Betting U – Improve Your Betting With Football Advisor
11 Aug 2019 Jon Roberts
Building A Successful Betting Strategy
The Gateway to long-term profits (when done correctly!)
Over the years we have continually tweaked and optimised the system builder part of Predictology to be as user-friendly and intuitive as possible.
That said, there is a big difference between knowing how to use a tool and how to build a profitable, sustainable betting strategy. And it is this that we will look to address today.
Over the course of this educational guide, we will explain to you exactly how to design a system so that you can have complete confidence in it and it’s ability to profit for you in the long term by applying it correctly.
Potentially much of the advice below will contradict popular system building advice and we hope that you will see the logic in our methods.
By following our steps you will be able to build something that avoids the pitfalls of back-testing, while enabling you to create quality, long-term working systems.
What is a Betting System or Strategy
Developing a successful system or method involves locating the niche trends of profit which evolves from the dynamic created between ratings, filters and form factors.
The rules must be logical if a strategy is to have chance of forward success. For instance, if your system requires Man Utd to have won their previous came 3-0 or 2-1 only. It is clearly an illogical and back-fitted system. It basically says that it is only profitable to back Man Utd when they have won by particular score line in the previous game – the sample size is likely to way to small and what logic is there that a particular score line will have an impact on their chance of winning the next game.
To begin with, you will need a basic idea or a theme for the strategy that you want to create. It may be teams who won their last two games, or teams who’s rating is X points higher than their opponent, or average goals scored and it’s relationship to Under / Over 2.5 goals betting market, or one of thousands of themes that you can test and develop within the Predictology platform. It can be form based, sequence based, league position based, or ratings based or a combination of each.
Great sources of ideas can football books, magazines, forums and general observations that you make.
Once you have a theme you will need to divide your testing data into three roughly equal parts and then use one of those parts to begin building your system with. We will explain what to do with the other two parts later.
For example, you may want to focus a system on the last six seasons of data. You would divide this up into the three groups; Year 1 and Year 2, Year 3 and Year 4, Year 5 and Year 6
It is our strong belief that the number one problem with anyone trying to build a system is that they work backwards. Usually most ideas are born from the observation of current form and betting patterns. The temptation is to test those observations over the current period for instant gratification from the results that will reflect what you already know; that is that your theme or idea currently works. Once users establish that they have this ‘goldmine’ (from current testing) they then back test the idea, usually over all the data at once and then commence back-fitting whatever they can in order to come up with a system.
The more logical and correct method is to start at the beginning of your data and work forward towards current dates. A good reason to start at the beginning is that many ratings go through changes. Ours constantly improve for the better. Using the oldest of your data means that you are building the systems using the worst of the ratings first and moving towards the better, as you would in real life.
(Side note: In certain leagues we have data that goes all the way back to 2003/04, where for other leagues, we may ‘only’ have a few seasons. It is reasonable to argue that results and betting strategies from 2003/04 may not be relevant to how the world of football operates in 2020. Our advice is to pick a reasonable sample size – 5-10 seasons- as your core testing period. If you do this, while also ensuring your strategy has sufficient total bets, you can achieve the confidence you are looking for – more on this also in a moment).
Another issue with some rules is their availability such as betting odds. Having a system based on these is perfectly legitimate and logical, but the problem is that you need to be able act upon these selections.
For this reason, we include “maximum price” and “average price” in our odds filters. The maximum is the best price you could have got from our 20+ tracked bookmakers – while the average is the average of this, naturally – If you base your system solely on the maximum / best price, you are unlikely to always be able to get the best price and therefore your results may be an over-estimation.
When looking at lay systems, the additional challenge is factoring the price of what it would be to lay the bet on the betting exchanges such as Betfair. We use some pretty complex calculations to estimate the true Betfair price when running a ‘lay’ system. However, it get’s harder to assess this the longer the odds and although we are consistently updating our lay bet calculations, it is important to factor and live test this before placing real money behind any lay betting models.
Importantly, you should also consider that the market is never truly formed until the match kicks off. As a result, systems based on the final price can produce a slightly different set of selections post-match.
Number of Selections
There is often a tendency to strip out the negative results from a system test in an effort to create a seemingly perfect system. Beware – There is no such thing as a perfect betting strategy!
Doing this creates a mirage, in that this random group of results (often small sample numbers) are highly likely not to repeat. Other problems exist in respect of small sample systems, in that not only do such systems have too few selections per year, they are also subject to high levels of variance which makes betting for profit unsustainable.
The more selections you have the more likely it is that you have identified a true pocket of profit that is currently under bet by the market. The more selections per week on average, the better and more stable the performance of your system will be.
We would suggest that you only consider systems that contain at least several hundred selections per season. Any less and you are most likely seeing variance that is distorting the results.
There are of course limits to this, you can’t for instance expect thousands of selections from strategies based around a particular team, cup competitions like the Champions League, and in other similar circumstances the number of selections will be naturally limited.
Other factors depend on things such as when you are likely to bet and use your system and your betting preferences.
Just remember that more is better although less doesn’t necessarily mean it doesn’t work.
First Step To Building A Profitable Betting Strategy
As we mentioned earlier; you will need a basic idea or a theme for the strategy that you want to create. You may already have some specific in mind or you could start by jotting down on a piece of paper, a few theories that initially come to mind. Or, you could look at some the example training systems that we include by default in the system builder.
Once you have that, pick the time period you want to assess your system over and divide that in to three parts. The first part being the oldest date period that you have.
At this stage you select and set the main rules and filters of your system in the System Builder.
Once you have done so you will be able to test your new system over the first (oldest) of your 3 data segments by selecting the seasons (years) that you want to test over.
When the test is finished you’ll then have an idea as to how successful the logic behind your system is over the data that is not responsible for the birth of the idea.
Most likely the result will be less than you had hoped for and you will then need to either need to overlay additional logic over your selections to see where you can improve; or start with a new hypothesis.
We have built a number of analysis tools into Predictology to make this progress easier for you.
First we provide you with a summary analysis of your system and it’s performance. Hovering over each section will uncover additional information such as number of bets
If things are looking promising. You are now able to click on Extended Analysis.
On the Extended Analysis, you can first see our proprietary system rating score*. Underneath that, you can see a detailed breakdown of all the key metrics related to your system, by league and season.
– Winning Bets
– Strike Rate
– Implied Odds
– Winning Odds
– Profit Per Bet
– Profit staking 2%
– ROI and ROC
– Longest Winning and Losing Runs
*The rating is an objective assessment of the quality and confidence level of the results of the strategy.
It is calculated based on a series of parameters including number of bets, average odds, ROI, ROC, Edge, and an assessment of skill versus luck. Strategies with a negative profit level will not receive a rating”
At this juncture, the analysis will be able to point you in the four different directions.
1. Everything looking positive – move on to stage 2
2. A particular league or country is looking problematic – test further in stage 2
3. Add another logical filter to Batch 1 and test again
4. Scrap the idea and start a new hypothesis
Option four is self-explanatory.
With option 3, you may wish to tweak your existing filters or export the data for offline analysis. Beyond this, you could add one-two more rules to your system (no more than that) to refine further. Important, only change or add one rule at a time.
Ideally, what you are looking for here is the change that will have the LEAST impact. By that we mean the option that will eliminate the least winners and the least selections.
The natural tendency is to focus on profit over selection quantity but we want to avoid that where ever possible. By avoiding this you will most likely retain far more selections in your system and less rules than doing it the other way round.
As mentioned, apply one change at a time before re- testing and re-analysing your system. We have optimised the platform so that systems generally only take a few seconds to a minute to populate, so it is very quick and easy to test many things, one at a time.
When running each test, keep a note on the total volume of bets and total number of winners. If a change makes a significant difference, you may want to make a note of that filter, remove it, and try something with LESS of an impact.
Also, if you find a filter that may have promise but you want to keep testing other ideas. We have added an easy duplicate option, allowing you to keep your original and not lose it’s settings, while using the duplicate(s) to test additional theories.
For options one and two, we are ready to move into Phase 2.
Now that you have your potential strategy in a shape that you are happy with, using the first oldest segment of your data, it’s time to test how this method will likely perform moving forward.
Run a test over the second oldest segment of your past data and compare the results. (You may also wish to use the system duplication option for this so you can easily see both sets of data side-by-side).
If the results are similar then you are likely to just move to the third Phase. If not then you will need to see what is causing the change by detecting the problems before taking steps to rectify them.
The reason for running the data over the second segment as a separate test rather than clumping it all together, is that the two separately will present a more accurate picture of any performance issues and allow you to focus onto what matters. For instance, you may think that the English Premier League is really profitable for this model – however, when you look at it by season, you realise that, although profitable, it was the profit from Batch 1, that covered the profit of Batch 2.
This also helpful metric when testing additional filters as you can quickly see how they interact Batch 1 and Batch 2 independently.
The other benefit of doing it this way is that you will see and get a feel for a real life (forward going) performance of your system, importantly one that doesn’t cost you in losing bets. (Remember to check out the Longest Winning and Losing Runs, plus the Drawdown in the Extended Analysis section for a greater picture of this).
Using the comparison method you are likely to extend the number of selections and drill down to filter settings that will produce a better result. When you are finished then run a final test this time over the two combined periods, again giving consideration to duplicating or saving your system details before moving into Phase 3.
Phase 3 – Testing, Comparing, and Finalisation
The final Phase should, if we have done Phase 1 and 2 correctly, be more a formality.
Run a test over your third segment and see the results. Hopefully you will see a result that will please you in that you really don’t need to do much else other then analyse and look for any possible obvious anomalies which can then be eliminated. (E.g. selections that are priced way above what you would bet at).
Compare these with the Phase 2 analysis and decide if you are going to take any actions. Apply the same elimination process, one filter at a time as before.
You should have a hard look at any systems that need a lot of extra work at this stage and ask yourself “How likely is it that this will continue?” If you are forever tightening up the rules, will this continue into the next stage where you will pay for each loser? Some systems are just not meant to be. The less time that you spend working with your system here in Phase 3 and even Phase 2 the more likely it is that you will experience similar results when you are betting.
Now run a final test over your entire period of past data and enjoy the results of your labour. You will now have a system that is likely to continue to perform into the future in a similar manner as it has when testing Phase 3 for first time.
By following the steps in this guide you will have take all reasonable measures to ensure that you have a system that should be sustainable, with solid logic, and that has a good chance of continuing to deliver profits into the future.
Remember the very best betting systems and strategies are logical and have the least amount of rules as possible.
One of the greater issues with many systems is that they have few selections, as we now know, the more selections we have the better, however less isn’t necessarily bad.
The one major problem with such systems is the expected performance and the reality of mathematical variance and your mental capacity to handle this long term. Let’s say that you have a system that produces 50 bets a year and has 50% strike rate. Your expectation is to win every second bet, however such system could go through a losing run that is 11 bets long!
Based on one bet on a Saturday, you could actually go through nearly 3 months of not winning a single bet. Having 11 losses in a row is very real possibility! Actually it is an inescapable reality, IT WILL HAPPEN AND IT WOULD BE QUITE NORMAL BEHAVIOR for a system with such a strike rate regardless of what rules the system uses and whether you believe it or not. Don’t believe us? Search “Expected Losing Runs” on Google and you will soon see…
Ultimately, by following the steps in this guide you will have take all reasonable measures to ensure that you have a system that should be sustainable, with solid logic, and that has a good chance of continuing to deliver profits into the future.
Remember the very best betting systems and strategies are logical and have the least amount of rules as possible.
To your betting success from the team at Predictology.