I only came across this video recently and have used the method this season. Winning just over 80% of the time with just a minor tweak of my own. Cheers Peter
BTTS and corners usually have good odds with a decent probability of the outcome occuring... tricky part is knowing the teams/matches with the high probability.
Very interesting model, certainly needs to add some form of weighting to allow for team difference but has got the cogs turning and I'm definitely going to look into it more 😀
Very interesting indeed. I have an extremely manual model running for this and while watching I just said to myself..."this just makes like much easier for me"
Thanks Peter. I'm envisioning an Excel work book with all the league table data on the first sheet and then blank templates on subsequent worksheets that perform the calculations as the league table updates. Time to get my thinking hat on.
I love BTTS betting although I shamefully use it for accumulators. Love crossing of the sides that have scored on Saturday afternoon. Often get let down by 1 team not scoring though.
By home team overage goals ( for btts) in a game , did you mean home team scored goals +away team conceded goals ? And opposite for away team overage goals ? As explained in a previous video about overage goals . Regards Dennis
This was only because we were using a complete league season. All teams had played each other and all form was known. Ideally you would create ratings for each team, home and away and how they would interact with each other.
Hello, can you guide me on how to calculate the average number of goals scored by a team? For example, you can calculate the probability that a team scores an average of 2 goals in a match as shown in the video. Thank you.
Does this work also for over/under X goals ? I mean, for over 2,5 goals you are literally looking at top right corner above the point where both teams have a 1,5 goals scorred or there are also other spots (eg. for one team 3 goals in average and 0,75 for the other)? What about this model applying to tennis market to predict the number of games or sets in a match taking into consideration de average number of games /sets played by each player ? On a trading market, is the model usefull to make a decision upon entering with a back (if market odds are higher than the model suggests) or with a lay (if market is below the model probability / odd output) ???
Yes, it's perfectly possible to do this for over and unders. I am going to do a video on this at some point. We've done all the hard work for you on Tennis trader where you can input some stats and it will pump out all the relevant stats for you. So if you want to understand the Tennis side of things watch some of those videos and do a deep dive on Tennis trader.
Hi peter, how many match would you suggest for the average goals for the home team and away team? Average for The last 5 matchs? Average for The last 10 matchs?
You need a few reference points. Matches played are useful, but also consider the teams they played and how they played against them. So I tend to look at those sort of factors, rather than just a number.
So you've created a model, and the Betfair price is the same most of the time you would say ok the market is correct but if it is slightly out of kilter you need to do some more digging to find out why it might be out and if you can't find any reason then you may have found value, or you are missing some unknown critical information, which in that case you may be getting bad value. How do you reconcile that?
There is masses of sites now for this sort of data, but 'understat' is useful. I actually generate my own versions of xG data. But that's generally not required.
This is based on the Poisson distribution right? If so, are you confident that the formula can be used here? A poisson distribution requires that each event (goal) is independent of the events preceding it. I think most people would agree that a goal being scored in a match influences the chances of another goal being scored in a match. Or is the concept of independence here applied per match rather than per goal?
On a goal, no goal basis it would work. But as you head up the chain of modelling, Poisson would become less accurate because of it's realiance on independence. I don't think most people would really understand the signficance of fitting a distribution onto a problem however. So I'm not sure it's something I can really cover in these videos unless people have a real grasp of PDF's.
@@betangeltv Yes I totally understand that this is not the right forum to go into detail about probability questions, and the Poisson is probably a reasonable model for a close approximation. One question though. Do you think a Poisson analysis on goals for and against with correct weighting for recent games, home form, away form etc. would be enough to potentially build a profitable model that overcomes a 5% commission?
I started on Poisson about 35 years ago, so there is plenty of refinement to be had! The lack of dependancy means it often understates the draw. So that fits logically as if a team is a goal down they may go for it later on. That tends not to show up with a model like this.
Maybe I misunderstood, but did you take into account the statistics of average goals scored, which are calculated as follows: in one game the team scores four goals, in the other game the team does not score goals and on average 2 goals are scored in the game? Wouldn't it be more objective to see how many games have scored at least one or at least two goals from the last 10(or ever a different number of games) home games?
I tend to focus on the strength of the home and away teams and the number of goals they are likely to score given the scenario. This will be a blend of a number of things. Ultimately, your input into these sorts of models is where you will find an edge against others.
Pretty sure there are not many teams that average 2 goals every game away from home,City,Utd,Liverpool maybe a few others would be it,i know Chelsea didn't score many goals last season.
I did have some worked examples in the video but removed them to make the video shorter. Some teams will have certain characteristics, in terms of being slightly defensive or free scoring. So it's worth building up your model with these variations.
Hi, I just tried out predicting goals using your model(looking for over 2.5 and over 1.5). Out of 14 games 9 were over 2.5, one was over 1.5, two were 1:0 and one was 0:0. I'm wondering what goal chance percentage is the safest for over 1.5 goals?
I've got lots of data, but not done much with it. I tend to trade that just before the start of the race and I'm out before it's underway. So never really looked beyond that.
Referring to this video... how many matches cam be used as data model? I can get data from 1998 or up to the first matches of Premier League 1920, 1930, 1940..an estimate of number of games enough to give me a clue
You have to blend data to get a resonable forecast. Matches close to the current date are most likely to be the most accurate. But older data can help you form an opinion on the likely range you should be seeing. So it will always be a blend of the two.
Hmm definitely something in this along with some of the other football using maths videos, looked at the Blackburn Hull game last night both pretty bad at scoring goals, Blackburn at home and Hull away scored 5 or so each in the last 6 home and away, all indicators would lead to a nil nil which was the final score , if only id been brave enough to put my money on correct score instead of under 1.5 goals. West Ham Arsenal game today, went through the btts calculation and yes both did before half time, each team looked to have a 85% of scoring two goals each, low and behold the game has just finished 2-2. I bet both teams to score before the game btw. Get a couple of these right out of say 6 games and youre well in profit, looking at the Forest United game in half an hour because of Forests poor scoring record its not a btts game however if you deduct the home side goals from the away sides goals you are left with 3, add to the fact forest have only 48% chance of scoring using the exp formula, could United win this game 2 or 3 nil, interesting to see.
You could make a nice little spreadsheet , for all games of the season using updated stats every week. giving a nice little prediction for all games for that competition season..
It's important to have an edge in the market, so I'd recommend using a model like this but creating your own spreadsheet and variations in the model and data collection. That way your model is unique to you.
@@betangeltv Verry true, sadly my days on betfair will come to an abupt end this month as it will be no longer available in the netherlands and residents of the netherlands. after september 30th. So iam getting every thing in order to get an edge on the new dutch regulations on betting and trading.
@@betangeltv Well this is betfairs disision . as the dutch gambeling laws will become more open to other forms of gambeling. And even other betting sites are now free to aply for a licence ..
That's interesting, so Betfair could operate but choose not to? Do you have any background to this I can read. I'd happily take it up with Betfair for you.
Unfortunately it can't that simple. you need to also take into account the chances of both teams conceding also. for example Leicester are playing City next weekend, and the stats show that based on their last 40 matches Leicester score 1.79 goals at home and therefore have an 83.3% chance of scoring on average, but city are not an average defence. How would you factor this in Peter?
I'm presenting a base model here. You can expand it in many directions. The numbers work, but what inputs you put into the model can contain many variations. I would take the home goals and the away goals the away team has scored and combine them. But you would also need to understand the strengths and weaknesses of each team during that match. But the concept of taking forecast goals and using them as the base is sound.
As much as I wanted to accept this explanation of yours BTTS I can't do it. When it comes to horse trading there is no one who can challenge your knowledge but what you have just shown is simply not true.The percentage you get based on your formula is not a chance that shows the possibility of BTTS in the next game.People are constantly trying to predict a football match based on statistics ... Football doesn't work that way. Following the statistics, everyone will have a positive trade or bet. It's like when you say that in horse trade, if we do them randomly, we will have a 50/50 result. .That is the case with statistics in football.But a consistently positive result was not obtained on the basis of statistics.Statistics only show what happened before, not what will happen in the future ... For some sports and situations it may be more relevant, but in football it is simply not so.I don't want to show disrespect with this comment, but football simply doesn't work that way.
I've been analysing football for many years, much longer than other markets. What I've presented here is a way to analyse and forecast what the odds on BTTS will be. If you come up with a model that forecasts goals you can plug it into the model to see if you find value or use this model to get a feel for what the market is discounting. This and similar methods are how I've bet and traded on football for a very long time. But ultimately, you should do whatever you think gives you an edge.
Is there room to skew this? I.e for example your research produced information that Liverpool score 3 goals against bottom 6 sides, but only 1 against top 6 sides?
@BigTCG, yes you should skew it. The only reason I didn't is I am using season long data. The model will tell you how many goals are being discounted by the market, but you need to tweak that to suit if you are projecting forward. So plenty of scope for adjustments and tweaking.
For a BTTS bet, surely it would be more productive to consider the number of games each team has scored in home and away respectively, rather than the average number of goals as advocated by this model? The mean can be highly misleading, especially in games between inconsistent teams who draw a blank in many games yet have a decent average overall because they have scored freely in a reasonable number of others.
That's not how I would calculate the average goals and the ratings of each team. It would distort the inputs. The number of goals likely to be scored is in a fairly narrow range overall, so it's just tweaks and refinements to this, that I look at.
@@betangeltv hahahaha, must be one of those that wants a “silver bullet” or “a magical method to never lose “. Don’t worry chap go to exchange, put your money there we will all be waiting for you
