Quote:
Originally Posted by Macindoc
I was trying to prove that an average error of 10.4 is not good, only slightly better than picking random numbers. All of those inputs in the JFresh model make it no better than some simple model that has nothing at all to do with the sport or the league. So there's something wrong with it. If the JFresh model is any good, it should be measurably better, regardless of how that result was reached. If you're inputting a whole bunch of information that you think is relevant but your results are no more accurate, then maybe what you think is relevant isn't, or isn't weighted properly. It's not the concept of the JFresh model that's the problem, it's the model's design.
|
Then you failed. You didn’t even pick random numbers, you picked specific ones using a set formula where you already knew the result you were trying to accurately predict… and his was still measurably better.
Predicting 1312 events with a 88.6% average accuracy is pretty good or at least interesting.
Predicting 32 datas points you already actually know and only hitting 88.6% accuracy is actually kind of terrible. “I put it on easy mode for myself and knew the answers and was almost as accurate as the complex model that didn’t!” isn’t really a flex.
But as I said, I think you’re just missing the point here, which is fine, though I think my math teachers would have spit on me if I told them how you reach the result and whether you can do so consistently doesn’t matter, all that matters is getting pretty close if you already know the answer you’re trying to find lol.