2018 Summer League HDAFU Tables: Standard Deviation Adjustments

In the quest for further accuracy using the HDAFU Tables to compile successful portfolios, the forthcoming 2018 Summer League campaign now features Inflection Point adjustments using Standard Deviation technique.

What is Standard Deviation and how does it help?

Standard deviation is a measure used to quantify the amount of variation within a set of data values. It measures how far a set of numbers are spread out around the mean, or average value. In this case, five seasons’ odds data.

Calculating the standard deviation values takes into consideration larger deviations from the mean. In other words, the ‘outlier’ values; those that are furthest away from the mean.

Thus, standard deviation provides us with a ‘margin for error’ allowance and also a more rounded perspective of the statistical conclusions to be formed. In short, standard deviation provides us with a wider safety net for predicting future results.

2018 Summer League Standard Deviation Adjustment Figures

Once you have calculated the two inflection points (see User Guide), use the standard deviation tables below to adjust these figures to provide a wider, more accurate scope for your predictions:

Summer League Inflection Points Standard Deviation Adjustment Figures

Random Example Calculations

Example 1: Brazil Home Wins – 2nd Half Season Analysis.

Let’s say your chosen system is home wins in Brazil for the second half of the Série A season. Your two inflection points are 1.98 to 2.56.

1. Divide both figures by 1 to find the implied probabilities of the odds:
1/1.98 = 0.50505, or 50.505%
1/2.56 = 0.39062, or 39.062%
2. Lose the percentage signs for the time being from the implied probability figures:
Lower odds/higher probability threshold becomes: 50.505
Higher odds/lower probability threshold becomes: 39.062
3. Look up the adjustment value in the table above. The middle table shows the second halves of each Summer League season. The home win value for Brazil is 0.677
4. ADD the home win value of 0.677 to the lower odds’ threshold: 50.505 + 0.677 = 51.182
5. SUBTRACT the home win value of 0.677 from the higher odds’ threshold: 39.062 – 0.677 = 38.385
6. Add the percentage sign back on and convert both figures back into odds values. Round-up or down during this step only:
1/51.182% = 1.953, which rounds-down to 1.95
1/38.385% = 2.605, which rounds-up to 2.61

The new inflection point range adjusted for standard deviation is therefore 1.95 to 2.61.

Example 2: Sweden Underdogs – Whole Season Analysis.

Inflection points: 4.16 and 7.20

1. 1/4.16 = 0.24038, or 24.038%
1/7.20 = 0.13888, or 13.888%
2. 24.038 and 13.888
3. The bottom table shows the Whole Season adjustment values for each league. In this case, Sweden’s standard deviation adjustment is 0.420
4. 24.038 + 0.420 = 24.458
5. 13.888 – 0.420 = 13.468
6. 1/24.458% = 4.088, which rounds-up to 4.09
1/13.468% = 7.425, which rounds-up to 7.43

The new inflection point range adjusted for standard deviation is therefore 4.09 to 7.43.

Conclusion

The 2018 Summer League seasons will be the only time you will have to carry out these calculations manually.

In future, the standard deviation calculations will be incorporated into the HDAFU tables themselves and will calculate automatically.

Last Update: 23 January 2018

14 Responses to “2018 Summer League HDAFU Tables: Standard Deviation Adjustments”

1. AM0751
29 January 2018 at 4:26 pm #

Hi Right Winger,

Some questions about the summer league campaign, and a generic question to finish.

1. I know you mentioned before you approach your summer and winter leagues as one big portfolio. Does this mean you will spread the high risk, medium risk and low risk systems for the summer leagues taking into account your winter leagues as well, or you separate this between summer and winter leagues?

2. A question that has been nagging me for quite some time. As above, you advocated for spreading risk by using high risk, med risk and low risk systems in a mix. You also said that anything else than a nice balanced portfolio would struggle to make money.

Hypothetically, given a scenario where one has deep pockets and can sustain a long losing streak without too much difficulty; wouldn’t it be more lucrative to just take the most profitable systems of each league and go ahead with those, regardless of high, medium and low risk? Or is there another mathematical reason, beyond being able to sustain long losing streaks, for this mixed portfolio approach? My simple thought would be that as you have deeper pockets to be able to sustain long losing streaks, the most lucrative system should be chosen in each league…

Very interested in what you have to say about that.

• Simon
29 January 2018 at 8:50 pm #

AM,

The points you have made are ones that also occurred to me too. I’d really like to hear RW’s thoughts.

I had always in my mind thought that the two portfolios; summer and winter leagues, would be distinct, but similar in make up with regards to risk. For example, you would be able to find plenty of “good” and “sufficiently profitable” systems in all risk categories i.e, low, medium, high. Thus if the summer and winter leagues were mirroring each other in risk balance then all would be well, and they could run in overlap without having any negative effect on the overall picture. In essence more bets overall, with the same balance of risk.

Having now had the chance to analyze the summer leagues I find myself seeing that the majority of sufficiently profitable systems are in different bet types than the winter leagues.

Of course, my point above is made with recognition that the summer leagues in general are smaller in size than the majority of winter leagues.

For example in the winter leagues, Germany, Poland, Spain, England, all have very large profits accumulating in the underdog bet types – ranging from odds 3 to 9+ France and Italy showed large profit systems in the Favorite bet type. Netherlands has a very profitable draw system.

The summer leagues that I have analysed, 9 in total, have shown me that Home wins are showing very good profits in the last 5 years, favorites seem to do well, but the profits are small. A few leagues show that away wins are profitable.

Aside from Brazil, not many leagues have large profitable underdog systems, at least not to the extent where it is the best system on offer in the league. By “best” I mean that it meets most if not all of the criteria set out in the user guide.

When I look at the overall set of systems I analyzed, the ” best” ones tend to be home wins away wins and favorites. I didn’t find too many good draw systems (as the odds ranges can be extremely narrow in some cases), although there is one I will definitely be using. Underdogs, as mentioned, only really Brazil, has large profits here and would be worthy of including in a portfolio as a high risk system.

So when I look to compile the portfolio, based on what is going to make the most money (in theory), I look to the systems that historically have profited more in the last 5 year’s, season on season. These tend to be lower risk systems from the home win, away win and draw bet types. Contrast this to the winter leagues where there were plenty of high risk systems with lucrative rewards, making them worthy of inclusion.

The portfolio I compiled for the summer leagues 2018 has an average hit rate of 48%, which is 4% higher than that which I settled for the winter league campaign. I organized the campaigns too by harmonic means and the systems (spanning whole, 1st and 2nd half) range as follows:

2.027
2.105
2.16
2.26
2.335
2.406
2.492
2.583
3.041
3.043
3.058
3.15
3.173
3.44
3.57
4.814
6.712

It occurred to me that I don’t have many systems that would be high risk in the sense that my winter campaign had many more systems of an underdog nature with odds ranges showing higher harmonic means.

Perhaps I am missing the point, but for the sake of introducing more high risk systems into the portfolio, I would likely have to omit lower risk but more profitable systems. After all, if there were more profitable underdog systems I would have included them already.

Perhaps this is just the way it is this year, or perhaps the smaller leagues show different patterns with regards to win ratios and odds setting.

Personally I don’t feel 48% as a hit rate is too high or too low risk. I do feel the portfolio I have compiled (first draft) is balanced. But, yes, I would definitely like to know peoples thoughts in sacrificing potential profits for the sake of “balance” or “more balance”.

• jo
30 January 2018 at 2:02 am #

Hi Simon,

it was hard to find a lot of high odds strategies to balance my 2017 summer portfolio.
When I was building my winter leagues portfolio based on low to high risk strategies, I assessed the risk of my summer leagues too. I would say the portfolio balance was not the best:

low risk 3
low-medium 3
medium 6
medium-high 3
high 1

The biggest loss I was in was around -1050, I placed 560 bets and the end result was +7600, 100 units flat stakes. Some of strategies had differences in hit rates more than double over 5 years and looked unreliable to me, but no problem for the end result. The only one high risk strategy had max 5 bets a season over 5 years, so it’s hardly a strategy at all, but it was little to lose and a lot to gain, so, I included it and it failed. I had only 1 draw strategy, it ended +190. In the end, the portfolio still made decent money in my opinion. I agree with the idea that high risk strategies makes the best/real money, out of 4 medium-high and high risk strategies 2 failed, but the remaining 2 made up 50% of final profit. Although it’s just 1 season, it doesn’t seem to me that not the best balance of portfolio is a waste of time. I think you should have high odds/high risk systems included, but maybe it’s not worth to pursue them at any cost to have “right” balance, especially if you don’t like something regarding their performance over 5 years. But, I did not like Brazillian underdog hit rate differences from 16% to 52% over 5 years, more than triple, no problem again, Brazillian underdog was the best performer and produced profit at least 6 seasons in a row.

• Right Winger
15 February 2018 at 10:55 pm #

Hi AM0751,

1. Look at the Summer and Winter Leagues separately and ensure they are both balanced. If you balance them both separately the overall result will be balanced too.

2. Yes, I agree with you totally. If I had a huge bank balance, didn’t mind losing it and had no emotions towards the situation at all, then I would certainly go for the higher risk systems highlighted by the tables, because that is definitely where the money is.

Bookmakers make the majority of their money on shorter priced favourites that don’t win often enough to justify their prices. Underdog prices have to be increased to balance the books, so underdogs do carry huge value a lot of the time.

But unfortunately, I don’t have the resources, and I am only human. I also understand losing streaks and to endure a streak of 36 consecutive losses in the EPL underdog system of the 2016-17 Winter League Campaign was pretty heavy going I have to tell you.

Just imagine a whole raft of such systems experiencing streaks such as that at the same time! No thanks, I’ll stick to the investment banking approach and spread my risk.

Nice thought though!

2. acepoint
28 January 2018 at 9:40 am #

»Effectively, we are comparing each season’s average odds with the overall average for five seasons and then interpreting the filtered result based on the latter.«

I don’t get it. I took the MLS data (from your provided sheets) but calculate 1.648 as standard deviation for the average full season home odds. In your table I see 1.845.

And I don’t understand, why the standard deviation is added to the low end (and added to the high end), which widens the interval. I would have expected that shortening the interval by the SD will give more precise results.

• acepoint
28 January 2018 at 9:43 am #

of course I meant »subtracted from the high end«. Sorry for this error.

• Right Winger
28 January 2018 at 12:17 pm #

Hi Acepoint,

I am sorry but I have no idea how you have carried out your SD calculation and therefore cannot comment on your result. But I can assure you that the 1.845 figure is correct in accordance with our formulas.

If you have bought the Soccerwidow Odds Calculation Course you will realise that the SD formula we use is specifically designed for small data sets such as five seasons’ results in a league. It is slightly different from the standard recognised method of calculating SD that you will find in most textbooks, but absolutely valid and correct just the same.

I shall not go into the method of calculation we use as that would entail repeating what is in the course, which is rightly the property of those who have paid for that information.

But to answer your second point with a question, why would we restrict the filtered inflection points and ‘shorten’ the interval, when we are attempting to bring outlier results into play (in other words, those slightly outside of the filtered range)?

To shorten the range would be to cut down on the opportunities, thereby reducing the chances of winning on bets that are rightly to be considered part of the strategy, whilst the whole point of SD adjustments is to increase the chances of winning by encapsulating all the relevant data up to the edges of the historical results’ boundaries; not restricting our tool to just the partial, filtered result.

When you filter, you will undoubtedly be left with different numbers of historical results in each season. In other words, 20 bets may qualify for your ‘system’ from 2017, maybe 17 from 2016, 23 from 2015, and so on. The SD adjustment is aimed at the different odds (prices) of these qualifying games, not the match results. With SD we are attempting to balance the odds mis-matches with the adjusted differences between each season and the overall harmonic mean of all five seasons put together. The adjustment in most cases will be small, but is significant in terms of fine-tuning for accuracy.

Let’s take an example. Two identical (or near identical) matches in two different seasons (one in your historical analysis and one coming up for analysis). The historical one is priced at 2.20 and the upcoming game is at 2.25. The only difference between these games is that the weight of money in the market at the time is slightly different, generating the different prices. Your inflection point cut-off is 2.23. Using just a filtered analysis you would ignore the game priced at 2.25, and potentially miss out on a winner. Adjusting your odds range using SD in accordance with what has happened in the previous five seasons might see your inflection point change slightly to 2.27. The 2.25 would then be within your range, and quite rightly so, as it is within the narrow band of differences between each historical season. Once we’ve adjusted our inflection points in accordance with the price difference averages of each season, these ‘outlier’ matches become valid selections.

Effectively, with SD adjustment, we are standardising five different seasons’ odds and presenting them as one block of information. You could say we are ironing out the differences between those five seasons and producing a standardised block of data as a benchmark for future predictions. As I mentioned in previous comments it is not an easy thing to explain in layman’s terms, but I hope at least the tenet is understandable.

You don’t need to apply SD if you don’t want to, but we’ve gone the extra mile to provide this information for those who do.

Hope this helps and thanks again for your comment.

• acepoint
28 January 2018 at 1:07 pm #

Hi Right Winger,

»I am sorry but I have no idea how you have carried out your SD calculation…If you have bought the Soccerwidow Odds Calculation Course«

Among other HDAFU sheets I bought the recent 2013-2017 MLS and the Over/Under book (German: Basiswissen Sportwetten) by Soccerwidow (is this the same you mentioned above?). And I took the numbers from the MLS sheet you provided. Of course I see and accept your point of not telling all secrets here in public. IMHO your page provides more than enough knowledge for free and I’m grateful for this.

Nonetheless I stumbled over the difference in SD in this certain case and was curious because I not only want to consume but understand when I put money into bets.

Meanwhile I figured out that in Excel, where you can use two different methods for the SD, one is called »calculated« (for basic population) the other »estimated« (for random sample), I took the calculated before which shows 1.648. The »estimated SD« formula gives 1.842 which is at least very close to your number. Basic population were the five numbers for Home 2013-2017. Nonetheless I’ll check the book again.

I see your arguments regarding the second point and though I can’t say I agree totally I’ll at least take a deep thought on it ;-). Thanks again!

• Right Winger
28 January 2018 at 2:24 pm #

Hi Acepoint,

Yes, I think you have the SD formula now. The auto-Excel function does give slightly different figures to all of our figures above. Our 1.845 was a pure, manual calculation to three decimal points on all of the seasonal numbers used for the sake of pinpoint accuracy.

And thanks for the compliment! Yes, perhaps we do give away too much for free. In hindsight that’ll be £5 please! 🙂

• jo
28 January 2018 at 5:13 pm #

Hi Right Winger,

once you have inflection points range, let’s say it’s 2.00-2.50, and after adjusting SD it’s 1.97-2.53. Is it then a good idea to take a fresh copy of HDAFU and filter results at 1.97-2.53 range? To see how the strategy worked with wider safety net in those last 5 years?

• Right Winger
28 January 2018 at 5:36 pm #

Hi Jo,

No, no need. The new range is a cosmetic adjustment of the original inflection points for use going forwards, not looking backwards.

Look backwards out of curiosity only but don’t pay too much attention as all you will see is a mechanism to predict the future more accurately, not the mechanism in action.

Hope this is clear.

3. Scott
27 January 2018 at 7:02 am #

Hi Robert – yes, all very clear thank you. Thanks for taking the time to write such a detailed response, it’s definitely helped with my understanding.

4. Scott
26 January 2018 at 12:27 pm #

Hi – thanks for this. I’ve now been through all of my summer selections and adjusted them for the SDs as indicated above.

I’ve checked and re-checked the calcs, but in one of my strategies, the inflection points move quite markedly – from 6.01 to 7.6 all the way out to 5.75 to 8.07. I guess this is because the higher risk strategy results in relatively lower probabilities, so adding the SD to these has a relatively greater effect….?

Couple of questions if you don’t mind, please.

Firstly, is a shift like that shown above reasonable? I’m sure I’ve done the calcs right, it just seems like quite a significant widening of the odds range.

Secondly, is the wider range now my target range when it comes to finding bets on oddsportal – i.e. I should move to the new post-SD adjusted range? (I’m sure it is and this is a daft question, but just wanted to check!)

Thanks again for all the help and the insight.

• Right Winger
26 January 2018 at 8:31 pm #

Hello Scott,

Don’t worry – no question is daft – I’m just the same as you, always seeking reassurance before committing.

6.01 is an implied probability of 16.64% (i.e. 100/6.01) whilst 5.75 is 17.39%, so the difference is just 0.75% (17.39-16.64). Between 7.60 (13.16%) and 8.07 (12.39%) is virtually the same difference. To answer your first question – absolutely reasonable. To answer your second question, yes, the new figures are your new spread for selecting matches via Oddsportal.

The odds differences look much larger than they actually are but if you put it into perspective, £10 at 7.60 wins £6.60, whilst £10 at 8.07 wins £7.07, so the difference is relatively small change. (Less than 10%). At lower odds ranges, the difference will of course be much smaller.

Ultimately, the Standard Deviation (SD) adjustments should widen your net for future predictions and provide more accuracy in the selection process.

Effectively, we are comparing each season’s average odds with the overall average for five seasons and then interpreting the filtered result based on the latter. This gives you more scope to pick up the outlier selections, which are also value bets and a valid part of the statistical tool you have created when selecting the system in question.

You’ll see in the new section of the 2018 Summer League tables just how close each season’s harmonic mean odds are. It’s quite an achievement for the market to register such small differences in a season of 380 games (e.g. Brazil), five seasons in a row. SD standardises the differences and provides a benchmark based on all five seasons’ odds spreads.

It’s a complicated thing to try and explain but SD is accepted and sound statistical practice, and if we can create as much accuracy as possible, the end game should also bear more fruit.

We haven’t had time to build-in the adjustments automatically this time round, but this feature will be in place in future versions of the tables (i.e. you will no longer have to calculate your final inflection points manually as per this article – we’ll arrange it so you just dump the two figures your filtering produces into a couple of cells, which will calculate the SD adjusted figures for you).

I hope this is clear and thanks again for your question.