A YEAR OF CHANGE & DISCOVERY
As 2020 has now come to an end we want to reflect on the development of Akka throughout the year taking each month one-by-one. As an overview, our service has gone from being barebones pure AI predictions to now offering stats on 100s of leagues and a healthy prediction record to go with it. However, let’s start off with January.
We appropriately labelled January as a dry month (potentially a sign of things to come?). The reason being it was the first ever month we’d seen our AI have a negative ROI. After such a successful 2019 this was tough. Rather than letting this get us down, we assessed the situation. We realised that our service needed more than just predictions, we needed the stats.
As a result, we focused on around 5 leagues to add to our (now infamous) Stats Zone. These leagues were the top leagues across Europe with a few extras for good measure. The inclusion of these stats was done because we wanted our subscribers to have as much information as possible when making their picks. We didn’t want them to be limited by our AI predictions but rather make their own, informed predictions.
With our slow January we hoped this was not the beginning of the end of our AI. Success in February was vital, and we got it. In fact, for the English League 2 if you had followed our Over 2 predictions you would have doubled, yes doubled, your money. Pretty impressive stuff.
We also introduced 9 new leagues in February from across Europe to our Stats Zone. Once again our aim was to provide our subscribers with as much information as possible so they could make more informed picks.
Now for March, when true hardship hit. Sport was cancelled, the world was in turmoil and lockdown was no longer just something you heard in Prison Break.
Rather than licking our wounds and feeling sorry for ourselves, we decided to innovate. We did this by looking at other sports and seeing the feasibility of predicting them using an AI. From this came our UFC AI. This AI is one which very much goes under the radar but has been hugely successful since its launch. The AI has been slightly adapted since its conception in March but has been very profitable week-in week-out.
It sounds like the first AI we tried was successful but this was not the case at all. Some subscribers may remember our CS:GO AI which was an AI for eSports. This was unfortunately not consistently profitable enough. We also tested out Snooker, Darts and many other sports until we came to our UFC AI. This highlights how we strive to only deliver the best predictions to you.
April was the month where we focused more heavily on ensuring the profitability and continued success of our UFC AI. This process is very time-consuming. It is mainly the testing of different models and algorithms to see which are the most profitable, but also more importantly most consistent. The process itself can be very difficult given there are a lot of dead ends and there’s no guarantee that you can find a profitable model.
As well as that we needed to test different types of predictions, such as fight winners and how fights will end, to see which can and can’t be predicted to a good level of accuracy. All of this testing and changing, testing and changing took up all our time in April.
May was a rather strange month. Some sports had decided to come back while others had not deemed it safe enough. It was the first time we had seen football matches being played consistently without fans. This was the new norm.
In terms of Akka, this was a month of monitoring. Our greatest worry was how the new style of the game would impact predictions. The lack of crowd is always going to impact the home advantage of teams, especially teams whose stadiums are deemed a fortress such as Liverpool at Anfield.
Due to this high level of uncertainty we in fact did not charge any subscribers nor did we offer any predictions. We realised this was a vital time to determine the impact this pandemic would have on our AI and how we could minimise it as much as possible. Therefore, in May we just watched from the side-lines.
In June many seasons came to an end, especially in Europe. In turn, we were then able to assess the season-long profitability of our AI for a complete season (albeit an unusual one). The stats don’t lie and we found ourselves being very pleased with how the AI did. We actually out predicted the bookmaker’s odds on a consistent basis, while also being profitable with our Over 2 predictions once again.
We also realised we needed to step our stats game up. No longer would we have a measly 7 leagues from across Europe we wanted more. As a result we invested heavily in a new data provider so we could offer stats on over 200 leagues worldwide. We also introduced our Quick Stats tool which is now a fundamental part of our service. This tool aimed to provide a quick and easy way to find the best value markets across the world in 1 click. A busy month to say the least.
This was arguably one of our most proud moments of the year as we were able to see that our AI was beating the bookies as we had hoped. Also, the work our team put in to add over 200 leagues to our site was amazing and shouldn’t be underestimated. Sleepless nights, let’s say.
July was another month of improvements. As we had now introduced so many stats we wanted to make them as suited to our subscribers as possible. We spent the month trying to get feedback from subscribers as to what they wanted and making the required adjustments. We believe we have hit a good balance currently but we are always looking for suggestions and improvements.
In July we also mentioned our Tennis ATP AI. This AI looked extremely fruitful however due to technical restrictions it is currently dead. We aim to take a fresh look at the AI in the new year given how successful it was during the short period it ran for. This once again shows that making these AIs isn’t plain sailing unfortunately.
Over August/September we saw the return of the top European leagues. This was a true test for our AI given that teams had no pre season and the state of the game was still very different to anything we’d ever seen before.
Despite this, our AI had some good success in the English leagues being relatively profitable. The same couldn’t be said for the Dutch leagues. The beginning of the 2020/2021 season indicated to us just how difficult this season could be, with these fundamental changes in the game of football. Fortunately, our AI was still out predicting the bookmaker’s odds so we could take that positive from the situation.
October was a month of new features. During the month we listened to feedback from our subscribers and made tweaks to our Quick Stats Zone which greatly benefited our subscribers. An example of this was removing those games which had good stats but were unavailable on bookmakers sites. Also, by adding odds of the different markets for the stats it was now easier than ever to find the best value matches and FAST.
In October we once again saw how the pandemic would impact predictions specifically with our Dutch Eredivisie predictions. We found these to be inaccurate especially given in the 2019/2020 season it was the most profitable league. How quickly things can change. On the other hand, our English Championship predictions came out swinging with a 29% ROI across our Over 2 predictions. Very impressive given the circumstances.
The penultimate month of 2020 came around and so did the success of our UFC AI. Over November we were able to truly see how good the UFC AI could be especially with some special new additions to it. We found through backtesting that our UFC AI could have an ROI of around 8%. This would have been over a 2 year period which is extremely impressive. It shows the consistency of the AI and gives an insight into the potential future profitability of the AI.
This AI is still developing upon each event so we suspect it will only keep getting better. This is especially true as it seems that the new format of the UFC events without fans has not really impacted the performance of the AI. Things you love to see.
The final month of the year was pretty quiet at Akka HQ. We allowed ourselves to settle and prepare for the new year. Nothing major was worked on, nor was nothing major added to our service.
Instead, we compiled the positives and negatives from throughout the year to then integrate into this blog. We have written a few goals for the new year and hopefully achieving these will help elevate our service to the next level.
We thank you all for your continued support and love for this great project. We hope that this year hasn’t been too tough for you, as we know it certainly hasn’t been a smooth ride for us.
2021 is the year of Akka.
Akka = The name of our AI.
ROI (Return on Investment) = Profit as a percentage of total money invested.
Calculation: (Money Won – Total Money Staked)/ Total Money Staked
Over 2 = Any of our AI’s predictions where odds were greater than 2 and less than 3 (decimal odds).
Expected Accuracy = Converting bookmaker odds to an accuracy.
Calculation (based on decimal odds): 1 / Odds
Stats Zone = Part of our Football Package offering stats for 14+ European Leagues.
Backtest = Use the AI to predict past data showing the accuracy levels if used in the past.