A Marked Improvement
A Marked Improvement
If you read my previous article, you will be aware that the pursuit of data driven betting excellence continued last week and even though the Premier League programme was one leg shorter than normal, that did not stop “PremBot” and hist younger brother from trying their luck once more.
Which means with another weekend of top flight action quickly appearing on the horizon, it is time to see how the mixture of gut instinct and data driven football predictions fared in the weekend that has just passed.
HOW DID WE DO?
As always lets take a look at the gut instinct element first and after an incredibly poor performance the week before, one hoped that there would be some form of marked improvement and now its time to find out if that was the case, by running through the nine-fold attempt.
Everton – N
Manchester City – Y
Manchester United – Y
Chelsea – Y
West Brom/Crystal Palace Draw – N
Leicester – Y
Tottenham – Y
Liverpool – Y
Brighton/Southampton Draw – N
A respectable six out of nine, but if we’re honest these were not the bravest picks and once again it is the ‘swing’ games not involving the top teams which is where my gut was not made of cast iron.
However, that will at least boost the percentage total and from 107 predictions, 56 have been correct. Giving me a success rate of 52.3% thus far and sneaking just above average, but to be honest we should be working much nearer 60% to have any real ‘success’
Still, just random picks is not really part of the project and therefore, it is now time to look at how both the data driven predictive models fared.
ALL ABOUT THE DATA
First up lets recap what “PremBot” picked out last weekend:
A respectable 5 out of 9, so just above average in terms of overall performance
While lets take a look at how its younger brother “PremBot Jr” fared:
A relatively decent 6 out of 9, so the second iteration of the model was the winner this week – with Crystal Palace’s differential proving to be the edge here. Although in hindsight that Liverpool/Wolves draw prediction was wide of the mark, but I have to play the cards as they are dealt.
How does that mean, when we compare all the predictive attempts:
Pure Gut Instinct: 56/107 (52.33%)
PremBot 1: 19/39 (48.71%)
PremBot 2: 15/29 (51.72%)
Hybrid Model (Combining First 68 matches and then PremBot only): 57/107 (53.27%)
Ignoring the hybrid model, as that’s only a measure to mop up every single game throughout the season. The gut instinct is currently winning in percentage terms, but there is a much bigger sample at play here.
So we can attest that its not an absolute shambles, however there is still work to be done and it is work that will be carried out before another round of Premier League data driven football predictions.
Happy punting and thanks for reading – to discuss this or any other projects you have, feel free to message me at either @dantracey1983 on Twitter or email me: dan@realfootballman.com
(THESE ARE NOT TIPS PER SE, THESE ARE JUST DATA DRIVEN FOOTBALL PREDICTIONS IN A PURE TEST AND LEARN MODELLING CAPACITY)