• Alastair Beaumont
• Kolby Chapman
• Graeme Nathan
• Cole Peterson

• Home team (str)
• Away team (str)
• Home team's score (int >= 0)
• Away team's score (int >= 0)
• Match winner (str):
  • H: home team
  • A: away team
  • D: draw
• Date of match (date - dd/mm/yy)

• Relegation score (float in (0, 1))

This is calculated as 1 - (pts/team_max). It is a ratio in the range (0, 1) where the higher a team's score the closer they are to being relegated.

The team_max is 3 times the number of games that time has played, representing the theoretical maximum score the team could have at that point in time.

We will create a baseline algorithm to use as a baseline comparison against a machine-learning approach.

We will implement prediction algorithm that always predicts the outcome of the upcoming match as the mode of the previous X matches of each team. This will serve as a comparision, along with the theortical randomized guessing rate of 33%, for our machine learning implementations.

If both teams in a match are predicted to have the same outcome as one another, then this will be considered a prediction of a draw.

We will take two similar approaches in this area.

First, we will use our data without the relegation score. Second, we will repeat the process with the relegation score. In both cases, we will use one season as training data and another season as testing data.

We will investigate whether or not a machine learning approach is any better than randomly guessing the outcome. Additionally, we are going to check whether or not a team's closeness to relegation has an impact on their playing. Should we find that such an affect does appear to occur, we will investigate the threshold where the affect starts.

• Build baseline prediction and calculate its accuracy for different lengths of history set
• Redefine our closeness to relegations stat to include the theoretical remaining points a team could achieve so we can measure the closeness to relegation more accurately.
  • thinking is that, if you only have a few more chances to make up enough points to leave the relegation set, then you are closer to relegation.
• Define what measures to use when checking if a team is playing 'better'; is it simply number of goals scored?