S-TIER technique
Wasserstein Distance — measuring the backtest-to-live reality gap
100k ft — What is this?
Wasserstein Distance — measuring the backtest-to-live reality gap
50k ft — Why care?
Uses Wasserstein Distance and Sinkhorn Divergence to mathematically measure drift between backtest distributions (blue) and live market conditions (orange). Detects distribution shift that kills more strategies than bad fees.
10k ft — Key takeaways
• High backtest Sharpe doesn't guarantee live performance
• Wasserstein distance quantifies strategy drift
• Monitor distribution shift to kill dead strategies early
5k ft — Tools & techniques
Python
Distribution Shift Detection, Optimal Transport, Sinkhorn Divergence, Wasserstein Distance, algorithmic trading
1k ft — Deep-dive / try it
The most painful chart in trading. 📉
Your backtest has a Sharpe of 3.0, but your live account is bleeding? That’s called Distribution Shift, and it kills more strategies than bad fees ever will.
In this visualization, I’m using Wasserstein Distance and Sinkhorn Divergence to mathematically measure the "Reality Gap" between historical simulations (Blue) and live market conditions (Orange). If these distributions drift too far apart, your model is broken.
Stop guessing if your strategy is overfitting. Measure it.
👇 Comment "Wasserstein" below and I’ll DM you the Python repo to build this yourself.
#quantfinance #algorithmictrading #econophysics #python #math
quant.dhawan
20w
👇 Comment "Wasserstein" below and I’ll DM you the Python repo to test this yourself
Transcript excerpt
(no speech detected)
Appears in
- Guide: Physics Models → Market Alpha: When Scientists Become Traders
- Deeper dive: Physics Models → Market Alpha: When Scientists Become Traders
- Guide: Strategy Autoresearch — the judge is the product, not the agent
- Deeper dive: Strategy Autoresearch — the judge is the product, not the agent
- Guide: Own the Thermometer — $100 sensors as prediction-market edge
- Deeper dive: Own the Thermometer — $100 sensors as prediction-market edge
More from @quant.dhawan
- B-TIER Heston Model — stochastic volatility for realistic options pricing
- B-TIER Cox Process — doubly stochastic Poisson for bursty market events
- A-TIER Hidden Markov Models — detecting market regime shifts automatically
- B-TIER Avellaneda-Stoikov — optimal spread calculation for market makers
- A-TIER Ornstein-Uhlenbeck Process — mean reversion for pairs trading