Home Soccer Soccer Betting I am trying to build a model. Help Plungeinto Soccer betting

I am trying to build a model. Help Plungeinto Soccer betting

3 min read
8
0
10

Plungeinto Soccer betting. We give soccer results and expectations. Pick soccer league and you will discover statistics, picks, tables and data for all your betting needs. All football data on this site is free.

Hello all. I have been interested in soccer for many years, and I find modeling it to be fascinating. I feel as though I understand the value sets of soccer quite well, and I know what to look for when it comes to using which data for my model. I understand the concepts of which data to use and which data to overlook however I am having trouble understanding which variables to plug in. I have been trying to use Poisson distribution, but I am having a torrid time at it. The technicalities of mathematics and statistics are not my strong point. I also thought about using Bill James models, and more easily straight forward baseball models and translate them into soccer terms. I found this video and channel to be helpful, but the author of the video doesn’t explain how he created his model.

As soon as I find out how to plug in my values. I am sure I could make a decent model.

Soccer Betting Guide – Reviews, Tips, Bonuses and Promotions! This soccer betting guide has been arranged to give first time and inexperienced soccer bettors and additionally any individual who is knowledgeable at putting down soccer related betting and bets a more profound understanding into the advantages and entanglements of setting any kind of soccer wager. We will be telling you how you can spot conceivably compensating wagering openings, also helping you locate the best betting site to put down your wagers with, to guarantee a you a much good betting position.

AliDropship is the best solution for drop shipping
Load More Related Articles
Load More By plugeldadmin
Load More In Soccer Betting

8 Comments

  1. vaiix

    September 26, 2017 at 3:04 pm

    [Using R, Advanced](https://teouchanalytics.wordpress.com/2017/07/08/episode-1-feature-engineering-and-some-data-to-play-with)

    [Using Excel, Intermediate](http://www.petermckeever.com/how-to-compile-odds-for-football-a-beginners-guide-part-i-getting-data-and-finding-xgoals)

    [Using Excel, Very basic](https://differentgame.wordpress.com/2017/04/29/an-xg-model-for-everyone-in-20-minutes-ish)

    I’ll echo what others have said here, you’re either going to be producing something that’s too basic and therefore ineffective, or delving in to something you don’t understand as you haven’t the statistical knowledge.

    Even the advanced model above isn’t going to produce repeatable results in its current form – the whole data modelling football community agrees that you need location tracking data to produce a model that may hold some significance, but the data isn’t readily available to the public.

    You may want to look in to game state, for example, would you include any goals in analysis that are beyond 2-0? If a game is 5-0, do you weight goals after 2/3-0 as are both teams even trying? What if the losing team scores late one to make it 5-1, do you count that since both teams are just seeing the game out?

    What about possession, is all possession useful, or just when it’s in dangerous areas? What about shots from 5 yards out, is one from an inconceivable angle counted the same as one from directly in front of goal?

    Variables are endless, and those that are considered vital are unavailable to us mere peasants. Build your own model for enjoyment and interest, but don’t expect to uncover anything to make you a repeatable profit.

    Reply

  2. Millkovic

    September 26, 2017 at 3:04 pm

  3. the_wizard_of_odds

    September 26, 2017 at 3:04 pm

    > As soon as I find out how to plug in my values. I am sure I could make a decent model.

    This is a minor mistake in your thinking. Choosing these variables is the hardest part. From there everyone can make a *decent* model. We are basically facing the following problem of some input X, some Transmission-function G and and an output Y

    Y = G(X)

    X is a vector and Y is either a vector or a simple number (scalar). G could be anything, a neural network a simple 1 dimensional equation like y = g*x or something completely different.

    Y is whatever we want to compute. Some people try to compute the odds, some people try to compute the likelihood of goals some people try to predict the weather. Now we have to think about, what we need to approximate this output. If we have the odds as Y for example we would need the mean goals a team scores as X (among many others). This X is maybe computed from all games this team has played this season (we call these games samples). Now we have fixed X and Y. Finding this transmission function G is a machine learning problem. One (old fashioned way) to find G is by using the Poisson distribution as described by Dixon and Coles.

    You could for example find it by linear regression (probably the simplest way to do it, yet kind of effective). Let us assume for now (to make my explanation simpler) that we only have one input X, the number of goals the team scores in the mean. And we assume that we have only one output, the probability that this team wins. Now we find the transmission function G by regression or whatever technique you like. Now G could be 0.3. I.e. if a team scores 3 goals in the mean our model says by y = G(=0.3)*X(=3) that the probability that this team wins is 90%. I completely made the numbers up (please don’t use it like this).

    Now this model is obviously to simple but now you can make your inputs more complex, i.e. you simply take more data into account (for example the mean goals they receive). Now your vector X has two entries for each game. The art of making these models is choosing the right input. The problem is basically, the more entries your vector X has, the harder it will be to find the transmission function G accurately with the same number of samples (this is called curse of dimensionality). There are now sophisticated techniques (called feature extraction (e.g. by Singular Value Decomposition)) to reduce this input dimensionality.

    Reply

  4. imthirstybuymeadrink

    September 26, 2017 at 3:04 pm

    Hey guys. Davicing and the wizard of odds I really appreciate your input. I am going to share with you my thoughts regarding my interest. Sorry if I go off a tangent.

    I guess what I will do is try to explain what variables and value sets I want to utilize for my model, and hopefully you can share with me how to set it up.

    Let us use The Man U opener as case study.

    Manchester United (Home) vs West Ham (Away) as an example. Man United is at 1.3 to win, the draw is 5.4 and a West Ham win is 11.1. My inclination is that odds for Man U do not have any value. Best case scenario the odds are just about right for Man U, but probably negative value.

    Finding out the offensive strength and defensive strength is quite straightforward. What I would do is find the mean of goals scored and goals allowed. And come up with a scoring strength value and defensive strength value which would be based how much better or worse than the average strength (goals scored goals conceded value of the league.)

    Although not perfect it would have to be based off of last season’s tallies. However I would tweak it slightly based off of the strength of new players and departures. For example I would calculate what how many goals Lukaku scored last year minus Ibrahimovic’s.

    Ultimately, another reason why I think Man U odds at 1.3 are quite low is because of the overreaction of Lukaku’s arrival. Lukaku is better than Ibra. but not by that much based off of goals scored. After doing quick research. Lukaku is healthier than Ibra was last year. As Lukaku played around 20 percent more games than Ibra, but only has a 4 percent better chance than Ibra at scoring per game. I was surprised by the 4 percent improvement as the media hype would make you think otherwise.

    I don’t think I need my model to be that accurate. It simply has to be able to isolate value. I just find it very hard to believe that Man U has better than a 77 percent chance at beating West Ham at home and West Ham has less than a 11 percent chance at beating Man U away. This might be an ad hoc rationale, but if I am correct in realizing a market inefficiency regarding my belief that 77 percent is far too high for Man U, and 11 percent chance for West ham to win is a little low, if I can create a model, that will justify my beliefs and I add a few tweaks and extra variables I think I could create a model that would generate positive R.O.I.

    Another idea of an extra variable or tweaking I would add would be i the event of a striker injury, where a weaker alternative would start up front. I would simply minus the offensive strength by 10 percent for example.

    I’ll end with, this is sort of a question/idea/ thought out loud. An advanced model as possible with as many data sets and variables is obviously desirable but my belief is too complex of a model with too many data sets could actually negatively affect the model to the point of making it inaccurate. Is this true?

    My inclination is a simple, yet straight forward model, rough around the edges yet efficient is better from a time put into the model/benefit analysis as a complicated model could be inaccurate due to using too much data that would be considered white noise.

    This may seem like a shit example, but a model that inputs variables such as refs, striker performance trends during certain monthes would just be considered useless data.

    Reply

  5. davicing

    September 26, 2017 at 3:04 pm

    Use poisson distribution models as a learning tool for more advanced models but don’t ever bet with real money using one of them.

    Also, you don’t need to have the knowledge necessary to solve the poisson equation with a pen and a paper, you only need to understand what it does.

    That youtube video explain what it is the most spread around poisson distribution model. Basically it calculates what it is called the “Scoring Power” of both the Home and Away team based on the strength of their attack and defence and it calculates the percentage for a given number of goals.

    In excel the formula would look like this

    =POISSON.DIST(2,1.25,FALSE)

    This would calculate the probability of a team with a Scoring Power of 1.25 to score two goals

    Reply

  6. ExeterQuickly

    September 26, 2017 at 3:04 pm

    You can use the POISSON function in Excel to work out the probability of a game finishing 0-0, 1-0 etc. all the way up to 10-10 and use this to calculate probabilities of win/loss/draw. This is only the start though, there are other factors that you need to consider such as transfers, injuries, suspensions, etc. and you will have to decide how much to adjust the probabilities based on this.

    Reply

  7. essbeck

    September 26, 2017 at 3:04 pm

    Basic thing that you need.

    A constant. Something that is the same for each match so you can compare your analysis.

    Keep detailed track of your results for your own sake so you can figure out what type of bets that work and doesnt work.

    Reply

  8. zombrex2311

    September 26, 2017 at 3:04 pm

    I’ve gotten really interested in creating a model. I’m just tired of throwing bets away without any thorough study. Sadly, my computer doesn’t have excel, I have to buy it. So is there any other way to utilize the Poisson without having it to be on excel?

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *

Check Also

Real Sociedad 2-[2] Barcelona – Luis Suarez

Plunge into Soccer and get all the most recent football news, betting tips, and advancemen…