For funds and traders with a mandate
I use a self-built, AI-native research framework to independently research, develop and validate systematic trading strategies against your investment mandate. Bring me your target returns, risk limits and markets. Within a week, I'll research and test strategy candidates through my framework and deliver a clear assessment of what holds up, what doesn't, and what deserves further consideration.
2 · Case studies
Case studies against published fund mandates, not past clients. Each one takes a brief that a fund or platform has made public, builds a strategy for it in about a week, and tests it the same way as the strategy log: rules locked in advance, data after 30 Sep 2025 kept for testing only, double trading costs and one-day-late trades.
Results are backtests, not live track records. Expand a card for a short summary, or open the one-page PDF.
Mandate: little stock-market exposure, hold up in crashes (early 2020, 2022), beat T-bills + 2% after costs. Result: up in Q1 2020 and 2022; beat cash in 19 of 19 years; beat T-bills + 2% by 4.0%/yr before Oct 2025 and by 0.7% in the test year.
Note: evidence is moderate after multiple-testing adjustment.
Full report (PDF)Mandate: hard stop −8.5%, target worst fall 3–5% including COVID, strong Sharpe, positive every year. At 2× leverage the worst fall over 2007 to Sep 2025 is −6.0%, and the −8.5% stop is never hit.
Caveat: Sharpe above cash (1.23) is under the 1.5 bar; a bad year could still breach a 5% loss limit.
Full report (PDF)Mandate: market-neutral pod ~20%/yr net, Sharpe above 1.5 (less if diversifying), little crypto-beta. Test period return matched the return target with almost no bitcoin link while bitcoin fell 30%.
Why further research: Sharpe under 1.5, a deep drawdown from 2022 to January 2024 (about −33%), and multiple-testing leaves luck on the table.
Full report (PDF)Mandate: daily ranking of ~300 coins scored on next-~20-day returns; pays for originality. Re-run on the full perpetual-futures universe, delisted coins included, with point-in-time liquidity rules: rank IC 0.098 (Jun 2020 – Sep 2025), in line with 0.095 on the original 85-coin list and well above the typical 0.01–0.03. Using the signal a day late makes it slightly weaker, as it should.
Why further research: almost all strength is the well-known “calm beats jumpy” pattern; the futures-payment part adds only 0.016, so originality is not proven.
Full report (PDF)Mandate: 8+ months live history on a trader-seeding platform and, for GOLD, roughly 1-year return above 20% with return / worst-fall above 2.5 at the platform’s fixed risk level. Test year clears the bar (+24.6%, ratio 4.2; +23.0% / 3.5 under double costs and late trades).
Why further research: needs up to ~3× leverage (2× cap drops return to +16.3%), and would have qualified in only 6 of 10 past years.
Full report (PDF)1 · Research framework
Nine steps from hypothesis to conditional live deployment. Tap a step for detail.
Ideas originate from newsletters, books, academic and industry research, exploratory data analysis, and TQG-assisted idea generation.
Define data source, universe, sample period, information availability, and a frozen out-of-sample split before testing. Search stays before the study-specific cut (often 2025-01-01).
Implement and backtest with safeguards aimed at mitigating lookahead and survivorship bias; their effectiveness is validated rather than assumed perfect.
Seek a stable high-Sharpe plateau, not a single best point; choose a representative/median configuration or an ensemble using development data only.
Fees, slippage, and a 2× cost stress on OOS. When a trial matrix exists: DSR, PBO / CSCV, shift nulls, and beat-random-search checks. Each row’s Robustness Score (0–10) ranks out-of-sample delivery — a comparison hint in this log, not a forecast of future profits.
Publish qualifying research tracks and archive rejected or no-edge hypotheses with transparent decisions. Every hypothesis receives a verdict.
Monte Carlo simulations and stress scenarios estimate plausible variability, drawdown ranges, and failure conditions for candidate systems.
Paper trade and monitor operational behavior, realized signal generation, and divergence from research assumptions before any capital commitment.
A conditional future stage under explicit capital, operational, and risk controls. This site does not imply live trading where it has not occurred.
Experience
Next
Send your mandate, or a short description of what you are looking for, and I will reply with scope and the fixed fee for a one-week trial. Also open to fractional research engagements and collaborations.