BA (Mathematical) Finance student at University of Zurich · concurrent mathematics at ETH Zürich · Zurich, Switzerland
One question through most of this: how do you tell a real edge from a lucky backtest?
Strategy-Zoo — 780 multi-asset configurations, enumerated and committed before any of them ran, then corrected for having tried all 780. Best Sharpe 0.67, deflated 0.639. Zero survive Bonferroni, Holm or BHY at 5%.
Volatility Risk Premium — Does the market overpay for insuring against turbulence? Confirmed on 84–86% of trading days across three measurement windows. A term-structure de-risking rule cut losses in the two real crashes tested (Feb 2018, Mar 2020) by roughly 70–76%, at no cost to everyday profit.
Equity Factor Timing — Can the value spread time HML? Four pre-registered rules on Kenneth French data 1983–2026, block-bootstrap CIs. One looked significant until it was corrected for the other three. Null.
Monte Carlo Option Pricing — European, American (Longstaff–Schwartz) and Asian options by Monte Carlo, with pathwise Greeks and variance reduction.
Sporitelna Deposit Forecasting — Retail loan and deposit forecasting for Slovak households, built in R at Slovenská sporiteľňa (Erste Group) and used internally for scenario planning.
The trial count gets written down before the first backtest runs, confidence intervals are block-bootstrapped, and null results stay published.
Python · R · LaTeX · SQL valachy.mar@gmail.com · LinkedIn