Testing the Poisson Model Prediction of the Number of Goals Scored in Premier League Games 1

Testing the Poisson Model Prediction of the Number of Goals Scored in Premier League Games 1

I was just asked a couple of hours ago about the Poisson model and looking at the over/under gold market so so far I’ve always been talking about trying to predict what the match result is gonna be you know whether the home team’s gonna weigh in whether it’s a draw or the away team while here I want to look at whether we get more than three goals I was kind of making some kind of response or reply when I thought actually it makes more sense just to make a series of videos because actually I haven’t studied it I haven’t looked at how how it does the basic idea is we’re gonna use thePoisson model we’re gonna get the data from soccer stats and we’re gonna make predictions we’re gonna make say two things about each match so one I’m gonna look at whether it’s a cur so how often you know when when the Poisson model says I expect the home team to win or say when the Poisson model says that I expect three or more girls is it typically right – if I bet using the Poisson model does it work so if the Poisson model says something is gonna happen with 70% probability so I expect it to happen but the betting market say it’s gonna it’s more like 50/50 then there’s value there so then I die if I used to trust my model I work bad so so for each match I’m gonna try and make a forecast and then make a decision about was it better or not we get the data from soccer stats so I’m gonna be looking at the Premier League so I clicked on England I want Goal’s data so I click on goals I need the goal scored and conceded by the team so I clicked on scored conceded town and I’m gonna use home away because some teams have you know different behavior when they’re playing home in a way so this gives me better resolution I think but again the best way to check is to see how well it does so you know to make make predictions using the average over the entire season make predictions using Home & Away you know and we want the one that does better I’m you can use the same program which I used for predicting results I you can get that program since it’s from like my course speaking or the language of data science so this program the code for that is included in this course so you can just stand out and it also tells you how you can get the program r and install it, which is free.

So mean the r program. it’s also on sale at the moment so very reasonable. so I’m gonna look at three matches Liverpool Bournemouth Southampton Chelsea and burned olestra for each I’m just gonna get the data from soccer stats and I’m just gonna plug them into the program in r in the burnley-leicester match we can see that we expect less than three goals nearly 70% of the time so the model expects there to be less than three goals in the southhampton-Chelsea it expects what’s very 50/50 it expects three or more goals 48 percent of the time less than 3, 51%, because remember this model we plug in the data from soccer stats so this tells me that Southampton on average score one goal a game at home, but they concede one point three one for the league as a whole the average is the home team scores one point 5 4 and the way as we’d expect sir that kind of shows us that typically the home team wins and these are the stats for the away team so Chelsea typically score one point seven three goals per game and concede one point zero seven and what this model does is it kind of normalizes it so say Chelsea typically score one point seven three that’s against a team that would concede one point one eight at home because the league average is one point one eight but in this case Southampton concede one point three one so Southampton have a worse than average defense and that’s why we multiply we expect Chelsea to score more than normal so we can see here the mean for the array team is one point nine one so the model expects Chelsea to score more than they normally do that’s just about you know that’s the internal mechanics of this portal model so as I said it’s 50/50 and for the Liverpool match it expects three or more goals but 57% of the time so you know we are we always have to decide I mean is that sufficient I mean is that is my rule going to be I bet on three or more goals if it’s more than 50% or do I have a threshold do I go I need it to be 60 or 70% so again remember I’m looking at two things for every match one what is the prediction and to do I bet for Bournemouth there’s a clear result it was nearly 70 percent that it’s gonna be less than three goals for the Chelsea match it’s basically 50/50 and for the Liverpool match it’s more than 50 so it is I would say that that’s high enough that I’m gonna expect there to be three or more goals so I’ve talked to three model predictions again I might change as we go through week by week so at this stage I’m I’m interpreting 56 57 percent probability as being enough to tell me that there’s going to be three or more girls but maybe I should be needing 70 or 80 percent I mean I don’t know the second part of the check was to look at the odds to go is this a good pair we can do that by just dividing one by the odds so that’s one point six three and remember if I just click on the probabilities or the odds we can just inverse the access and be given the probabilities so the betting market is saying or thinks that we’ll have less than three goals 60% of the time while my pass or model is saying it’s gonna be 68 so there’s you know there’s a potential that these arts are a bit off according to the postal model so I’m happy to bet – we’re gonna have under three girls I’ve also done a count repair because what you can do on a betting exchange is lay the odds so you can beat the bookie and so what I’ve done is I’ve laid the odds at one point one so that I can kind of reduce my bet the idea being that if we’re saying Neil all a half time then the odds for there to be less than three goals will shoot that shoot down so the odds will maybe go down to one point one one way – and if I sell the – a cheaper price I make profit That’s just you know the particulars of how I’m implementing the bet.

For Chelsea Southampton the odds are 5050 50 so it’s slightly more likely to be under three goals. chelsea Southampton two more goals were slightly under oh no this is saying it’s more likely to be yeah it’s more likely it’s more likely to be less than three goals and it’s more likely to be under three goals according to the betting market, I wouldn’t bet on this because there doesn’t seem to be any value it seems to agree with what model is saying and finally if I look at Bournemouth Liverpool here the odds for over so the model said that it expects there to be three or more goals but it was like 58% I think so it’s nearly 50/50 so the odds really id-1 odds of closer to at least are a bit high for me so if anything I would I would take the counter bet I would bet that there’s gonna be less than three girls because this price is too high sake I will do that I think you can okay you can also kind of shop around with this so yeah if the I was I was able to match my bat at even though if I just clicked I would’ve gotten the idea here is that I think that these odds are too high and now what I’m gonna do is take the counter bet okay so there I if anyone wants a pair that it’s less than three goals even I’m willing to take that bad and I clicked on keep so as the game goes on this bet will stay there so maybe if at halftime or after twenty minutes if it’s still in all people will start thinking oh maybe Liverpool are tired after their Champions League they’re safe in their top four maybe this isn’t going to be a goal fast and so these are two will start throwing up and these hearts will start going down and hopefully I’ll be able to find a match for this pad and then I’m happy because then I can’t lose then if it ends up being a low-scoring game I win if it’s a high-scoring game I don’t lose which is the kind of that I like so I went on a bit in this one because again I just got the question from the viewer so you know it wasn’t not clear what is exactly it is that I was I wanted to do here so I would say from the point of view of going through the five weeks of looking at the over unders the idea is I make a prediction for each result and I make a bet or I avoid it so from the point of view of this series I’m gonna imagine I will imagine that I bet 10 euro or I didn’t bet in reality I didn’t but just to keep score so I would bash that we’re gonna have blunder 3 goals for the Liverpool Bournemouth I would bet that we’ll have under three goals for the Burnley Leicester and I I wouldn’t bet on the Chelsea match because there was no value so we’ll see how that goes and then we’ll be able to see whether the this model like this implementation of the Poisson model at value whether whether it’s actually the numbers it produces are Acker and provide a way to make money fingers crossed



Comments are disabled for this post.