docs/MiniPaper.md

Vector error correction model information leadership analysis across 20 equities and 20 ETFs:

• plots_and_figures/vecm_leadership_data.csv

Runtime comparison for fitting implied risk-neutral density using Gaussian process regression with fresh starts vs incremental low-rank updates:

• plots_and_figures/example_RND_incremental_nvda/incremental_comparison.csv

Risk-neutral densities across all 40 underlyings at 10:00 AM on November 5th, 2024:

• plots_and_figures/example_RND_10am_1105/

Implied volatility surfaces - smooth market-implied vol, unsmoothed, and double-exponential Kou jump diffusion fits:

• plots_and_figures/iv_and_kou_jump_diffusion_fits/

Kalshi vs Fama factors - comparisons of Kalshi-derived electoral probability with intraday computed Fama factor returns:

• plots_and_figures/kalshi_and_fama/

Kalman Filter & Logit Transformation to denoise raw presidential prices and map to a martingale:

• plots_and_figures/kalshi_kalman_filter.pdf
• For context, see docs/Greg_WkngIdeas_on_Binary_Martingales.md
• Reference: Election Predictions as Martingales: An Arbitrage Approach by Nassim Nicholas Taleb (NYU, October 2017)

Regime decomposition - moment matching to get lognormal superimposition on Tesla risk-neutral density with binary outcome of election:

• plots_and_figures/TSLA_regime_decomp_1500.pdf

VECM leadership analysis - examples of leadership plot using vector error correction model across all 40 assets, lag 1 on a minute basis:

• plots_and_figures/vecm_leadership_analysis.pdf

See section 4 below for environment setup.

Main dashboard files (serve live Panel dashboards):

• graphing/Calculating_Binary_RiskNeutralDensities.py - 📹 Demo Video
• graphing/Leadership_and_FactorModel.py - 📹 Demo Video

Demonstration dashboard - calculates Fama factors intraday as well as microstructure features:

• graphing/Demonstrate_Factors_and_Microstructure_Features.py

Additional test files:

• graphing/test_fama_and_microstructure_features.py

Supporting libraries:

Risk-neutral density extraction:

• rnd_extraction/

CBRA algorithm (Conditional Block Rearrangement Algorithm):

• mv_rnd/
• Reference: "A model-free approach to multivariate option pricing" by Carole Bernard, Oleg Bondarenko, Steven Vanduffel (Accepted: October 2020, Springer)

Fama-French factors - connecting to Wharton's WRDS to calculate Fama factors:

• fama/

Information leadership research (not completed due to time constraints) - historical ML pipeline to predict one-lag-ahead information leader using microstructure features up until the observed time.

Requirements:

• Python (3.12+, lower could work too, only tested at 3.12)
• Install: pip install -e . (uv is faster, if you have it)
• python download_data.py (will extract data.zip into root directory of the repo),
• then run the Main dashboard files above from the root directory (don't cd into graphing/)