The System

System structure

• 1. data pipeline
  • information from the book; trades
  • quantitative game and player data relavent to sport association
  • qualitative news sentiment data
• 2. runs models Models folder |_ example_news_sentament_model.py |_ example_arbitrage_model.py
• 3. finds buy-sell signals
  • normalized data to translate into execution
• 4. executes trades
  • logs into broker w/ selenium
• 5. PnL data
  • shows results with graphs and helpful multiples

the strategy might be composed of two general trading systems:

• arbitrage; making the market more efficient
• news sentiment; based on highly relavent and ripe data

news sentiment

• classification method

brainstorming new sentiment:

• what are the possible new items that have high importance
• is there an objective way of determining that
• should players have a preformance rating to determine what news is important and how

brainstorming ai:

• how can we use ai to find buy-sell signals
• is there a way of formatting data in a clean way that the neural network can understand
• is there enough data to train an ai model
• how long would it take to run and would it be competetive
• can we use a general/trained LLM model or might it be better to only use quantitative terms
  • can both be used but for different scenarios; news sentiment & regular buy-sell model

Weird set-up

To fix the MRO error, you need to patch the pytorch‑forecasting source code that defines the combined callback. In your environment, locate the file:

vbnet Copy venv\lib\site-packages\pytorch_forecasting\models\temporal_fusion_transformer\tuning.py Then find the class definition that currently looks like this:

python Copy class PyTorchLightningPruningCallbackAdjusted(pl.Callback, PyTorchLightningPruningCallback): ... Change the inheritance order so that the Optuna integration callback comes first:

python Copy class PyTorchLightningPruningCallbackAdjusted(PyTorchLightningPruningCallback, pl.Callback): ... Save the file and re-run your script. This swap resolves the inconsistent method resolution order error by ensuring that the MRO is defined in a consistent way.