Following up the successful 2016 Summer League Campaign, here is our complete report on the 2016-17 Winter League Campaign, based on 18 different top flight leagues.
2016-17 Winter League Profit Curve
We will try to avoid repeating what was said in the summer league article so that it is all new information for you here.
For the sake of completeness, you should still read and digest both reports for a full idea of our strategies and thought processes.
And you will find the current stock of available HDAFU tables via this link.
2016-17 Campaign Report
Measures of Risk
Before beginning to plan any portfolio or placing bets, you will need to review your analyses and rank the systems you have found according to risk exposure based on the values of the upper inflection point odds.
This will allow you to compile a portfolio with a healthy balance of risk, which is essential to the success of any investment plan.
Here is our rough guide:
- Low Risk (Probability 45.00% or more; upper inflection point maximum odds of 2.22)
- Low-medium Risk (Probability 44.99%-35.00%; upper inflection point maximum odds of 2.85)
- Medium Risk (Probability 34.99%-22.50%; upper inflection point maximum odds of 4.44)
- Medium-high Risk (Probability 22.99%-16.00%; upper inflection point maximum odds of 6.25)
- High Risk (Probability 15.99% or less; upper inflection point odds above 6.25)
Discretion is used if a system does not fit these parameters or crosses two or more classifications – In these cases, the harmonic mean odds of all the games in the set is used as the benchmark to guage risk. The Excel formula for a range of odds in cells A1 to A100 would be: =HARMEAN(A1:A100)
Measures of Success
You will also need a definitive framework to be able to judge the final results.
For us, the final results of any league fall into four distinct categories:
- Systems that achieve a six-season-high (i.e. profits larger than any of the previous five seasons). (Over-Achievers).
- Those that make a profit over and above the size of the initial stake (£100 in our case), but fall short of six-season-high results. (Achievers).
- Strategies that break-even or, record a tiny profit or loss up to the size of the initial stake (£100 in our case). (Zero-Sum).
- Leagues that make a loss over and above the size of the initial stake. (Losers).
You can already see that two of these outcomes are favourable, one is neutral, and only one is detrimental.
2016-17 League-by-League Review
Let’s have a brief look at each league to see how our 22 systems fared. (Alphabetically according to the tab codes in the workbook):
1. AUS1 – Austria Bundesliga – Whole Season System
Only eight of 28 bets won, but this was enough to see a profit of £882.00. Hit rate and yield were both below the calculated averages and the resultant profit figure was lower than any of the five previous seasons’ figures.
Result: Achiever
2. BEL1 – Belgium Jupiler League – Whole Season System
Up to, but not including Europa League and Championship Group splits
Both the estimated hit rate (33 out of 97 bets won) and yield figures were surpassed by wide margins, leading to a six-season-high profit figure of £3,108.00.
Result: Over-Achiever
3. CZE1A – Czech Republic 1. Liga – First Half Season System
Up to the 04/12/2016 winter break
Nine out of 22 bets won, and whilst the hit rate and yield both outstripped expectations, the profit figure settled at £1,247.00, the fourth largest in the last six seasons.
Result: Achiever
4. CZE1B – Czech Republic 1. Liga – Second Half Season System
From 18/02/2017 start of the second half of the season
This one suffered its worst result in the last six seasons, but the resultant loss was minimal at -£310.00.
Result: Loser
5. DEN1A – Denmark Superligaen – First Half Season System
This was the only system employed in this league purely because the format of the second half of the 2016-17 season was to change from previous seasons.
Although hit rate and yield were both below estimates, the system still recorded its third highest total for six seasons at £1,238.00.
There were twice as many bets than expected (largely due to the fact that around 40% more games than usual were played in the first half of the season to accommodate the new second half season format).
Result: Achiever
6. ENG1 – England Premier League – Whole Season System
This system suffered the longest losing streak we have ever encountered, almost 14% worse than expected, for a very painful 36 straight losses. (A six-season-low).
However, the situation was mostly recovered by three big winners all carrying odds of over 12.00, for a final loss of just -£32.00.
The last six bets of the season lost, but had only one of these been a winner, this system would have returned a profit. Small margins.
Result: Zero-Sum
7. FRA1 – France Ligue 1 – Whole Season System
Hit rate and yield were both below par, but the league turned in a steady performance for a profit of £1,634.00, ranked fifth largest in the last six seasons.
Result: Achiever
8. FRA2 – France Ligue 2 – Whole Season System
The same story as France Ligue 1, but with a profit of £2,801.00, for its third largest profit figure in six seasons.
This one was unusual for a much higher number of bets than expected. (215 vs. 119 estimate – Profit was at £2,523.00 after 119 bets).
Result: Achiever
9. GER1A – Germany Bundesliga 1 – First Half Season System
Up to 21/12/2016 winter break
Another below par system, but one that still achieved a profit of £1,098.00. (Fourth largest in six seasons).
Result: Achiever
10. GER1B – Germany Bundesliga 1 – Second Half Season System
From 20/01/2017 start of the second half of the season
The very rare inclusion of a system with two non-profitable seasons (the oldest two) in the previous five.
It featured one more bet than expected, which won, to total one more winner than expected. Hit rate and yield both exceeded estimates for a profit of £1,294.00. (Fourth largest in six seasons).
Result: Achiever
11. GRE1 – Greece Super League – Whole Season System
The hit rate here was almost 8% below estimate and resulted in the worst performance for six seasons, and a loss of -£819.00.
Result: Loser
Hi again, RW. I am very excited to the upcoming season in applying all the knowledge learned here in the last few months.
A couple more questions:
1) I found very healthy inflection points in 2nd divison leagues of England, Italy, France and Germany. You held to France only. Is there a specific reason? Also, these other 3 have very good liquidity. Why not them? Aren’t they better and more liquid than Croatian, for example?
2) About odds movement: would it be correct to collect data from footballdata (friday 18h) and oddsportal (closing odds) to compares odds movement per league and per bet type (HDAFU) to try to antecipate betting movements? That would help, in my point of view, to place bets before the one hour ante post. Does that make any sense or it is impossible to antecipate those movements due to news about the teams at the last hour or so? Does this enormous effort compensate? Also, oddsportal collects opening odds x closing odds. Now, that would be enormous work because one cant copy paste these data. Please, some kind words about this topic.
Best regards!
Hi Gabriel,
Your question is worthy of a series of articles, so I will try and be brief!
We tend to look at only the top-tier leagues in each country because they are far more stable in their composition than any of the leagues below (and therefore give more reliable results).
For example, in France Ligue 1, 65% of teams have been ‘ever-present’ in the five seasons from 2012-17 (and also in the five seasons from 2011-16). Compare this with Ligue 2, and you will see that the latter has only a 35% rate of ever-present teams during those same periods.
The fact that the composition of lower leagues is so greatly affected by teams leaving/joining as a result of relegation/promotion creates a good deal of statistical noise. Results just aren’t as reliable because there are fewer games involving the same ‘established’ teams each season.
Ligue 2 is the only second tier we look at purely because of the prevalance of one particular bet type, which is almost unique to this league.
In reply to point 2, unfortunately, Football-data.co.uk collects its data at two separate times during a week. Fridays for weekend games and Mondays (I believe) for midweek games, which is why we need a more reliable benchmark for odds.
Oddsportal includes time stamps for when the odds were recorded and also red/green arrows indicating which way the odds have moved (from the opening odds of each bookmaker to the odds currently displayed).
If you study Oddsportal from the opening of the market in any particular round of games not yet played and watch/record the odds as they change during the ante post period, then you will have a good idea of how the market moves. I can say that with a lot of experience watching odds movements, the trends are fairly predictable, and help us judge when we can place bets at optimum prices.
This also helps us avoid congestion on days where we can see there may be a glut of bets all needing to be placed within a short space of time – we can place them further in advance and still get prices within the inflection points, sometimes a lot better than the closing odds.
Therefore, there is no need to collect data from more than one point during the ante post period for the inflection points analysis.
I hope this helps.
I won’t mention the particular league this question relates to so as not to give away any specific content, but was wondering if I could get an opinion on something please.
It’s a first half workbook, and of particular interest to me is the draw bet type, which gives a reasonably profitable medium risk (I think) system. The profit curve increases smoothly from odds of 3.22 up until 3.38 and then levels off briefly before dropping off and then increasing again smoothly.
So it seems there is sweet spot in the lower odds range, followed by a brief period of statistical noise and then another decent profit increase. So I looked at 3 things in detail here; the first rising portion of the curve, the second rising portion, and then the entire curve.
I found the following:
Odds 3.22 – 3.38: Profit 5072, zero odds 2.43, hit rate 41.13%, yield 35.97, 28 total bets, losing streak 8.
Odds 3.5 – 3.68: Profit 4659, zero odds 2.73, hit rate 36.6%, yield 30.45, 31 total bets, losing streak 9
Both of those sets of odds ranges were in profit 4 out of 5 seasons. The gap between the 2 systems is the statistical noise you mentioned in the article.
Then I looked at Odds 3.22 – 3.68 to see what effect the statistical noise had on the outcome, and found this….Profit 7536, zero odds 2.93, hit rate 34.17, yield 17.28, 87 total bets, losing streak 9. Interestingly it was in profit 5 out of 5 seasons.
My first option would be simply to go with the system with inflection points 3.22 -3.68, which to me looks good. I could also go with the entire odds range of 3.22 – 3.68 to access the higher profit at the expense of hit rate, yield, and having to place more bets.
Then I thought to myself, well why not just run both systems for the first half? They don’t conflict with each other, they are the same bet type in the same league, all I’m doing is cutting out an unprofitable section between 2 very steady portions of the curve. And by doing so I would increase the 5 season profit by 4659 and only need to place 30 extra bets, taking the total estimated bets for this first half system to 59.
Does this sound feasible to you given the figures I have found? Or would I be better served here by just going with the 3.22 – 3.38 system?
Thanks.
Hi Simon,
For us, the decision is not to run split systems in the same league.
As mentioned in Page 1 of the User Guide, we have found that the synergy effect gained of running only the best single system from a league is detrimentally affected if you begin running competing systems in the same league.
Therefore, we would suggest choosing just one of the three options you have outlined. Either go with the longer 3.22-3.68 cluster (more bets but higher potential profit), or choose one of the two segments within this cluster, i.e. the 3.22-3.38 or 3.50-3.68.
Of the two smaller segments, the first is better (higher profit, fewer bets).
Ultimately, the choice is yours and is probably easier to make once you have a rough idea what the risk profile of your portfolio is after you have highlighted the best systems in each league.
Remember, try to keep the portfolio as a whole as balanced as you can. Choose whichever system adds or maintains balance to the portfolio as a whole.
I hope this is clear!
Thanks for the reply, that makes sense. Focus on the best only.
At present the 3.22 – 3.68 odds cluster is the one I had chosen for that league for the 1st half, and would be one of the medium risk systems. Then for the 2nd half in that league there is a solid favourite system that I had chosen as a low risk.
I think I’m almost there really when it comes to the portfolio – but am just tinkering to see where I can make things better, if that’s possible.
There are 20 systems I have assembled in total; 4 low risk, 5 low-medium risk, 6 medium risk and 5 medium-high/High risks. In fairness I have modelled it in large part on the successful one you illustrated, to get a similar balance.
I’m not a high risk type of person though, preferring moderate risk with a tinge of something higher risk included. Some of the underdog systems I have analysed with big profits but come with between 20-30 game losing streaks are something I’d rather stear clear of! With what I have now, 6 of the systems have double figure losing streaks, but all under 20. That’s palatable for me especially as the lower risk systems balance that out with the longer winning streaks.
I know I won’t make the portfolio perfect (doubt that is possible anyway) since this is my first attempt, but overall I’m happy with the way it looks so far. All that remains is to see how it fares in a couple of weeks when the first games start.
Hello again Simon,
Don’t worry, there is no such thing as the ‘perfect portfolio’ but I’m afraid you will need to get used to the ‘high risk’ element as it is a vital constituent of the whole mix, especially if you are going to achieve profits worthy of the time you’ll need to run the operation.
Bookmakers rely to a certain extent on the psychology of the majority of punters to make their profits. People tend to back lower priced favourites (usually under-priced) because they think they are more likely to win, and also because they are scared of backing a rank outsider.
Running a mixed risk portfolio allows you to gamble on the higher prices (and reap the higher rewards), because the lower priced bets provide a safety net.
It sounds like you have a good understanding of the concepts we have outlined, so, good luck – hope it all goes well for you!
Hi Right Winger,
can you explain more in detail how do you compile Inflection Points graph? Odds are ranked from 1 to 200 in both summer and winter leagues campaigns. In Summer league campaign every number is repeated 4 times 7 times in a row, and every 8th number is repeated 5 times in a row (8, 16, 24 and so on). As I understand it’s because you get round cluster size number for odds rank no. 8 – 33.000, then for 16 – 66.000 and so on. Or, whenever you get closest to round number, for instance in winter league it’s odds rank 7 and cluster size of 63.980. Regarding my own portfolio, to make my own inflection points graph with data taken from HDAFU tables of the 2016-2017 winter leagues season, there are 1452 bets. 1452 divided by 200 equals 7.26. I tried to calculate the right sequence how to spread and repeat odds rank numbers evenly across those 200 and it didn’t work out. I finally finished with 202. But numbers of odds rank in some cases are repeated for 7 times and every 8th number for 8 times and then I have intervals where same odds rank numbers are repeated for 7 times 15 times in a row. A bit chaotic, but just to finish “ranking” around 200 because I struggle to calculate the right sequence. It was not the easiest thing to explain, but I hope you understand what I mean.
Hello Jo,
Yes, I know exactly what you mean and I think you have described it rather eloquently.
And I’m afraid my answer is going to disappoint you.
I’m sorry but this is one of our ‘trade secrets’, and we are not prepared to divulge or give away for free this part of our intellectual property.
The clustering sequence is just a small part of the HDAFU table compilation process. There are other bits of the jigsaw you haven’t mentioned, which are also essential to know and get right when putting the tables together.
And some of the formulas needed to calculate vital areas of the tables will not be found in the actual tables themselves.
Of course, from a personal perspective, I am 100% proficient in putting the tables together. I’m very fast as I know exactly what I’m doing, and even have the raw data set-out in an optimum design to copy and paste nicely into the template.
But even at a lightning pace, it still takes me at least 20 minutes to put a single table together (more than an hour for one league), and then more time to review that everything is okay and cross-checking nicely.
Today’s product is the culmination of several years of work. We allow people to buy the tables but I’m afraid we will never reveal the recipe for making them.
And I would strongly advise against anyone trying to copy them as, if you don’t have the various formulas for certain parts of the table, you will not even know if the information you receive from your end product is correct or not.
The probable knock-on effect would then be reliance on a totally false picture, which will only lead to disappointment (at the very least) when it comes to investing money on it.
This is therefore our health warning – put together your own tables at your own peril!
There are over 100 steps involved, many of them requiring to be performed in a certain order – cracking the exact code and sequence is probably akin to the old statistical theory of countless monkeys on countless typewriters. (The Infinite Monkey Theorem). Good luck to anyone who can do it!
Sorry again Jo!
No problem Right Winger, I understand.
Regarding my portfolio, after your previous comment I made a lot of changes and now instead of low risk strategies, medium risk strategies makes it unbalanced: 19% low risk, 25% low-medium risk, 51% medium risk and 3% medium-high risk strategies. 31 strategy in total. I checked everything thoroughly and making any better balance would mean to exclude some strategies or leagues from betting. What can you say about portfolio when higher risk strategies, in this case medium risk, are dominant over the others? Is it a good idea to exclude some of them to make the portfolio more evenly balanced? Just two of all strategies have zero odds higher than first inflection point, but no more than 0.04.
Hi Jo,
I think I would be a little uncomfortable with a portfolio biased towards medium risk strategies.
You have a fair enough representation of low/low-medium risk systems, but the higher end, in other words, medium-high and high risk strategies is hardly represented at all.
With the portfolio you have suggested I imagine that you may see a profit, but it won’t come close to the type of figures we achieved in the 2016-17 Winter League Campaign.
You need the higher odds bets to make the money. The smaller odds bets create the safety net to allow you to speculate on the higher risk bets. This is a key element of the portfolio approach.
The old adage ‘speculate to accumulate’ applies to any form of investment – life is one big gamble in general. If you don’t include room to speculate in your portfolio, then I imagine you will perform a lot of work placing many bets, with perhaps a less than satisfactory outcome at the end.
You could be looking at a yield far less than 10%. Too many small odds bets will make it harder to recover losses. See also my answer to Simon from yesterday at 1.51pm.
In summary, I think your suggestion has the potential to make money, but you’ll have to have an exceptional season to see a profit size that will make all the work worthwhile. If your season is as bad as the reality check on page 5 of this article, you may even lose money.
But if you’re looking for an introduction to gain a little experience running a multi-risk portfolio, then go with it. If it is going to lose, I think it will lose slowly, giving you the option to cut and run if you get fed up with it.
I hope this is useful.
Hello Right Winger,
i am in the final stages of the analysis and i wanted to ask you for an opinion about the spread of risk of my portfolio. I have just selected the best systems (as far as i can tell) from each Championship, 39 systems in total. I have selected 4 low risk, 8 low-medium, 11 medium, 3 medium-high and 13 high risk systems. The ratio in % is: low 10%, low-medium 21%, medium 28%, medium-high 8% and high 33%. Looking at the spread on your campaigns i notice that i have slightly more systems at the upper and lower end and less in the medium range. So i want an opinion from you.. what do you think about my spread? Do you find it balanced or should i fine tune it to balance it more?
Thanks, i’m really looking forward to the beginning of this season (in a week or so for the early championships)!
Hello Daniele,
I would be happier with a more even spread across the board.
With fewer bets at the lower odds range, your portfolio will find it difficult to make-up for the losing streaks at the higher odds range.
Whilst Jo’s portfolio (see 6th July above), has a chance of making money (as it is biased towards lower risk), I think your current arrangement will struggle more.
Increase the low risk and medium-high risk bets and reduce the high risk bets for a better balance.
Good luck!
Thanks for your answer.
I noticed though that in this article you stated that you chose a spread of 4 low, 6 low-medium, 6 medium and 6 high-medium/high, for a total of 22 systems. Looking at the spreadsheet of the campaign I see that the inflection points guidance as per Page 1 of this article does not exactly match what is stated above, and if we analyze the maximum inflection points of each system we count 3 low (max inflection point of 2.22), 2 low-medium (max inflection point of 2.85), 8 medium (max inflection point of 4.44), and 9 medium-high/high systems. Could you please clarify on this matter?
For example, a system in your campaign have odds from 2.13 to 7.40. Don’t you classify it as high risk system? Or another system has odds from 4.25 to 5.25. Don’t you classify this as a medium-high system?
I understand that the first system spread across all 5 categories, and the second have a spread between medium and medium-high. So how do we decide in which category they should fit in?
Thanks!
Hi Daniel,
As mentioned in the article, the odds splits for the risk analysis are a ‘rough guide’ to get readers thinking about the composition of their portfolios, and to encourage people to balance their portfolios with an even spread of risk.
The classifications stated for our 2016-17 portfolio are based to some degree on discretion (also mentioned as an adjustment factor above), and you will see from the 2016-17 Campaign workbook that some of our systems did indeed cross the thresholds of several odds’ bands.
When this happens, we can never be truly exact but we can say from the start that the majority of the bets will fall into one or another of the odds brackets.
Using the example you mentioned, in a system covering a range as wide as Greece between odds of 2.13-7.40, we know from the outset that there will be more bets at the lower end because it is a home win system. Indeed, the analysis showed the harmonic mean odds for the matches between these inflection points as 2.97, which further validated our decision to classify this system with low-medium status.
The harmonic mean for the 110 bets placed was just over 3.00.
As a reminder, the harmonic mean formula in Excel is =harmean(a1:a10), if the range of odds you are looking at encompasses cells A1 to A10.
Perhaps our readers will come up with better ways of gauging the risk content of their portfolios, but the overarching message we are putting across in this article is that a healthy balance should always be sought.
It’s not an exact science, maths never is, but hopefully we have guided you in the right direction.
Another great question – thanks again!
RW,
I think you did a great job with explaining how you gauge the risk of the system within the context of a portfolio of systems.
When getting the balance in mine as even as I could I relied to an extent on the workbook of your previous campaign to see how you yourself classified a system, taking into account things like the odds range, hit rates, yield, number of bets e.t.c. So it was interesting to see based on these things how you rated your own systems. I then looked at my own systems to see how they looked in terms of risk.
I found the low risk and low-medium risk the easy ones to classify. I look for those systems to have a good length of winning streaks, low length of losing streaks, 50-60%+ hit rates, and of course meeting the 4 of 5 seasons in profit and 10%+ yield. Most of these systems have relatively small odds ranges.
The ones I found a little more difficult to classify were those that had odds range more spread out, like the Greek example you gave. So applying the harmonic mean to it, gives that extra view point that I found helpful to see how risky a system is based on odds the majority of bets placed will be at.
What I had done was analyse each league to identify the systems and then make a spreadsheet that listed each system for whole, 1st half, 2nd half, and show each of the systems features i.e hit rate, zero odds, yield e.t.c
From that I selected those that I felt gave a balance and listed them in order from low risk to high. Applying the harmonic mean odds to each system then showed that the odds increased from low to high, which I think validated that I have my systems in the right sections of the risk profile and have a decent balance.
That’s something I found useful in getting something I was finally happy with.
Now, very much looking forward to the weekend to take things “live”.
Good luck everyone and I wish you all the best for the upcoming season!
Hi RW – sorry, but I’ve got a question you’ve probably been asked a million times about placement of bets.
I’m currently sorting out getting some money onto Vodds in preparation for the winter season starting, and obviously anticipate being able to access the upper end of the odds range as a result.
My questions are:
1. In terms of bet selection, should I only be placing the bet if the max odds on oddsportal are within my strategy range? I’ve got a summer league campaign running at the minute and I tend to use the upper range as a guide but I always check if the top price looks anomalous – if it does, and if I can access odds within my range, I’ll place the bet. However, I’m never sure if I’m “breaking protocol” here and should be using the top price as the decision point, anomalous or not….I’m conscious there might not be a right answer to this but could you please expand a bit on how you weight your decision on selection.
2. TIming is the other question. I’ve tended to look at the odds an hour or so before K-O and made my decision at that point. However, on occasion I’ll have a cheeky look just before K-O and can sometimes see that the odds have shifted markedly and I’m now in a position where I should be placing a bet. Should I ignore this and not place the bet, or is this now deemed an appropriate selection based on the late odds movement?
Thanks again!
Hello Scott,
To answer your questions seriatim:
1) Yes, place the bet in accordance with the highest odds shown in Oddsportal, and yes, it is always a good thing to check for and preclude anomalous odds that haven’t updated for a while.
As previously mentioned, we try and iron out the anomalous odds when putting together the analyses, and for the sake of good order, you should try and do the same when judging whether to place the bet if it is close to the outliers of your odds range.
2) The HDAFU analyses are based on odds as close to the end of the ante post market as we can get them. If towards kick-off you realise that the odds have shifted a bet to within your inflection points then, yes, it is a valid bet to place.
I hope this clarifies.
Hi – thanks for your reply, and apologies for my delayed response to it. All makes perfect sense, thanks again for taking the time and for your excellent website.
Hi Soccerwidow
Please could you explain why you went with a system in Denmark despite the 2016/2017 season being the first season after major structural changes?
I would have thought that that was risky in the sense that historical data is now not directly comparable to the new format.
Asking because there’s other leagues with the same problem for the upcoming season, eg Bulgaria / Ukraine.
Been having some good success with a summer leagues system thanks to your work, with roughly £1,200 profit since March.
Thanks
Hi Michael,
I think I have answered this one before to some degree. (Page 2 of this article I believe?).
We were aware that the Danish league format was changing in the second half of the season, hence why we chose to include just a first half season system.
The new first half up to the winter break included straight-forward league matches, just as any other season before. The only difference likely was that there would be more matches encapsulated by our inflection points.
Therefore, for all intents and purposes, the league was no different up to the break than before – just league games during the same part of the year in accordance with the analysis.
Thanks for sharing your Summer League success story – keep in touch and let us know where you end up!
All the best.
Hi Right Winger,
since Danish league had only 1 season in new format, would it be better to skip the 2nd/whole season strategies from betting in the upcoming season?
Hi Jo,
Personally, I don’t think it matters a great deal.
The only important thing with the new format in Denmark is not to include any Relegation or Europa League play-off games in your strategy.
Also, expect fewer numbers of games entering your inflection points in the second half of the season as there are fewer ‘league’ games after the winter break now.
Check out our Summer & Winter League Calendar for a full summary of what to include and what not to bet on in each league.