We decided not to remove this post from the site because of the many questions and answers in the comments section. For those of you with curious minds it may also be quite interesting to see how our HDAFU Tables have developed over the years.
Read about: The Power of the 7th Generation of The HDAFU Tables
Our HDAFU tables have evolved tremendously over the years. They are a complete statistical analysis of historical performance over the previous five seasons of the Home win, Draw, Away win, Favourite win and Underdog win (H-D-A-F-U). They serve to identify the most profitable odds ranges in each bet type.
To help you understand why we value this product so highly, here is our Definitive Guide for using the 5th Generation tables to their maximum potential.
It’s difficult for us to put into words how important the HDAFU tables are to us and our own betting adventures. But what we can say is that we have complete confidence in them to do their job. And from testing them in a live setting, we know that they are an extremely reliable method of building lucrative betting portfolios.
Quite simply, they are the best and most user-friendly tools available for nailing down value betting systems in every league you apply them to.
They reveal the DNA of a league, and provide a hidden level of detail that makes finding and exploiting the sweet spots so easy and so rewarding.
The next six steps will probably change the way you think about betting…
Step 1 – Observation
By default, the HDAFU tables open on the Data Tab showing the financial summary of each bet type. Here are the figures from an example league.
(Click on the image below to enlarge it in a new tab):
HDAFU Table Data Tab: Whole Season Results Summaries
The totals along the top row show the effects of betting on every match over five seasons. In our example league the totals are (from left to right) Home win (-7,329), Draw (-835), Away win (+8,236), Favourites (-2,594), and Underdogs (+3,501).
You can see from this graphic that away wins look the most promising backing system with a profit of 8,236 units from 100 unit stakes.
To customise the stake amount enter what you want in the Fixed Stake box at the top of each bet type in the Data Tab.
The image above shows the full five season cold analysis. If you enter a different stake amount the financial values will change, but the percentages will always remain the same. This being the case, we have fixed these percentages as a benchmark to better gauge the improvements we will make with our filtering exercise later.
The Odds Toggle is for testing the effects of the odds you are getting when playing the systems for real – you can ignore it during your analysis.
You can also leave the betting exchange commission rate at zero. Again, use it for backing system monitoring purposes when you start betting on or paper testing your systems of choice.
Okay, we fancy away wins in this particular league but let’s now have a look at the Inflection Points Tab to see if this backs-up our observation.
(Click on the image below to enlarge it in a new tab):
HDAFU Table Inflection Points Tab: Inflection Points Overview
Away wins are certainly financially the most profitable bet type but the profit curve doesn’t really begin rising until odds of 3.30 are reached. Overall profit at this point is 463 units and this rises to a peak of 13,502 units around odds of 8.60.
These two points on the graph would therefore be our two inflection points: Odds of 3.30 where the curve begins to rise; Odds of 8.60 at the pinnacle, the point at which profits begin to fall again.
However, notice there is a big portion of the away wins curve which is a zero-sum game. This ‘hole’ in our profit curve begins around odds of 3.75 (6,653 units). At this point, the curve falls away again, encounters what we call ‘statistical noise’, and only recovers at odds of around 6.52, when the profit figure surmounts its previous high at 7,184.
HDAFU Table Inflection Points Tab: Inflection Points Odds Intervals 3.75 – 6.52
In between these two points is the potential for a lot of wasted effort and not a lot of gain.
We can see the extent of this by scrolling down and looking at the inflection point intervals.
(Click on the right-hand image to enlarge it in a new tab):
This image shows the start of the 3.75 odds sector at the top and the end of the 6.52 odds sector at the bottom.
The yellow column indicates the running total of matches up to each cluster of matches.
We can see that our two odds of 3.75 and 6.52 encompass roughly 330 matches – the difference between 1,115 indicated at the 6.52 break-off point and 785 at the starting point of the 3.75 cluster.
That’s 330 bets over a five season period that are simply not worth making; or 66 bets in a season.
HDAFU Table Inflection Points Tab: Inflection Points Odds Intervals 3.30 – 8.60
We can see this clearer by looking at the same snapshot between our original inflection points of 3.30 and 8.60.
:(Click on the left-hand image to enlarge it in a new tab)
In this odds range, we have roughly 587 bets (1,221 minus 634). We now know that more than 56% of these (330 bets) are not worthwhile making.
This leaves only 257 bets but the away win profit sectors between the inflection points seem to be split into two areas of the curve: from odds of 3.30 to 3.75 (medium risk system, accounting for around 160 bets), and then from odds of 6.52 to 8.60 (high risk system; around 100 bets).
If we were to continue our analysis of away wins we would eventually see that the three elements (the medium risk sweet spot, the high risk sweet spot, and the statistical noise in-between) combine to give us a bumpy ride.
Our expected hit-rate will be tempered by that area of noise, and yield will be lower because of the size of the zero-sum area and the number of pointless bets within it.
This means a lot of unpaid work to perform, placing many bets that maintain the status quo and not much else. On top of this, the losing streaks will be greater.
Therefore, why not split into two systems in this league?
The synergy we have mentioned before about many systems supporting each other is what makes the HDAFU betting systems so viable.
However, we also mentioned that you should find the single best system in a league to play alongside the other best systems in the other leagues within your portfolio.
In our away win example, we would need to choose the better of the two systems we have identified. Either backing away wins at odds between 3.30 and 3.75, or between 6.52 and 8.60. Choose one or the other, not both.
We recommend never to play multiple systems in the same bet type. The synergy effect is diminished as ultimately, one of the two systems is not the best we can find.
Ideally, we are looking for synergy between the absolute single best systems in each league within our portfolio, without creating a situation where one system supports another within an individual league.
With different bet types in the same league (e.g. 1×2 market and over/under goals market) this is not an issue, but we would go as far as avoiding the conflict of interest between HT and FT 1×2 systems in the same league, for example.
Away wins initially looked great but is there something better?
Have a look once again at the Inflection Points graphs to try and see what it is.
As is typical of an underdog backing profile, the high risk/high return nature of this bet type produces a noisy curve, one full of jagged peaks and troughs. There are only small rising areas to analyse. Anything you can analyse into promising profits will contain few betting opportunities in a season, with long runs of losing bets to cope with.
Backing the favourite has one area between odds of 1.90 and 2.10 but we can see at these odds not a huge profit is created over five seasons (less than 3,500 units).
The home win is a misery for backing. Again, the sweet spot is between 1.90 and 2.10 but the profit is less than 2,000 units.
That leaves us with backing the draw. There is a large, rising area in the curve beginning at draw odds of 3.32 (-2,008 units), and peaking at 3.65 (7,170 units). It represents a potential profit chunk of 9,178 units over five seasons.
This is better illustrated by superimposing our inflection points onto our graph – We are interested in only the portion of the curve in-between the red arrows:
The shape of this curve is what you should be looking for when identifying the first system to analyse in your leagues of choice.
It is the classic, gently rising curve from bottom left to top right. It is relatively smooth, with a far smaller amount of statistical noise.
Therefore, this is the bet type we will analyse as our example.
Next Page: Step 2 – Hiding, Sorting & Filtering
One question though. How can I obtain Betfair ‘s odds from the data?
Thanks in advance.
I have a few questions regarding the HDAFU tables:
– Would it help to add in more seasons to the analysis?
– Would it help to weight the seasons?
– Do you backtest your picks? For example, omit betting on the first 25% of the season, use these matches to validate, and then start betting on them. And if so, how about splitting the season in half?
Thanks!
Thanks for the good questions, JVR,
Yes, we performed literally years of analysis before settling on five full seasons of data. It doesn’t seem to make a significant difference adding any more. In fact, adding further seasons tends to dilute the data pool with information that becomes more irrelevant the older it is.
So, five seasons is enough to provide a data pool which carries statistical significance, and that’s all we need when analysing any snapshot of past performance. You can add more seasons if you want (and create more work for yourself), but the fringe benefits of the extra stats become more and more negligible the further back in time you go.
And don’t forget, maths is not an exact science. Our aim is to get as close to the answer as we can, knowing that there is never an absolutely correct one.
My feet are different sizes for example. Most peoples’ are. Almost all people buy their shoes from a shop. We buy shoes that fit as best they can but they are never a ‘perfect’ fit for both feet. But they do the job they are intended to do. This end result metaphor is all we can ever hope to achieve with a statistical analysis – it will never be pinpoint accurate is what I am trying to say.
Weighting the seasons again doesn’t make much sense as we want to compare ‘apples with apples’ and see how they stack up against each other on their own merits. By doing this, we get a more rounded picture of what is likely to happen in the future, without any bias being placed on past performance.
For example, there is no point identifying an anomalous season and compensating for it. We want to see the effects of that season in all its ugliness as such a season could quite easily repeat itself in the new season we are analysing for.
I could write a book longer than the Bible with the amount of back-testing we have performed over the years, and the benefits of this are that we now split the seasons at their natural mid-season interval. The 2017 Summer League HDAFU tables started this trend and in a few weeks, when the Winter Leagues end in Europe, we will again be providing three tables for each league (the whole season analysis, first half split, and second half split).
Hope this helps and thanks again for contacting us!
Hi Right Winger,
how do you determine if betting opportunities or hit rate has too big deviations?
Hello again Jo,
I’ve been trying to backtrack for the answer to your question as, after many years of analysis and betting, identifying deviations almost becomes second nature.
Firstly, I think that ‘anomalies’ is probably a more useful word to use than deviations.
The number of potential bets covered by a system and the expected hit rate from it are usually pretty uniform when looking at a filtered set of data from five seasons.
Anything that stands out from the pack should be looked at more closely. Was it an anomalous season that caused the deviation? Is there a history of anomalous seasons in that particular league? Are there any other reasons for the anomaly?
Certainly, if the betting opportunities are anomalous in one of the seasons analysed this can point to a change in odds setting for that season, perhaps due to the composition of the league (i.e. an historically strong team is relegated and replaced by a far weaker team, which has a knock-on effect in the odds compilation).
Filter the results in your chosen system by team, and see if there any anomalies there. You may even be able to filter out certain teams from your analysis – ones that continually let the system down, for example. See if this makes any difference to the deviations you can see.
I hope this helps!
“The number of betting opportunities and the hit-rate should both show fairly consistent figures throughout the five seasons.”
How do we know that it’s consistent ? Minimum number of bets for a season ?
Hello again Sellig,
By consistent we mean that the numbers of bets in each of the five seasons should be more or less equal.
This also applies to the hit rate, which should be fairly even over the five seasons, with no huge anomalies.
Hello,
Great Job Right Winger !
How many bookmakers do you select on Oddsportal to find the best odds:
only premimum bookmakers automatically selected by Oddsportal;
Every bookmaker;
Another selection?
Thanks a lot.
Hi Sellig,
Good question.
We have no choice with those automatically selected by Oddsportal as they do not allow you to deselect them.
We add a selection of the more reliable bookmakers to this list, but not all of them.
We scrape our odds from a selection of around 30 bookmakers in total (including the compulsory ones).
There is a problem with Oddsportal where odds do not update from certain bookies, or sometimes they update after the match has started with in-play odds, which have to be discounted.
Our odds are therefore carefully checked and compiled – it takes hours to scrape the odds we want and to check that they are viable.
I hope this helps!
Hello Right Winger,
how do you deal with championships that uses a playoff format in their season? Do you analyze only the regular season, or maybe you analyze every playoff separate from the rest of the season?
Thanks for your answers!
Hello again Daniele,
We don’t include play-offs or any other non-regular league matches.
Play-offs are akin to cup matches and are far more volatile than the regular rounds of a league season.
We also prefer to deal with the premier leagues in each country only. In lower leagues, there is too much statistical noise created by having promotion at one of the table and relegation at the other.
The content of lower leagues therefore becomes far more diluted by the fact that there are many more teams competing over a five season period than in the top leagues where only relegation is a factor in altering the statistical composition.
Hope this is self-explanatory, and thanks for your question.
Hi Right Winger,
what do you do if offered price in the last hour before kick off time is a little bit out of your desired inflection points range? For example, inflection points are 1.5-2.5, price 1 hour before kick off is 2.55. Do you come back later to check again if the price moved in your inflection points range, or do you check the price only once?
Jo,
The last hour before kick-off, when the large majority of our bets are placed, can be a very hectic time depending upon the number of bets to be considered for inclusion.
At the other end of your question, if a price is a point or two below our first inflection point, then we will use an exchange to request a higher price (taking commission into consideration). If it gets matched then fine, if not, then it usually remains outside our spectrum.
With prices that are slightly higher than our second inflection point, we have to keep an eye on any movement. We check the prices constantly during the final hour. If you adopt this habit you will quickly see which ones are active (and more likely to move) and which ones remain stable.
But we are not talking about huge numbers of bets to monitor. It might be one or two in a league. The rest will already have been placed (as well within our range) or discarded (as well outside our range).
It’s a valid question and, with time, you will develop enough of a sense to make these kinds of judgements. For us, it was like being in a maze, exploring avenue after avenue until finding the shortest path to our goal. Our sense of direction has therefore been honed over several years.
Hopefully, we have saved others a lot of this time already by outlining exactly what our current processes are.
Jo,
Yes, the two new teams are mentioned in the MLS table (Backing by Teams tab).
I really don’t think it will make much difference. The MLS is definitely an unique league from the perspective that the distances between teams equate to many long trips for the away teams.
I truly believe this is a huge contributory factor to the percentage of home wins in this league. It’s always high compared to other leagues, and I expect it to be high again in 2017.
We will see if there is any noticeable change at the end of 2017, but I am really not expecting any surprises.
With other leagues, I don’t think it makes much difference, but we do have a tendency to rely more on the whole season analyses for leagues that experience expansion or contraction, rather than splitting their data into first and second halves of the seasons.
Hello again Jo,
Thanks for your comment regarding Matchbook.
Soccerwidow has just informed me that she is aware of their problematic commission rate structure (I have to admit it was news to me), and that Smarkets is currently the only exchange that we use. I’ve therefore amended my comment above to reflect this. Thanks again.
Regarding Marathonbet, we have also read a few negative stories but with stakes as small as ours we haven’t encountered any problems yet.
But bear in mind that negative comments are usually the only ones you read about, and never the plaudits of the 99.9% of satisfied customers.
I think it also helps that the 2016 portfolio hit-rate overall was less than 50% and substantially less than this with Marathonbet (as they are underdog specialists).
For definitive opinions on bookmakers we use: Sportsbookreview
Historically, we have used exchanges to withdraw bookmaker winnings and we may do so again in the future, but following the 2016 campaign, we had money spread around various accounts and had not withdrawn anything.
The overlapping 2016-17 Winter League programme of course utilises these funds too.
Of course, if we encounter problems we will let our readers know but, for the relatively small amounts we have been betting and winning, things have been pretty plain sailing so far.
However, I am sure that as our portfolios grow larger and larger we will experience choppier waters, but the thing is we don’t necessarily need to increase the stake amounts to ridiculous levels: We can instead achieve enhanced results with a greater number of bets.
Hi Right Winger,
it’s always a pleasure to read your comments.
Today I learned some news about MLS – it has expanded to 22 teams. And there plans to expand it to 28 teams in the future. It’s definitely not the first time when you deal with situation like this. I checked more previous seasons of MLS and it looks like it gets expanded every 2 seasons in the recent years. How leagues expansions affect betting results and HDAFU tables reliability?
Hi
Very interested in your tables but I have a question.
You use the odds and place the bet one hour before the match starts, but which bookmaker do you use? There can be a difference in odds between bookmakers that could either make an unwanted selection part of the system or take a desirable bet out.
Kind regards
Timmyp
Hi Tim,
Thanks for your question.
With such a small starting bank, we had to keep our initial number of accounts low to ensure there was enough money in them without unnecessary transfers between accounts.
As the bankroll grew, we were able to spread the money around more accounts. Right now, we use just one betting exchange, Smarkets, but I would say that they account for less than 10% of our bets.
The multi-bookmaker platform, Vodds, definitely comes in handy (includes access to Pinnacle), whilst the rest of our portfolio of accounts includes Marathonbet, Bet365, BetVictor and 5Dimes. We will undoubtedly add others as bankroll grows further.
Regarding unwanted selections, this really isn’t a concern. If we can get a price that sits within our two inflection points then that’s job done. It doesn’t matter if it then drifts out of one end the scale.
The important thing is that a bookmaker was prepared to offer odds within our threshold at a time which was close to our analysis point (i.e. within the final hour of the ante post market). Of course, it’s of even less concern if the odds we buy are above our zero odds for that system (i.e. a value bet).
As mentioned in the 2016 Campaign article comments section, we are scrutinising whether a wholesale approach to picking odds within our inflection points threshold as and when they become available is viable (rather than limiting ourselves to the final hour of the market).
We will let you know when our exercise on this question is finished at the end of 2016-17, but for the time being we are still telling everyone to do their best as close to the end of the ante post market as possible.
Thanks for your question and your interest in Soccerwidow.
Hi Right Winger,
Matchbook has complicated commission charge – 1.5% for accepted and 0.75% for posted odds regardless you win or loose. If you win you pay 1.5% or 0.75% from your profit, if you loose, you pay 1.5% or 0.75% from your stake or potential winnings whichever is lesser. How do you calculate real odds for Matchbook after complicated commission fee is deducted? Formula would be very helpful. I didn’t find it anywhere.
Also, have you had any problems with Marathonbet, their odds are probably the best in the market, but I read a lot of reviews how dodgy this bookie can be. You wrote somewhere in the comments that you withdraw winnings through betting exchanges using matched betting technique. But with dodgy bookmaker you might win just a few times and they won’t give you the chance to show loosing pattern. I would like to have an account there, but after reading reviews online I don’t know is it worth the risk to get money frozen for I don’t know how long or even confiscated for some hypothetical “problem”.
As usual, I make some progress with my excel tables to accelerate the filtering process. 🙂
Glad to see your new articles in soccerwidow site and congrat with the 2016 summer leagues campaign.
With your new free excel sheet’s datas, I investigate the Paraguay league and found far greater profit amount in the underdog market than in the draw market. What’s your opinion about that?
But my main question: Over the years have you found an automatic odds scraper process to download the odds form oddsportal?
Sadly, I just use the odds from betexplorer, but the odds from this site (also higher odds) are smaller than yours from oddsportal.
Hi VB,
Glad to see you understand what you are doing.
Yes, underdogs look good in Paraguay too, but it’s a high risk scheme with potentially large losing streaks. Just be careful of including something like this in your portfolio – there are an awfully large number of bets, which could see your longest expected losing streak occur more than once.
Personally with underdog systems, we prefer shorter runs of potential bets, with not so many needed to be correct!
Regarding odds, we still scrape from Oddsportal. Getting a whole season’s data for one league such as the EPL (380 matches), takes approximately 12-15 minutes.
Betexplorer odds should be similar to Oddsportal as they are (or were) run by the same company, although I have noticed some differences since Betexplorer recently changed the theme and appearance of their website…
Thanks for your question!
Hi Right Winger,
in Summer league Inflection points simulation table of MLS 2016 season you bet at odds of 1.84-3.71 Home win. After reading this article, I understand that zero odds must have been below 1.84 after sorting, filtering etc. of 11-15 MLS seasons data. When I do my own sorting, filtering etc. of 11-15 MLS seasons I end up with 2.105 zero odds. Since it’s above first inflection point value of 1.84, that would suggest this system is not going to work. Am I missing something? Can you tell what was your zero odds after you sorted, filtered and etc. 11-15 seasons of MLS?
Jo,
Thanks for your important comment. I have amended the article as a result.
Zero odds should ‘preferably’ be below the first inflection point. It’s more important to get close with this criteria on your medium and higher risk systems such as the draw, away win or underdog.
With home wins and favourites where the hit-rate is higher, it is not so vital as you will usually have many more bets to contend with and the weight of numbers should be good enough to give you the hypothetical value from concentrating on the sweet spot.
Yes, USA was 2.10 zero odds as you mentioned, but achieved a nice profit regardless. Again, the 100 lower priced (i.e. below 2.11), higher probability bets (50 won) provided the synergy we needed to get into profit with the 142 higher priced, lower probability bets (65 won) in this league.
Nice profit in this league during 2016 season, but look at the low yield of only 5.59% – below our expectations partly because we surrendered some of our value by having the zero odds point for this league in the middle of our two inflection points. But, the hit-rate was higher than expected suggesting that the odds we achieved were worse than expectations.
Every league will be different – all we hope for is to be there or thereabouts with the majority of our systems at the end of the season. The chances of all the systems failing was less than 0.01% (i.e. no chance at all). The chance of all the systems being in profit was also relatively small. But, like the standard distribution bell curve, we will end up somewhere in the middle (if the analysis is sound to begin with!).
Thanks again for you work Jo!
Hello
Ive seen that you are using the highest closer odds at the market to look for the inflection points. Since i switched over to use the asian bookmakers – should it be possible to only use the pinnacle odds from football.data?
Which leads to my second question. You are using the CLOSER odds. So wouldnt it be necessary to only bet the closer while doing this system? Especially the summer leagues would be a problem since a alot of matches start while europe is asleep.
So i would take the pinnacle (not pinnacle closer odds) and since theire collected on fridays for weekend matches and on tuesdays for midweek matches i would only place my bets at these days. Unfortunately football.data only provides us with winter leagues.
PS: What would you do if only bookmakers you dont have an account with offer an odd which is within your inflection points and all the others you use are over/under the needed odd? Would you still bet on it? Of course, this problem wouldnt occure only using pinnacle odds
thank you for answering 🙂
MB,
By CLOSER odds I guess you mean the odds just as the ante post market closes? (and not closer as in ‘nearer’?).
I suppose it really depends on your commitments but we certainly have no problem with the kick-off times of games. This is a full-time occupation for us, so we adjust our working hours to suit.
If you want to use Pinnacle odds, then you will have to use them in your analysis and re-program the HDAFU tables (good luck there!).
BUT, we no longer use football-data.co.uk odds for the tables because of the irregularities of when they are recorded. It is impossible to be consistent with any analysis when odds are taken at two separate times per week, with these ‘frozen’ figures sometimes three days old before the match even kicks off.
It would be a huge juggling act to get the analysis correct and then bet on the games around the same times the odds are taken. There is too much time in-between for odds to fluctuate wildly.
With betting in the final hour before kick-off, we get the advantage of seeing odds change with the team news as it is released, which is always a major factor in swaying public opinion and influencing the weight of money placed on certain outcomes.
Whatever you choose to do, just make sure you are totally consistent with both your analysis and then your follow-through.
Thanks for your question and I hope the reply makes sense!
hi ,
in this example, we will back the draw between 3.33 and 3.63 odds range and i found the match in draw odds of 3.33 before one day kick-off. i place the bet and in kick-off day for example the odd change and become 3.32 may be 3.31. how can i do now?thanks
Hello again Zinphyosein,
We place our bets within the hour before kick-off. If a bookmaker is offering odds within our range, we place the bet regardless of whether the price then drops outside the threshold.
If a bookie is willing to offer a price we need within the final hour of the ante post market then that is good enough for us.
What is more of an issue is seeing a price just below our minimum requirement, which never rises. In these cases, we sometimes use an exchange like Smarkets or a multi-bookie platform such as Vodds to request a higher price.
If it gets matched then all good and well, but if not, we cancel the bet as the match goes in-play.
Hope this helps!
hi,
i think there may be typing error .
backing home sweet spot is between 1.90 and 2.10 in inflection points graph and in article is between 1.90 and 3.10 for home backing.
Thanks for that Zin,
Duly corrected!