TLDR: Essentially black box trading inevitably fails because they're all trying to do the same thing. At a certain point everyone tries to liquidate, or buy, or perform same trades at the same time which can lead to a blow up.
I wonder if getting around this assumes some sort of diversity in algorithms/ML approach, and if that diversity is a realistic assumption
Can someone put in layman's terms the passage below? please. Assuming the reader has some machine learning experience but no financial knowledge. Thanks
#### numerai_training_data.csv
Use this dataset to train your machine learning algorithm. The first fourteen
columns (`f1` - `f14`) are integer features. Column `c1` is a categorical
feature, column validation indicates a dataset that you can use to validate
your model, and target is the binary class you’re trying to predict.
If you're looking at the training dataset, you'll see these 17 columns:
"f1" - "f14" are feature columns, represented as integers. They are inputs for you model.
"c1" is also a feature column, represented as a category. Feel free to ignore it if you're not sure how to best use it.
"validation" is a binary 0 or 1 that just says: 'if you want to validate your own model, so you can test your accuracy on your own machine, you may want to use the rows that have 1's here.'
"target" is the expected output. For this training data it is provided, for the tournament data you come up with it.
18 comments
[ 3.8 ms ] story [ 59.3 ms ] threadTLDR: Essentially black box trading inevitably fails because they're all trying to do the same thing. At a certain point everyone tries to liquidate, or buy, or perform same trades at the same time which can lead to a blow up.
I wonder if getting around this assumes some sort of diversity in algorithms/ML approach, and if that diversity is a realistic assumption
#### numerai_training_data.csv Use this dataset to train your machine learning algorithm. The first fourteen columns (`f1` - `f14`) are integer features. Column `c1` is a categorical feature, column validation indicates a dataset that you can use to validate your model, and target is the binary class you’re trying to predict.
"f1" - "f14" are feature columns, represented as integers. They are inputs for you model.
"c1" is also a feature column, represented as a category. Feel free to ignore it if you're not sure how to best use it.
"validation" is a binary 0 or 1 that just says: 'if you want to validate your own model, so you can test your accuracy on your own machine, you may want to use the rows that have 1's here.'
"target" is the expected output. For this training data it is provided, for the tournament data you come up with it.
Does that help?