Profit Is Profit
Profit Is Profit
If you read the previous article in the series, you will know that last weekend was the first live testing of my predictive model, one that I’ve decided I’m going to call ‘StatBot’ and with the most recent raft of Russian Premier League fixtures behind us, it is time to reflect on its performance.
Of the eight fixtures that the model predicted, StatBot managed to get five of them correct and the matches in bold italics are the correct picks (with my pick/actual outcome)
Krylya Sovetov Samara vs FC Tambov (Draw/Krylya Sovetov Samara)
FC Rostov vs Dinamo Moscow (FC Rostov/FC Rostov)
Lokomotiv Moscow vs Zenit (Draw/Lokomotiv Moscow)
Ural vs CSKA Moscow (CSKA Moscow/CSKA Moscow)
Rubin Kazan vs FC Ufa (Draw/Draw)
Spartak Moscow vs FC Orenburg (Spartak Moscow/FC Orenburg)
FC Krasnodar vs Arsenal Tula (FC Krasnodar/FC Krasnodar)
PFC Sochi vs FK Akhmat (PFC Sochi/PFC Sochi)
Now the target to break even each week is for four of the eight picks to be correct and therefore with five correct, it means we are off the mark with a profit. The 8 unit expenditure of £4, has returned £5.32. Which means the ROI in week 1 of the project is 33%.
Now admittedly we are working with small numbers and therefore our profit is going to be rather inflated in percentage terms. However, it has at least shown that there is something to build on in terms of StatBot.
At the same time, with three picks being wide of the mark, I need to ask myself what factors have stopped a clean sweep from being the case and here is what we can identify from the off:
- Perhaps the coding is too wide for matches that are predicted for a draw (as this where two of the three incorrect picks fell down) – something that can be amended with trial and error.
- Perhaps I need to build a form guide that looks at the home team’s last six home matches and compare against the away team’s last six away matches and use that in a form index
- At what point does this all boil down to luck and the fact that there are going to be shocks along the way, something that model cannot necessarily factor in
- Should I be building a model to find the edge or should I be building one that mimics what the bookmakers say in terms of odds.
- Also should I be moving away from singles and picking the three most likely picks for a larger treble – because single betting is not going to make me rich in the long term,
So here we have a number of action points or ideas that can be looked at over the next few weeks and of course there are still a number of additional layers we can add to the stats base behind it, as form itself as the key indicator might leave us a bit hamstrung.
However, with five of the eight picks correct and the fact that we’re in profit in week 1, means there is not a great deal that needs to be ripped to pieces at this early stage and with us being just a round of fixtures before the international break, there is little need to make any changes.
Therefore, by leaving things unchanged in terms of build, the next step is to test whether StatBot was just the welcome recipient of some beginners luck or whether or not, we do have the shell of something viable. Only time will tell.
Happy punting and thanks for reading, Dan.
If this has grabbed your interest and you would like to discuss/feedback then please feel free to drop me a message at firstname.lastname@example.org. While I am always looking for new football/data projects to work on and if you feel that my skills would be of use, I can be contacted at the same address.