Great effort friends.. Really good to know that the predictions so far have been close (one would not expect any player to score exactly as u predict; though inzy and lara did score the exact number of runs predicted) and at least you have been accurate in predicting the winner..
But just to understand how logical your modeling is, I would like to ask a few questions. 1. Since we know that different players may perform differently against different opponents and also in different conditions, it would make sense to give more weight to the performances of these players earlier in similar conditions. But how have u arrived at these weights? Is it something the model decides on its own or is it tweaked externally? 2. How have u gone about validating the models that you have built? 3. Why is it that some players have a range of scores while others are assigned just one particular number? 4. Also, considering this current prediction that you have made for Ind vs Bangladesh match, did u consider the fact that India has not played Bangladesh in I think the last two years? So does it not mean that your predictions could be more erroneous? (It is a different issue whether India wins comfortably or not, but how reliable is the estimate in these circumstances?) 5. Finally, though I know that the effort has a fun component attached and so it makes sense to keep it simple and free from technicalities; for the more obsessed souls like me, could you provide some statistical measures related to your model like say the level of confidence or goodness of fit of the model etc?
Hats off to all guys of fractal for this... whatever be the results u guys have at least created buzz about this industry of predictive modeling :P ....... sabby i think u r done with ur interviews thats y asked so many logical questions here :D .....or is it that u r also trying to bring modelytics in news :)...... One more question ..... is it possible for your model to predict team total also coz i think if u are predicting scores of individuals than u can predict team total also neways guys congos Keep it up
@nbdu (abhishek) yes my interviews are over and so i have more time at my hands now.. :D and about bringing modelytics in the news, woh bhi ho jayega with time.. :P
about predicting the team scores, it is more difficult because the scores cannot be predicted for the newer scores (lack of adequate data) and also, although the prediction turned out to be true for dhoni, i feel that it is more a kind of type III error.. :D what say analyst junta?
7 comments:
Great effort friends.. Really good to know that the predictions so far have been close (one would not expect any player to score exactly as u predict; though inzy and lara did score the exact number of runs predicted) and at least you have been accurate in predicting the winner..
But just to understand how logical your modeling is, I would like to ask a few questions.
1. Since we know that different players may perform differently against different opponents and also in different conditions, it would make sense to give more weight to the performances of these players earlier in similar conditions. But how have u arrived at these weights? Is it something the model decides on its own or is it tweaked externally?
2. How have u gone about validating the models that you have built?
3. Why is it that some players have a range of scores while others are assigned just one particular number?
4. Also, considering this current prediction that you have made for Ind vs Bangladesh match, did u consider the fact that India has not played Bangladesh in I think the last two years? So does it not mean that your predictions could be more erroneous? (It is a different issue whether India wins comfortably or not, but how reliable is the estimate in these circumstances?)
5. Finally, though I know that the effort has a fun component attached and so it makes sense to keep it simple and free from technicalities; for the more obsessed souls like me, could you provide some statistical measures related to your model like say the level of confidence or goodness of fit of the model etc?
All in all, a brilliant effort. Keep it up!
Hats off to all guys of fractal for this... whatever be the results u guys have at least created buzz about this industry of predictive modeling :P .......
sabby i think u r done with ur interviews thats y asked so many logical questions here :D .....or is it that u r also trying to bring modelytics in news :)......
One more question ..... is it possible for your model to predict team total also coz i think if u are predicting scores of individuals than u can predict team total also
neways guys congos
Keep it up
Things not looking good today
Well any prediction for India should consider the fact that Indians play right at the level of the opposition. team...
@nbdu (abhishek)
yes my interviews are over and so i have more time at my hands now.. :D
and about bringing modelytics in the news, woh bhi ho jayega with time.. :P
about predicting the team scores, it is more difficult because the scores cannot be predicted for the newer scores (lack of adequate data)
and also, although the prediction turned out to be true for dhoni, i feel that it is more a kind of type III error.. :D what say analyst junta?
Only The Great Indian Team with their Skills, Determination & Committment could get the Predictive Model wrong!!!
Not at all close... Hats off to India !!
Post a Comment