under,over ,seasonality,xg,,i'm in difficulty

[vhdgkl]

I am collecting a bit of data to learn the live football market... in particular I look at the line of the most probable odds of the under over(1,95-1,95), and during the match I check every minute the line goes down by 0.25 goals... then 0.5... then 0.75 etc etc... on the main bookmakers, including Betfair.(2,5 @1,95 start,2 @1,95 at 28',1,5 @1,95at 44' etc)

I have to say that the trend of the live odds of the under over is very similar to the initial odds...it is not affected too much by the development of the game.
In the event of a goal, the next price inherits the previous line (more or less).

I'm not good with PCs, so I do everything by hand (even if I write it down in Excel). I started a few days ago and I have collected 8 live games with their most likely goal lines. 8 games.
8 under in a row.all!
1 chance in 300 or so.
OK, it may be a coincidence, but I'm not convinced... they are all first days of various leagues... is it fair to assume that bookmakers make more money by overpricing the over in the first few days?
I used to accept that Pinnacle's line is right, but I'm beginning to think that it is built to make more money than the margin it imposes.
After all, how much money is put on the under and how much money is put on the over? A lot more on the over for sure...how can i trust the odds of some book, or even betfair, and accept that they are right?


This thing that I have collected 8 games, and 8 times in a row and 'came out under I was very very surprised ... and I think it might hide some trick to cheat the masses ... the fact is that if I can not trust a reference as pinnacle, or betfair, it all becomes an incredible difficulty ... you have suggestions to improve?

I'm also very surprised by another thing: during these 8 games (that I have watched), the xg's seem to put less energy into scoring goals when they have the chance...for example, if lewandovsky from a certain position scores once in 3, well during the first few days of the season he scores once in 5...

i need some help on how to move through this thicket


ps: match was...
aalborg midtillan(denmark) line 2,5....0-1
shaktar donetsk-inhulets line 3,25 2-1 ukrainian premier league
werder brema-hannover line 2,75 1-1 zweite bundesliga
wacker innsbruck-sc imst line3,25 0-2 3liga austria
energie cottbus-luckenvalde line3 1-1 3liga germany
cophenaghen-silkeborg line 2,75 0-0 denmark premier
lask- rapid wien line 2,5 1-1 austria bundesliga


roma siviglia line 2,75 0-0 ..ok it's' a club frinedly


it's too weird....there's some bookmaker trickery going on in my opinion...but how can I tell if this is the case?

[ShaunWhite]

Just be mindful of the fact that it's you and your 8 matches (basically a sample size of zero) and no real data skills vs Pinnacle and their PhDs and data scientists. I'm not saying you're not right, but if the Pinnacle figures look wierd to you it might be that it's your numbers that are wrong rather than Pinnacle's? Also prices aren't 1 or 0, just because something happens 8 times in a row it doesn't mean 2.0 was wrong or that 1.13 would have been right.

The first rule of (data) science is try and prove your theory is wrong rather than proving it's right. Try to think of as many reasons as possible why you could be wrong, address each one and then repeat your comparison with a different set of matches. That gives you a basic level of rigour and std in & out of sample testing. Good enough to start with and you can get more sophisticated as you progress. Keep collecting data, you'll need several hundred before it becomes genuinely useful.

[vhdgkl]

Just be mindful of the fact that it's you and your 8 matches (basically a sample size of zero) and no real data skills vs Pinnacle and their PhDs and data scientists. I'm not saying you're not right, but if the Pinnacle figures look wierd to you it might be that it's your numbers that are wrong rather than Pinnacle's? Also prices aren't 1 or 0, just because something happens 8 times in a row it doesn't mean 2.0 was wrong or that 1.13 would have been right.

The first rule of (data) science is try and prove your theory is wrong rather than proving it's right. Try to think of as many reasons as possible why you could be wrong, address each one and then repeat your comparison with a different set of matches. That gives you a basic level of rigour and std in & out of sample testing. Good enough to start with and you can get more sophisticated as you progress. Keep collecting data, you'll need several hundred before it becomes genuinely useful.

I took your advice.I enlarged the data sample a bit...probably not enough..

I took the 8 matches I had recorded on excel, and looked at which league they belonged to....
1)denmark superlig
2) Ucraynian Premier League
3) 2. bundesliga
4) 3. liga germany
5)austria bundesliga
6)3.austria 3liga

These leagues are on their second or third round.
i went to the odds of each match and took the predicted goal line, including 2.75 3.25 etc... it is not a perfect way, considering the 0.25. however, between a few for and a few against, there did not seem to be much difference. not many, but a few matches ended void, i tried to favour the over in case of a very doubtful line.

and so


1)denmark 12matches: 9under-3 over
2) Ukrainian 11matches: 6 under -5 over
3)2.bundesliga 15matches: 5 under-10over
4)3.liga germany 22matches: 15under-7over
5)austria bundes 9matches:5under-4over
6)3liga austria 10matches: 5under-5over


dividing by leagues:
more under than over in 4 leagues, more under than over in 1 league, 1 draw in a liga ...4-1(1)


dividing by matches:
45under -34 over 57%under


--------------------------------------------------------------------------------------------------------------------------

I also wanted to add a filter, removing the German leagues from the calculation, as they are part of countries that have the biggest leagues and so there may be more knowledge around...so, it is a filter I have tried to add,removing the german leagues
25under-17over 59%under

with or without the filter the result is similar.


it seems to me that there is indeed a tendency to misquote the over line,in the first rounds,by bookmaker.

the conclusion is that the data is too little? you can't really deduce anything from this small statistical sample? what would you suggest?