hatterson
Registered User
(Note, below I use the word "predicted" or things like it. I understand that the algorithm is just giving a percentage chance the home team wins based on the data, which isn't technically the same as a prediction, but I use prediction for simpler writing)
So I was curious how accurate the predictions of this algorithm were, so I decided to glance at the results so far. What we have is this:
It predicted a total of 57 games thus far (that are complete). Of those only 1 came in at 50%, I excluded it from the numbers because it doesn't seem fair to count a game that was called exactly as a coin flip one way or the other.
That leaves 56 games. Of those, 27 were picked correctly which is actually slightly less than half. Not great, but considering most predictions are only a couple percent away from coin tosses, it's well within the realm of possibility that it's just been unlucky.
Of the 56 games, the algorithm predicted that the away team would win 19 times (33.9%), and predicted that the home team would win 37 games (66.1%). In results, the away team won 24 (42.9%) and the home team won 32 games (57.1%)
Breaking it down to home vs away predictions. Of games where the home team was predicted at <50% chance to win (19 of 57 games), the algorithm was right 7 out of 19 times. Of games where the home team was predicted at >50% chance to win (37 of 57 games), the algorithm was right 20 out of 37 times.
I also wanted to see if it was clearer on the more sure predictions. I eliminated all predictions where the algorithm put the game within 10 points. ie, anything between 45% and 55% (inclusive) was excluded.
There were 11 games where the away team was predicted under these conditions, and 28 games where the home team was predicted. Of the 11 games for the away team, 6 were predicted correctly (55%). Of the 28 games for the home team, 16 were predicted correctly (57%)
So I was curious how accurate the predictions of this algorithm were, so I decided to glance at the results so far. What we have is this:
It predicted a total of 57 games thus far (that are complete). Of those only 1 came in at 50%, I excluded it from the numbers because it doesn't seem fair to count a game that was called exactly as a coin flip one way or the other.
That leaves 56 games. Of those, 27 were picked correctly which is actually slightly less than half. Not great, but considering most predictions are only a couple percent away from coin tosses, it's well within the realm of possibility that it's just been unlucky.
Of the 56 games, the algorithm predicted that the away team would win 19 times (33.9%), and predicted that the home team would win 37 games (66.1%). In results, the away team won 24 (42.9%) and the home team won 32 games (57.1%)
Breaking it down to home vs away predictions. Of games where the home team was predicted at <50% chance to win (19 of 57 games), the algorithm was right 7 out of 19 times. Of games where the home team was predicted at >50% chance to win (37 of 57 games), the algorithm was right 20 out of 37 times.
I also wanted to see if it was clearer on the more sure predictions. I eliminated all predictions where the algorithm put the game within 10 points. ie, anything between 45% and 55% (inclusive) was excluded.
There were 11 games where the away team was predicted under these conditions, and 28 games where the home team was predicted. Of the 11 games for the away team, 6 were predicted correctly (55%). Of the 28 games for the home team, 16 were predicted correctly (57%)