September Review

This blog is going to cover the performance of Akka when predicting Esports (Counter Strike: Global Offensive) for the month of September. Some interesting insights have been gained and applied to the AI to allow for future improvements.

Successes

Finding Market Value

Our hit rate for predicting Under 2.5 maps was 68%, giving implied odds of 1.47. This is much lower than the average odds for this market.

ESL New York

Akka predicted 7 games during ESL New York and correctly predicted 6 of those.

Kaizen

We have made notable changes to the AI from 20th September onwards and this can be seen with higher accuracies and a better ROI.

Improvements

Regular Updates

We are aware that sometimes Akka is not able to predict every game in advance. Due to time restrictions, etc. We aim to improve this so Akka is run multiple times per day.

Best of 1 Predictions

Akka is currently restriced to predicting Best of 3 games. However, we expect to add best of 1 games in the upcoming months.

Kaizen pt 2.

We aim to keep reviewing Akka to determine where it is failing and from this make appropriate changes. The aim is to keep pushing the accuracy of Akka as far as possible.

ROI Overall (Moneyline)

-22%

Implied Probability vs Hit Rate (Moneyline)

62.5% vs 55%

Hit Rate (Over/Under 2.5 Maps)

63%

AKKA STATS
(PRE-MODEL CHANGES)

ROI Overall (Moneyline)

-1.4%

Implied Probability vs Hit Rate (Moneyline)

62% vs 66%

Hit Rate (Over/Under 2.5 Maps)

63%

AKKA STATS
(POST-MODEL CHANGES)