from Amazon
Tuesday, September 08, 2009
Are forecasting systems useless?
This is the third of my three-part experiment. In this case, I did the following:
- I created 21 teams, semi-intelligently, and put them in a league. I did this by taking the consensus picks, turning them into dollar values, then randomly giving each player a dollar value within 5 dollars of that, and then sorting to create my player list. I did this 21 times.
- I put in Marcel in this league to make it 22 teams. I made Marcel draft first.
- Then, I replaced Marcel with another forecasting engine, and made that guy draft first (facing the same 21 teams as Marcel).
- I did this with all 22 forecasters
- I then created yet another league of 21 new random teams, and made Marcel pick SECOND
- And so on and so on
So, I ended up with 484 leagues in all, with each forecaster (and “amateur") playing in 22 leagues. With 484 wins available, the pro forecasters should win on average 22 times in total, if they were “average” compared to the semi-intelligent picks. They won only 12 times. Once again, teams 122 and 115 combined for 8 wins, and the other 20 pro forecasters managed just 4 wins.
IN CONTRAST, the semi-intelligent “Amateur 444” won 21 out of 22 leagues. “Amateur 565” also won 21 out of 22 leagues. “Amateur 762” won 19 leagues. There was three more with 17, and 3 more with 15, and on an on. 363 of the 462 amateurs won zero times.
The amateurs won at a greater proportions to their number of entries than the pro forecasters, and it was not even close.
Recent comments
Older comments
Page 1 of 343 pages 1 2 3 > Last »Complete Archive – By Category
Complete Archive – By Date