Football Data (CSV, JSON) - UPDATED 16/08/17

[Euler]

Red cards would obviously have a huge impact so I think it would make sense to list them.

[welshboy06]

Yeah that's what I thought.

So there are 3 options

1) I can either dedicate a fixed amount of columns to reds/goals (The way it is currently, just extend the Red's a further 5 columns)
Then loop through each and place them in to the designate columns.
e.g. 66, away, , , , , , , , , 31, home, 33, home, 70, home

2) I can loop through the reds then immediately loop through the goals (I would need to add the "type" of event, since otherwise you wouldn't know if it were a goal or a red.
e.g.
66, away red, 31, home, goal, 33, home goal, 70, home, goal

3) I can loop through the events, as they happen (e.g. goal, goal, red, goal) and list them like that. I'd again need to add the event "type" so people could see if it were a goal or a red.
e.g.
31, home, goal, 33, home goal, 66, away, red, 70, home, goal

[Euler]

I've traditionally looked at reds independently of goals, but I guess that including the red in the goal sequence would allow you to treat a red as a mini goal.

[welshboy06]

I've traditionally looked at reds independently of goals, but I guess that including the red in the goal sequence would allow you to treat a red as a mini goal.

Okay so there are now two sets of CSV's.

• Goals and Reds Separated (Each type having a dedicated 15 columns)
• Goals and Reds Together (Sorted in time order)

[Sickmund]

The possible influence of time of the 1st goal is certainly something I want delve in to when I'm finished looking at other areas of possible opportunities. This data will suit my research, thank you very much!

[Euler]

Thanks for the data, it's really helpful to combine much older data with the stuff I've collected more recently. Probably deserves a thread on its own to interpret.

[Westerner]

Thanks for this Adam. Great job.

Could well be the kick I need to take another look into this.

[welshboy06]

Thanks for this Adam. Great job.

Could well be the kick I need to take another look into this.

No Problem, any issues with the data or whatever just let me know. Now the fun part comes... analysing it all

[Dallas]

Some good work there Welshboy06

[Dallas]

In the interests of a bit of fun I thought I would do a simple program to quickly lookup what the prices of an outcome should be following a goal based on JonnyG's early goal theory and his way of pricing a market, below is a copy of the table which the program uses so that you can see most outcomes at once (just couldn't quiet fit them all on screen). Put simply when a goal is scored by the home team in the Premier League you just look at the minute it was scored then work your way along to see what the price of a home win, away win, O/U1.5 markets etc etc should be (according to JonnyG's theory) – this should save him tons of time copy and pasting when he’s doing his in Prem lge in-play analysis in the future .

For those who couldn't understand his posts the general gist was he believes the market often misprices the outcomes following a goal and that the correct way to price an outcome is to simply discount everything except for
"dividing the number of times an outcome has occurred by the number of goal scored on that minute".

So using the table below if the home team scores on the 3rd min the odds of the away team winning the match should be 18.67 because they have only won 3/56 times when a goal was scored on the 3rd min

PL.JPG

NB, im therefore not recommending anyone follow this as any sort of betting or trading strategy the purpose of this post was solely to highlight the way he thinks matches should be priced as I know it was very confusing for most ppl the way it was posted so i'll just leave it up to the readers if they think this is a viable and profitable way to trade/bet on football markets

Personally I would hope anyone that is using the goal timing data Welshboy06 has kindly scraped and shared will put it to much better use by finding themselves something far more relevant and useful which they can hopefully use to actually make a profit with

[Westerner]

Nice post Dallas.

Main problem I 've found when analysing goal times is sample size. If we breakdown into leagues like you've done with the Premier Leaguein we can see that the numbers are quite small for each 1 min bracket.

If we think about how each game has a different profile (e.g. strong home fav v weak away team) and start adding this to the mix the numbers decrease very quickly with each goal scored.

To get over this I decided to bracket goal times, which isn't ideal but it does bulk up the numbers to smooth over the bumps. Would be interested to hear if others have a better way once they start going over the data.

[spreadbetting]

Big thumbs up for the new CSV data welshboy, might be an idea to include the country and leaque in separate columns too as then the data could all be dumped in one file, or database

[Dallas]

Nice post Dallas.

Main problem I 've found when analysing goal times is sample size. If we breakdown into leagues like you've done with the Premier Leaguein we can see that the numbers are quite small for each 1 min bracket.

If we think about how each game has a different profile (e.g. strong home fav v weak away team) and start adding this to the mix the numbers decrease very quickly with each goal scored.

To get over this I decided to bracket goal times, which isn't ideal but it does bulk up the numbers to smooth over the bumps. Would be interested to hear if others have a better way once they start going over the data.

Thanks,

That imo that is another of the major flaws in JG's approach, he never factored in starting odds which as you have highlighted are a major part if the previous data of 10 games was from matches with a strong home fav and this current match has a weak home fav it renders it useless especially from a perspective of trying to price a market. Then break it down into odds and you end up trying to say 'well this has happened 2 out of 4 times in last 12 years so its 50% it will happen this match therefore odds should be evens!

I've had this type of data some time just never done much with it but when i do ill be looking a more longer term patterns and not at things like 'if a goal is scored on 8mins then the following will happen next but if its scored on the 9mins then something different will occur'!'

[spreadbetting]

I think there's always a tendency to backfit systems when you drill down data to the nth degree and start looking for patterns in what's effectively random data.

[deansaccount]

I was wondering, does anyone in here have a genuine edge in a premier league market that has been effective over a few seasons? I'm wondering if it's really possible as the premier league seems like it has very efficient markets.