Time For Data

Time For Data

If you read my previous article, you will know that the pursuit for data driven betting excellence is beginning to pick up pace and with the Premier League entering Novmber it is time to look at things via more statistical means.

That’s because with enough of a form guide now being built over the first seven weeks of the season, it is time to see how my first raft of data driven football predictions will fare and hopefully my good friend “PremBot” can get the better of the bookmakers.

To do so, he is going to have to get predict 10 correct Premier League results over the course of the weekend and just one failure will blow up the hopes for this week in particular.

WHAT DO WE KNOW SO FAR

Well we know that from the first 68 Premier League fixtures of the season, my gut instinct has not been all that bad. Then again, it has not been all that great either and with 38 predictions ending up as correct (55.8%), it is time to see whether data can take the reins from here on in.

With that in mind, I am still going to test my gut instinct each week, to see just how well I perform over the course of the campaign and will also use this as another method to compare and contrast results over the next few months.

Which means, if we are once again going purely on gut instinct, here is what I think will happen over the course of this weekend:

Brighton
Southampton
Everton/Manchester United Draw
Leeds
Chelsea
West Ham
West Brom/Tottenham Draw
Leicester
Manchester City/Liverpool Draw
Arsenal

A 10 fold, which is a very tasty 5,753/1 but is that too tasty? Usually the wider the odds from say 1,000/1 the more against the grain my gut feel predictions will be.

THAT IS ALL VERY WELL

Now gut instinct is fine, but this is what we have been building towards and in a moment “PremBot” will reveal his first raft of data driven football predictions for the season. Before that happens though here is the first form guide of the season in infograph form:

To those untrained readers the central colour code is as follows:

Orange – Hottest Form In The Division
Green – Best Form For That Game
Blue – Coldest Form In The Division

While the six-game run of form will then be indexed and depending on the combination of results, will also have some bearing on how “PremBot” works as a predictive model and this season, there will be something additional in play.

This time we are going to have constant weeks of model testing, not just chopping and changing throughhout and the first two models that should both be in play by week 13 or 14 of the season is these two

PremBot v1 – Form Taken By Last Six Results Home Or Away
PremBot v2 – Form Taken As Last Six Home Compared To Form Of Last Six Away

By doing this, we can get a better measure of what works the best and once we get to learn how these perform, we can also get a better scope of further variants of models and throw even more into the mix (Such as far more weight on away advantage instead of home for example)

COME ON, ENOUGH IS ENOUGH

Ok, the groundwork has finally been laid over the first seven weeks of the season and now it is time for “PremBots” first data driven football predictions for the Premier League, something that looks as follows:

If these 10 get over the line, it will be odds of 4,755/1 and although they look rather hefty if there is one thing that we have learned it is that the bookmakers do not get right themselves and to get a better idea of context, here is what the pre-match odds are saying:

Today, 06 Nov			1	X	2	B’s
16:30	Brighton – Burnley		1.89	3.49	4.57	13
19:00	Southampton – Newcastle		2.03	3.61	3.76	13
Tomorrow, 07 Nov			1	X	2	B’s
11:30	Everton – Manchester Utd		2.94	3.57	2.43	13
14:00	Crystal Palace – Leeds		2.94	3.41	2.49	13
16:30	Chelsea – Sheffield Utd		1.40	4.96	8.34	13
19:00	West Ham – Fulham		1.84	3.89	4.30	13
08 Nov 2020			1	X	2	B’s
11:00	West Brom – Tottenham		6.42	4.42	1.53	13
13:00	Leicester – Wolves		2.39	3.19	3.31	13
15:30	Manchester City – Liverpool		1.99	4.11	3.50	13
18:15	Arsenal – Aston Villa		1.71	4.04	5.04	13

When looking at my predictions compared to who the bookmakers think will win, mine are in line with 6 out of 10 and there is differential at the following games:

Everton vs Manchester United
Crystal Palace vs Leeds
Manchester City vs Liverpool
Arsenal vs Aston Villa

Therefore, it is these games that I will need my model to outwit the bookmakers and if the wins match, then it will be a stirling debut performance from my predictive model.

NEXT STEPS

In a couple of weeks, I will want to apply an element of double chance to a sub-model and for example you could back the “big six” to win where they all play in individual fixtures and add more ballast to the swing matches that could go either way.

Then again, if there is one thing we have learned thus far that everything seems to be going either way at present and it may just be better to stay with the all-out attack on the bookmakers and hope for an outrageous 10-fold win.

WHAT’S THE CODE

While for the really technical, here is the code that has allowed “PremBot” to serve up some data driven football predictions:

=IF(AND(S3-C3>=-5,AE3>50),V3,IF(OR(S3-C3<=-5,AE3>100),Y3,”Draw”))

Which means if the away team is five positons or less below the home team and their form index ranking is 50 worse off, then “PremBot” says it is a home win, while if the away team is 5 positions clear of the home team or it has a better form index by more than 100, then it is an away win.

While if neither of these criteria are met, the game will subsequently end in a draw.

The league positions and the form ranking could be tweaked for future reference or different model version and as always, who knows if this is the right combination of factors – then again, that is all part of the “fun”

Happy punting and thanks for reading.

(THESE ARE NOT TIPS PER SE, THESE ARE JUST DATA DRIVEN FOOTBALL PREDICTIONS IN A PURE TEST AND LEARN MODELLING CAPACITY)