16
It seems that the L1/LH effect noticed by John is not confined to the Moon; both lights show the effect.

If the Sun is L1 and LH, we find that again there are very few draws. Instead of the 26% draws we have on average, in the 40 games where the Sun was L1 and LH, only 3 (7.5%) were draws.

The effect is strongest with the Sun in Fire - just one draw out of 23 cases (4.3%).

17
GB wrote:It seems that the L1/LH effect noticed by John is not confined to the Moon; both lights show the effect.

If the Sun is L1 and LH, we find that again there are very few draws. Instead of the 26% draws we have on average, in the 40 games where the Sun was L1 and LH, only 3 (7.5%) were draws.

The effect is strongest with the Sun in Fire - just one draw out of 23 cases (4.3%).
Hi Graham,

I have now a dataset with 864 games (2010-2014 Portuguese Premier League), and I cannot confirm your results for Moon as L1/LH. In my case, for moon as L1/LH, I get this, which means draws are 25% for that case: {HouseTeam=10, VisitorTeam=5, Draw=5}

However, for sun as LH, I also get a lower number of draws:
AscRuler = moon: HouseTeam {HouseTeam=8, VisitorTeam=3, Draw=1}, aprox: 8.3%
AscRuler = sun: VisitorTeam {HouseTeam=5, VisitorTeam=6, Draw=1}, aprox: 8.3%

But I get low number of draws for many other combinations, such as (LH/L1):
- Jupiter/Mercury and Jupiter/Venus
- Mars/Mercury and Mars/Sun (This one is quite interesting, since all are HouseTeam winnings)
- Moon/Jupiter
- Saturn/Jupiter and Saturn/Saturn
- Sun/Moon and Sun/Sun

I include the Decision Tree generated by RapidMiner (it's a data-mining free software) below, so that you can maybe cross-reference with your data, if you wish.
GB wrote: If we can find a strong indicator for 2% of matches, and another for a different 4% of matches, and another for some more matches, etc. we can build-up a set of different indicators which might allow us to predict a significant number of matches.
About this quotation of yours in a previous post, I would just mention that care must be taken with that approach. This can lead to what in the Machine Learning and statistics area is called over-fitting (http://en.wikipedia.org/wiki/Overfitting). It is when you have too much rules to describe a model to a problem, that it start to only work good in your original test-set..


Jo?o Ventura

Code: Select all

Decision Tree by RapidMiner

HourRuler = jupiter
|   AscRuler = jupiter: VisitorTeam {HouseTeam=2, VisitorTeam=4, Draw=3}
|   AscRuler = mars: HouseTeam {HouseTeam=5, VisitorTeam=3, Draw=5}
|   AscRuler = mercury: HouseTeam {HouseTeam=7, VisitorTeam=5, Draw=1}
|   AscRuler = moon: VisitorTeam {HouseTeam=4, VisitorTeam=5, Draw=2}
|   AscRuler = saturn: HouseTeam {HouseTeam=5, VisitorTeam=4, Draw=3}
|   AscRuler = sun: HouseTeam {HouseTeam=7, VisitorTeam=2, Draw=5}
|   AscRuler = venus: VisitorTeam {HouseTeam=7, VisitorTeam=7, Draw=3}
HourRuler = mars
|   AscRuler = jupiter: HouseTeam {HouseTeam=2, VisitorTeam=0, Draw=0}
|   AscRuler = mars: HouseTeam {HouseTeam=3, VisitorTeam=2, Draw=2}
|   AscRuler = mercury: HouseTeam {HouseTeam=9, VisitorTeam=8, Draw=2}
|   AscRuler = moon: Draw {HouseTeam=1, VisitorTeam=1, Draw=5}
|   AscRuler = saturn: Draw {HouseTeam=3, VisitorTeam=1, Draw=3}
|   AscRuler = sun: HouseTeam {HouseTeam=6, VisitorTeam=0, Draw=0}
|   AscRuler = venus: HouseTeam {HouseTeam=16, VisitorTeam=7, Draw=5}
HourRuler = mercury
|   AscRuler = jupiter: HouseTeam {HouseTeam=4, VisitorTeam=1, Draw=0}
|   AscRuler = mars: HouseTeam {HouseTeam=4, VisitorTeam=4, Draw=2}
|   AscRuler = mercury: VisitorTeam {HouseTeam=7, VisitorTeam=9, Draw=4}
|   AscRuler = moon: HouseTeam {HouseTeam=21, VisitorTeam=15, Draw=11}
|   AscRuler = saturn: Draw {HouseTeam=8, VisitorTeam=5, Draw=8}
|   AscRuler = sun: Draw {HouseTeam=15, VisitorTeam=14, Draw=17}
|   AscRuler = venus: HouseTeam {HouseTeam=12, VisitorTeam=9, Draw=4}
HourRuler = moon
|   AscRuler = jupiter: HouseTeam {HouseTeam=3, VisitorTeam=2, Draw=0}
|   AscRuler = mars: HouseTeam {HouseTeam=14, VisitorTeam=8, Draw=9}
|   AscRuler = mercury: VisitorTeam {HouseTeam=11, VisitorTeam=16, Draw=12}
|   AscRuler = moon: HouseTeam {HouseTeam=10, VisitorTeam=5, Draw=5}
|   AscRuler = saturn: Draw {HouseTeam=5, VisitorTeam=3, Draw=5}
|   AscRuler = sun: VisitorTeam {HouseTeam=3, VisitorTeam=6, Draw=2}
|   AscRuler = venus: HouseTeam {HouseTeam=16, VisitorTeam=8, Draw=12}
HourRuler = saturn
|   AscRuler = jupiter: HouseTeam {HouseTeam=3, VisitorTeam=3, Draw=0}
|   AscRuler = mars: HouseTeam {HouseTeam=9, VisitorTeam=5, Draw=2}
|   AscRuler = mercury: HouseTeam {HouseTeam=9, VisitorTeam=8, Draw=3}
|   AscRuler = moon: HouseTeam {HouseTeam=6, VisitorTeam=1, Draw=3}
|   AscRuler = saturn: VisitorTeam {HouseTeam=3, VisitorTeam=3, Draw=0}
|   AscRuler = sun: HouseTeam {HouseTeam=10, VisitorTeam=1, Draw=5}
|   AscRuler = venus: VisitorTeam {HouseTeam=2, VisitorTeam=2, Draw=1}
HourRuler = sun
|   AscRuler = jupiter: HouseTeam {HouseTeam=4, VisitorTeam=1, Draw=4}
|   AscRuler = mars: VisitorTeam {HouseTeam=0, VisitorTeam=2, Draw=2}
|   AscRuler = mercury: HouseTeam {HouseTeam=13, VisitorTeam=12, Draw=7}
|   AscRuler = moon: HouseTeam {HouseTeam=8, VisitorTeam=3, Draw=1}
|   AscRuler = saturn: Draw {HouseTeam=5, VisitorTeam=4, Draw=5}
|   AscRuler = sun: VisitorTeam {HouseTeam=5, VisitorTeam=6, Draw=1}
|   AscRuler = venus: HouseTeam {HouseTeam=11, VisitorTeam=10, Draw=4}
HourRuler = venus
|   AscRuler = jupiter: Draw {HouseTeam=6, VisitorTeam=9, Draw=11}
|   AscRuler = mars: HouseTeam {HouseTeam=16, VisitorTeam=9, Draw=8}
|   AscRuler = mercury: HouseTeam {HouseTeam=22, VisitorTeam=10, Draw=8}
|   AscRuler = moon: VisitorTeam {HouseTeam=4, VisitorTeam=5, Draw=4}
|   AscRuler = saturn: HouseTeam {HouseTeam=19, VisitorTeam=10, Draw=10}
|   AscRuler = sun: HouseTeam {HouseTeam=10, VisitorTeam=6, Draw=3}
|   AscRuler = venus: HouseTeam {HouseTeam=7, VisitorTeam=3, Draw=3}

18
The issue of over-fitting is important. What I did not explain is that Ialways look at the statistical significance of the results. In the case of L1=LH=MO, the low number of draws compared to the average from the whole dataset was significant at 0.9% (p-value = 0.9%). In the case of the Sun the p-value for the draws was 0.3%.

Data mining can be very dangerous. If you use data mining software, you need it to give an indication of statistical significance for the results, and for that we need the unconditional probabilities of the various results (Home, Away, Draw) for the whole population, along with population and sample observation numbers.

I am happy that the results I quoted are significant and not the result of over-testing. As you will know, if you try enough possibilities, you will find some significant ones simply by chance. That is the definition of the p-value :)

19
GB wrote:In the case of L1=LH=MO, the low number of draws compared to the average from the whole dataset was significant at 0.9% (p-value = 0.9%). In the case of the Sun the p-value for the draws was 0.3%.
Hi again Graham,

I agree about the issue of statistical significance, which is usually set at a 0.05 threshold. However, in my dataset, the case for L1=LH=MO, still accounts for 25%. Although my dataset is 1/4 of the size of yours, I would expect that at this number of matches (864), I would also start to observe that kind of effects.

I have been using statistics for my work, but at a different level (I've been building statistical metrics for finding patterns in texts), so I'm not a specialist at this. But couldn't this be a case of a "false" statistical significance, i.e., something specific to your dataset?

(Maybe we should open a new thread for discussing our statistical approaches to football games?)


Jo?o Ventura

20
My results might well be specific to the English Premier League. With p-values of 0.005 and 0.003 they are certainly very significant statistically.

However there is an important general point worth thinking about. Matches at the same time do not all give the same result, even allowing for differences in odds. I have noticed that different leagues seem to show different patterns. So, the effect of a particular astrological configuration on a match in the Premier League is often different to it's effect on a match in the Championship, and this is different again to an FA Cup match.

This is why I concentrate on just one league rather than look at matches in different leagues and in different countries.

21
GB wrote:Matches at the same time do not all give the same result, even allowing for differences in odds. I have noticed that different leagues seem to show different patterns.
This happens within the same league. For instance, I have some cases of two games from the same (Portuguese) league that occur at the same time (locations not to far), and on one it is the house team that wins, the other is the visitors. Some times it is a victory followed by a draw..

This makes me think that there is something not really 100% true about the Ascendent being always the house team, and the 7th the opponent, at least as in the analysis I've making of this..

Also, theoretically there is something about "free-will" in astrology. For instance, even if we get the correct astrological variables to model a football game, in theory, if we account for some degree of "free will", we shouldn't get the so much desired 100% efficiency of the model..

(Interesting research indeed, not only for the ?/?/$ that one could win if one gets the right model, but for the learning and the demonstration of what works in astrology, and how "well" it works. :) )

22
Hi jventura
This makes me think that there is something not really 100% true about the Ascendent being always the house team, and the 7th the opponent, at least as in the analysis I've making of this..
That's right. If you don't mind I can send you a table for the Portuguese League. I use mainly home kit colors and sometimes city ruler's the teams represent (if I have), sometimes nickname associations to determine who is who. In the table you will see that Porto, for example, is always Jupiter regardless which sign is rising, Benfica - Mars, Academica - Saturn and so on.

Then apply your algorithm. You shall notice a very pronounced statistical significance when the lord of hour is the same as the significator of a team.

It is true for all leagues regardless of countries, the level of football, etc.

Kind regards

Janis

23
Janis Valkovskis wrote:If you don't mind I can send you a table for the Portuguese League. I use mainly home kit colors and sometimes city ruler's the teams represent (if I have), sometimes nickname associations to determine who is who.
Hi Janis,

yes, I would be most interested in trying a different approach! If you do not want to post your table here, you can drop me an email at skypluxweb@gmail.com.

Have you tested with the entries on your table, and had consistent results?


Thanks,
Jo?o Ventura