good teams are overrated and poorer teams are underrated). With these numbers I create a line that the model projects as the true line that should occur if the one particular game was to be played out over a thousand times. I bet on the line market within the NRL. I have trialled the head to head market, however I have found that there is more market inefficiency in the line markets. The edge that I quantify is the model’s projected line minus the market line and I bet that edge in units.
What else do you look at when assessing matches and determining bets?
Price is key. If the model doesn’t see any value, then I will not be betting, it is as simple as that. I strongly believe in not forcing bets. I love the quote “the best bet you make is the one that you never make “.
How will your service work? What types of bets will you send, how many bets per week, unit bank recommended, estimate of weekly units spent?
I will send out the model’s selections every Tuesday. If there is a lot of market movement (such as the line moving sharply in our favour), then I may notify customers of arbitrage opportunities (e.g. middles bets), that can be very lucrative.
What time will bets normally be released?
Midday Tuesday, after all of the major Australian bookmakers release their prices.
What information do you provide with each bet?
With each bet I will provide my rated price/line, the best & second best prices available with Australian bookmakers (second best of Australian bookmakers will be recorded for official results), and recommended unit investment, everything required to place the correct bet.
What are your previous results? What is a realistic Profit on Turnover percentage?
Previous results have netted on average 65 units profit per season at approximately 12% POT. The POT represents and aligns with the overall mathematical edge the model has over the market in my opinion. I have been very happy with my 12% over the last 3 years.
*Note Winning Edge Investments would like to point out that unlike our other services, the results stated between 2017-2019 are based on previous model output, not publicly verifiable results to paying members
Which bookmaker accounts are the best to use? What odds comparison sites are the best to use?
You can’t beat Topsport for value. TAB is a no brainer for whatever jurisdiction you are in to place a bet in person. The more options you have, the better.
Why sell your tips if it impacts the odds you secure with your own betting?
I bet almost exclusively with overseas bookmakers using Bitcoin. Hence I have no corporate bookmaker profile in Australia. The bets I place and where members will be betting means we won’t be competing with each other for prices, and won’t impact each other’s prices. We’ll all be able to get on at the same time at great prices as we’ll be betting in different jurisdictions with different bookies.
How will the odds be recorded?
The 2nd best price of Australian bookmakers only (excluding Betfair and all overseas bookmakers)
Do I bet straight away, or wait until close to the game start?
I bet straight away. More often than not the line closes in the direction that I bet (beating the closing line value is clearly a positive indicator for the model).
Will I be able to get a decent bet on?
The bets are on the line market, so yes, a decent sized bet is possible with a number of Australian bookmakers.
Are bets based on a 100 unit betting bank?
Yes, bets and staking are based upon a 100 unit betting bank. I take a different approach to betting that challenges the norms but is still realistic and protects against variance. I will be making a write-up available for members on this.
Any profit guarantee?
Yes a profit guarantee applies for all annual/season memberships.
How did we validate the model?
We reviewed the model and outputs in detail, noting the below:
Table 1 showed bets placed, the round, the team backed, the closing line, the projected line for that team in that particular game, odds (for my own recording purposes I use 1.87, when 1.90+ is easily obtainable with retail bookmakers in Australia). This clearly demonstrated that the model has beaten the closing line (not just the opening line when there is more value the earlier that you bet). For transparency and to prove it’s edge, we used the closing line to show that the model is beating this comfortably, not just the opening line. We noted the model doesn’t bet before round 4. It uses the early rounds for data validation, as the use of pre-season trial games for teams is statistically insignificant. Hence tips will start in Round 4.
Table 2 outlined the season summary for the model which demonstrated its nearly 60% strike rate for the 2019 NRL season. In the past three years, this model has averaged this 60% strike rate, which to me has demonstrated the inefficiency of the market and the effectiveness of the model. We can demonstrate my edge with mathematics: (60+41)!/(60!*41!*2^60+41) ~ 1.336% chance that the model is just flipping a coin (50% win probability) and getting lucky with the 60 wins, assuming that it should be 50% chance of winning with the line bet. Hence it is a 98.664% that the model has a true edge.
Table 3 outlines the theoretical edge on the season, simply through maths (strike rate - odds^-1). We noted the POT (profit on turnover) was higher than the edge calculation for the season, which demonstrates that the staking methodology was beneficial in making the most of the edge.
Table 4 outlined the teams that were bet on and how many times they were successful. The model favours betting teams that are considered of poorer quality to the market. The model finds very little value in betting the better teams (roosters and storm) of the past decade, as these teams are highly backed and the model itself rates those teams quite highly. In the 2019 season, the Raiders, Tigers, Sea Eagles and Knights were quite successful to follow and they outperformed market expectations overall.
Table 5 demonstrated the number of games the model bet on. The model was able to find value in half of the games of the season. The cutoff for the model to place a bet is a two point difference from the market. This is applied to minimise variance in the season.
Table 6 outlines the performance of each team over the course of the season against the line. Given each team plays 24 games plus finals, it is quite difficult to work in this market, as for one team to win, another must lose. As previously mentioned, the Raiders, Tigers, Sea Eagles and Knights were teams that were most successful for the model. General consensus is that each team should approximate a 50% cover rate for the line across the season, with the teams that have a higher cover % indicating that they outperformed the market expectation for them over the season.
Table 7 demonstrated the end of season rating for each team against an average team (the Knights epitomised the closest thing to an ‘average’ team). The Storm were rated as strong as the Titans were poor.