Using odds to predict scores

[duckmaster]

Hello

I decided to put all of last years data from the English leagues into a spreadsheet to show me which scores were most prevelant depending on the odds of the home team. I came up with the first chart (attached). The odds of the home team are shown at the top (odds according to Bet365), with the number of times each scoreline occurred following downwards. From this I made a second chart showing the percentage of times each scoreline happened, and from that I made the final chart showing the scores ordered in how many times they appeared for teams with those odds (bold and italic numbers are tied with the same colour of box in their column).

Lets take the 1.81-2.00 range of odds and find a game with similar odds for the home team. I'm going to go with Club Atlético San Lorenzo de Alem v Racing de Cordoba (mouthful!) in Argentina at 11.30pm tonight. According to my chart the most likely score is 2-1, with Skybet offering odds of 8.50 on that. But they have lower odds than that on FIVE scores (1-0, 2-0, 0-0, 1-1 and 0-1) suggesting my bet is a "value" bet.

I'm going to take the following bets:

Club Atlético San Lorenzo de Alem v Racing de Cordoba - home odds 1.80; most likely score 2-1; odds 8.00, 4 scores with lower odds

Instituto de Córdoba v Quilmes AC - home odds 1.73; most likely score 1-1; odds 6.5, 1 score with lower odds & 2 with equal

Club San José v Oriente Petrolero - home odds 1.22; most likely score 1-0; odds 11.00, 4 scores with lower odds

 

I'm going to take that as a £7 patent bet, covering the singles, doubles and treble with £1 each and see what happens!

If one wins I'll either lose 50p or make £1.50 (depending which one), if two win I'll make £63-106 profit and if all three then £847!

 

Let's see what happens!

[duckmaster]

First coupon

 

Club Atlético San Lorenzo de Alem v Racing de Cordoba - home odds 1.80; most likely score 2-1; odds 8.00

LOST - SCORE 2-0 (HT 1-0)

RANKING OF 2-0: 3rd

 

Instituto de Córdoba v Quilmes AC - home odds 1.73; most likely score 1-1; odds 6.5

WON - SCORE 1-1 (HT 0-0)

RANKING OF 1-1: 1st

 

Club San José v Oriente Petrolero - home odds 1.22; most likely score 1-0; odds 11.00

LOST - SCORE 5-0 (HT 1-0)

RANKING OF 5-0: UNPLACED (but was 1st in the 1.0 - 1.2 odds column?)

 

Will need some tweaks but that's a 50p loss from £7.00. Any suggestions?

[duckmaster]

To be clear I bet £7 and got £6.50 back so it's a return of -£0.50

[duckmaster]

A treble on the two wins and a draw would have been approx 8.00 or 7/1

[duckmaster]

Slightly different here: El-Masry v El Daklyeh, home odds 1.75, 3pm Monday (GMT)

Expected score 1-1

 

1-1 @ 6.00

1-0 @ 5.50

2-0 @ 7.00

 

£1 each

[liero1]

Interesting. What I don't get is why do you analyse English Leagues and then start betting on non English leagues.. Surely the market mechanisms can't be transferred 100% across leagues.. ?

[eddiey]

Great work. This is really interesting. 

[duckmaster]

7 hours ago, liero1 said:

Interesting. What I don't get is why do you analyse English Leagues and then start betting on non English leagues.. Surely the market mechanisms can't be transferred 100% across leagues.. ?

I forgot to put in the original post that I also entered all the stats from all major leagues from last season and the table didn't change that much. The sheet I've made has all data for the following leagues from last season:

England & Scotland - top four leagues

Germany, Spain, Italy & France - top two leagues

Holland, Belgium, Portugal Turkey & Greece - top league

Its 7,258 games total.

 

My theory is that the difference in scores between leagues is negligible. Whilst the quality of player in the English Premier League is overall far higher than the quality of player in the Dutch Eredivisie, the difference in quality between players in their own league remains about the same. For example, a top striker in a top six English team may be "ranked" 85/100, but the defenders he is coming up against are also ranked in the 80-90/100 range. A top striker in Holland may only be "ranked" 65/100, but because the quality of defenders is 60-70/100 the overall difference in quality remains the same. (note that these are rankings I just made up to make my point, they're not based on anything!)

A football match is two halves of 45 minutes for 90 minutes total, with 11 players on each team including 1 goalkeeper, plus a referee. Those are the only facts which remain constant whether the game is being played in England, South Africa, Brazil etc. When looking at statistics, and assuming the quality difference of players in each league is reasonably constant, the same scorelines should occur at approximately the same rate in one league as in another (there are other factors like number of teams in league, number of games etc as well of course, but the main ones are the fact the games are the same length and played under the same rules with a similar quality of players on the pitch).

Another interesting thing I have discovered with spreadsheets is this: assume the following scenario:

Team A is at home to Team B. Team B scores two goals in the first half. The score is 0-2 at half time. In the second half, Team A scores 3 goals to go ahead before Team B equalises. The score finishes 3-3.

In a 20 team league there will be 380 matches. The above scenario will occur approximately the same number of times per season as the season before that. It is impossible (probably) to predict the game in which it will happen, but it is extremely probable that, at some point in the season, it will happen.

 

15 hours ago, duckmaster said:

Slightly different here: El-Masry v El Daklyeh, home odds 1.75, 3pm Monday (GMT)

Expected score 1-1

 

1-1 @ 6.00

1-0 @ 5.50

2-0 @ 7.00

 

£1 each

Anyway this was 1-1 so £3 bet and £6 back, £3 profit

[duckmaster]

If anyone knows of a site where I can get historical data for other leagues that would be helpful. Particularly interested in India and China - would be interesting to see if the theory stacks up in the "wild west" of football"!

[doberman1222]

Check out this dataset https://www.kaggle.com/austro/beat-the-bookie-worldwide-football-dataset.

There is a file here (closing_odds.csv) which contains almost 500k football games (from "2005-01-01" to "2015-06-30" with odds), so it may be helpful.

[duckmaster]

I have made a new table which converts the percentages into odds.

What to do is look for the home odds column for the team you are interested in. The numbers which correspond to the scores are the ideal odds the bookies should be giving you for that particular score for that particular team. If they're offering higher odds, that's a good bet.

Looking at Liverpool vs Napoli for tomorrow, Liverpools odds are 1.62. So on the 1.61-1.7 column we can see the odds for the following scores are:

1-0 7.05

2-0 9.90

3-0 11.52

4-0 41.25

0-0 15.10

1-1 7.05

0-1 25.00

0-2 66.60

2-1 5.76

1-2 27.17

 

skybet are offering us:

1-0 10.00 (good!)

2-0 9.50 (bad but only just)

3-0 13.00 (good)

4-0 23.00 (terrible)

0-0 19.00 (good)

1-1 8.00 (good but only just)

0-1 19.00 (bad)

0-2 34.99 (bad)

2-1 8.50 (good)

1-2 15.00 (terrible)

 

The value bets here appear to be 1-0, 3-0 and 2-1 as Skybet are basically rating the chances of that score happening as lower than my chart suggests they are. i.e. my chart suggests it has a 14.8% chance, Skybets odds suggests it only has a 11.76% chance.

Then again its a Champions League game so not touching it, but just as an example!

[duckmaster]

Arsenal de Sarandi v Nueva Chicago in 2 minutes

1-1 6.00

2-1 8.50

3-2 29.00

2-3 41.00

0-3 41.00

[duckmaster]

will explain reasoning later, this is a bet based on a combination of my probability and value charts

[liero1]

@duckmaster A football match is two halves of 45 minutes for 90 minutes total, with 11 players on each team including 1 goalkeeper, plus a referee.

AHH! that's what it is all about, I was wondering

JK, great input here, good inspiration! We should compare your forecast/value calc with a poisson distribution from CGMBet or other resources.

[duckmaster]

TEAM	SCORE	ODDS	POINTS
WEYMOUTH	1-0	10.00	1
ST JOSEPHS	1-0	12.00	1
	2-0	10.00	1
OXFORD CITY	1-0	7.50	0.5
RAVENNA	4-3	101.00	0.5
	5-0	51.00	0.5
	2-2	19.00	0.5
	3-3	81.00	0.5
	4-1	29.00	0.5
SLOUGH	1-0	8.50	0.5
	2-1	7.50	0.5
GIANNA ERMINIO	3-2	29.00	0.5
ABERDEEN	3-2	51.00	1
	0-3	81.00	1
	2-3	81.00	1
	2-1	9.50	0.5
	4-2	101.00	0.5
	4-3	251.00	0.5
	2-2	23.00	0.5
	3-3	151.00	0.5
WESTON SUPER MARE	0-0	13.00	0.5

 

[duckmaster]

Above (and game above that, which finished 2-0) is based on expected odds according to my chart. If bookies odds are at least 50% greater than those odds (i.e. my odds are 10.00, theirs are 15.00), its a half point bet, if bookies odds are 100% more it's a 1 point bet.

13.5 points total... we'll see what happens 

 

edit: in another words, I think the 0.5pts bets are 50% more likely to win as the bookies do, and the 1pt bets are twice as likely. doesn't mean they will of course...

Edited December 11, 2018 by duckmaster
addition to previous post

[duckmaster]

All of them are goosed except Aberdeen which is 2-2

Will return 11.5pts from 13.5pts so a 2pt loss

2 mins left for another goal though! 

[duckmaster]

That'll do! Aberdeen 3-2 gives me 25.00 profit plus 0.5 stake back, 13 stake lost for a profit of 12.5

 

Where can I host an excel file online and I'll show you the calculator I've made?

[liero1]

just use google sheets @duckmaster ?

[duckmaster]

4 hours ago, liero1 said:

just use google sheets @duckmaster ?

Worried that the file size is going to be too big but here it is 

edit

calculator is wrong, please ignore for now. working on it

Edited December 12, 2018 by duckmaster
wrong link

[duckmaster]

Should say that this is all a bit fly-by-wire, I've never seen anyone do anything like what I'm trying to do here so don't blame me if it all goes wrong! 

[duckmaster]

6 hours ago, duckmaster said:

TEAM	SCORE	ODDS	POINTS
WEYMOUTH	1-0	10.00	1
ST JOSEPHS	1-0	12.00	1
	2-0	10.00	1
OXFORD CITY	1-0	7.50	0.5
RAVENNA	4-3	101.00	0.5
	5-0	51.00	0.5
	2-2	19.00	0.5
	3-3	81.00	0.5
	4-1	29.00	0.5
SLOUGH	1-0	8.50	0.5
	2-1	7.50	0.5
GIANNA ERMINIO	3-2	29.00	0.5
ABERDEEN	3-2	51.00	1
	0-3	81.00	1
	2-3	81.00	1
	2-1	9.50	0.5
	4-2	101.00	0.5
	4-3	251.00	0.5
	2-2	23.00	0.5
	3-3	151.00	0.5
WESTON SUPER MARE	0-0	13.00	0.5

 

@liero1 mentioned earlier that I looked at stats for certain leagues then bet on others, my theory being that with this market the scores will remain consistent over different leagues... BUT the only bet that won (or even came close!) was Aberdeen in the SPL whose league are in the stats I am gathering the data for!

 

Will stick to the leagues I've got the data for... so probably no more selections until Friday

[duckmaster]

Just trying out with champs league...

 

REAL MADRID	1-0	10.00	1
	3-2	26.00	0.5
	2-2	23.00	1
	3-3	81.00	1
	0-2	81.00	1
BENFICA	2-0	7.00	0.5
	3-2	34.00	1
	4-0	15.00	0.5
	2-2	23.00	0.5
	3-3	81.00	1
	0-2	51.00	1
MAN CITY	1-0	13.00	1
	4-0	12.00	1
	5-0	19.00	0.5
	5-2	51.00	1
	2-2	21.00	1
	3-3	67.00	1
	0-2	81.00	0.5
	1-3	81.00	1
	2-4	201.00	1
SHAKHTAR	1-0	13.00	1
	4-0	51.00	1
	4-1	34.00	0.5
	5-0	101.00	1
	5-1	81.00	1
	5-2	81.00	1
	3-3	34.00	0.5
	2-4	51.00	1
VALENCIA	3-2	34.00	1
	4-0	67.00	1
	4-1	51.00	1
	5-0	201.00	1
	5-1	151.00	1
	3-3	51.00	0.5
	1-3	26.00	1
	2-3	34.00	0.5
	2-4	81.00	0.5

[Kentucky110]

@duckmasterHi There have you managed to sort the link out yet?

Kentucky110