Soccer Stats: Distribution of goals throughout a match

I find the Internet is a good resource for this kind of information.

The best way is to keep your own records of matches, Betfair Live Score will give you all the goal minutes you need to record, or Flash Scores etc. That way you can also note the all important odds bands.

For example, last week I recorded 160 matches with 0-0 ko odds above 10.0, and 20 of them had a goal in the first 5 minutes - approx 12%, seems a lot doesn't it. Approx twice as many had a goal in the final 5 minutes.

But of the 52 matches with 0-0 ko odds less than 10.0, only 1 had a goal in the first 5 minutes - approx 2%.

For any other 5 minute sections you can then consult your records :)

Thanks for the advice Trev,

But recording by hand is too much trouble and this data must exist - the minute a goal was scored is fine for my purposes.

I think I'll email BF soccer results provider and see what they can do.

Thanks again.

Who said anything about collecting by hand? Write a bit of software to get the scores from betfair or some other live scores site (use more than one so that you can be sure the data is correct) and store it in a database (mysql or even access). then you can run as exotic analysis as you could ever want on the raw data and price your markets up based on it.

There are probably sample datasets out there already, might be worth trying to google them first to see how much use you could actually make of it.

There is a mathematical method for calculating goal time.

You use negative binomial distribution with your goal expectation to achieve a value. When the cumulative of these values passes 50 it is more likely than not that a goal will be scored.

have a look at soccerstats for some leagues

http://uk.soccerway.com/teams/england/manchester-united-fc/statistics/

not 5 minutes but 10 minutes interval with percentage of goals

Not sure what use stats from those websites will be for something like this.


If your GE is 1.6 for Team A then the chance of Team A scoring their first goal in the first 60 seconds is 0.0177

The probability of them scoring the first goal in the second 60 seconds (1-0.0177)*0.0177

Once you've done this in Excel you sum them and look for the cumulative total passing 50%

eg

Minute    Cumulative Total
50      0.49986
51      0.50939

So first goal for Team A is most likely in the 51st minute (or between the 50th and 51st) You'll find you may have to play around with it a bit though.

Couldn't you just imagine some geeky assistant to the Mgr. sitting on the sideline with his I-pad working on all these binomial distribution figures and screaming into the Mgr's ear that a goal is coming from the opposition in the very few next mins., and that he bleedin well ought to do something about it ?

http://www.soccerbot.com/fa/results/ukprem2011.htm has the times of all goals scored in the premiership.  You'd need to scrape it or enter the values by hand into an Excel sheet.

Hello there Rocket. I don't know if I am miscalculating this as it is not my area of expertese (apologies if I have got this wrong) but don't you factor in injury time into your scoring probability.

I assume that your expectation of Team A scoring a goal in the first 60 seconds (0.0177) is based on multiplying that by 90 to get 1.6, but would it not be fair to say that the average game is more likely 93 minutes long rather than 90.

Or are you just using a basic example to illustrate a point and I am looking too much into this.

Probability that's all.

Rocket
You seem to be giving out an extraordinary amount of  revealing analytical advice today by your standards of " secrecy " at least.
Or am I misreading you as usual.

UA I chose 90 minutes.

By all means use a league average if you like.

Ok thanks Rocket, I assume there are a whole load of other variables that you need to factor in as well as i doubt that each minute would carry the same probability (i'm asking for you to divulge any of these) but I understand the point you are making.

There are a lot of factors you need to take in.

The main one being a low GE may well give you a goal time higher than the match length.

Rocket,

I hope you're not just dividing 1.6 by 90 to get the probability of a goal in the 1st minute as U.A. said! That figure (0.017778) is rather, the expected number of goals in the first minute (assuming that goals are evenly distributed throughout the match). The probability of there being a goal in the first minute is, it is generally reckoned, best calculated using a Poisson distribution (in this case = 1 - POISSON(0, 0.017778, TRUE) (in Excel).

Do you reckon Bill Shankly knew all this stuff about when the opposition was most likely to score a goal ?

I've been reading through this thread.

About a second ago, I heard a ripping sound coming from behind me.  I realized it was the back of my brain

No, but you can bet John W Henry does.

lori which are the good bookies for IR basketball?

Yeah well the jury's still out on him, but not for good old Bill.

Thanks for the advice everyone.

Goals scored may or may not conform to a binomial distribution. I suspect they do approximate but with significant deviations.

I'm interested in where those deviations exist, if they exist at all.

http://uk.soccerway.com/teams/england/manchester-united-fc/statistics/

Thanks for the link, Lusitano

Goes part of the way toward what I want, and there does seem to be deviations. Although unfortunately they seem to lump extra time goals into the last 10 mins of play resulting in a strong final 10 minutes bias, which may be more apparent than real.

pxb     08 Feb 11 02:45 
Thanks for the advice Trev,

But recording by hand is too much trouble and this data must exist - the minute a goal was scored is fine for my purposes.
____________
laziness is the easiest route to failure .
why would you trust someone elses data above your own ?

It wouldn't be my data. It would be someone elses data I had transcribed.

Their data + my transcription errors.