Quantitative research
Systematic research into financial markets: allocation and rotation models, portfolio construction, risk management and backtesting, together with the data systems and dashboards that support them.
Research, not advice. SIGMA IV does not manage client money, does not give investment advice and publishes no trading signals.
- 1DataPoint-in-time, checked
- 2SignalRules fixed in advance
- 3PortfolioWeights and limits
- 4Costs and riskFees, slippage, drawdown
- 5ValidationWalk-forward, robustness
What has been built
Nearly five years of work, from the first data pipelines to complete portfolio systems.
- Allocation and rotation models
- Rules that decide how capital is divided across assets and when it moves between them.
- Portfolio construction
- Weighting, position limits and rebalancing schedules, with the accounting kept correct at every step.
- Risk management
- Drawdown, volatility and concentration treated as part of the model rather than added afterwards.
- Backtesting and simulation
- Replaying history with explicit assumptions about timing and costs; forward simulations without hindsight.
- Statistical validation
- Out-of-sample tests and robustness checks before a result is taken seriously.
- Market data analytics
- Collecting, cleaning and aligning exchange, derivatives and macroeconomic data.
- Automated research systems
- Software that runs data handling, simulation and reporting the same way every time.
- Financial dashboards
- Interactive dashboards that show the state of a market and the reasoning behind each number.
An analytics platform for digital-asset markets
The founder built a data analytics platform that brings several research views of crypto and equity markets together. Every panel can show its inputs and assumptions.
Market cycles
Several independent views on where an asset stands in its cycle: a power-law trend, harmonic and Hilbert-transform cycle estimates, a Lomb-Scargle periodogram, Monte Carlo forward simulation and a combined z-score.
Options and derivatives
Dealer gamma exposure for Bitcoin and Ether options, aggregated across Deribit, OKX and Bybit by open interest, with the gamma flip level and the strikes where exposure concentrates.
Macro and regimes
Macroeconomic series from the Federal Reserve's FRED database, including the yield curve, liquidity and labour data, alongside a market regime classification per asset.
Equity screening and allocation
Screening of large-cap equities on AAOIFI-style financial ratios for investors who follow Islamic finance principles, with risk-parity allocation for a chosen set of names. The score is quantitative only; it is not a religious ruling.
Data from Binance, Deribit, OKX, Bybit and FRED.
Models
A model portfolio that combines an equity ETF and gold with a systematic crypto part run by two rotation algorithms.
Multi-asset allocation
Model portfolio
An equity ETF and gold exposure alongside crypto rotation. The crypto part is split equally between Algorithm 01 and Algorithm 02. Costs are charged on every trade: 0.10% of the traded amount for the ETFs and 0.25% for crypto. Traded with real money since September 2026.
- Universe
- 50% equity ETF, 40% crypto rotation, 10% gold
- Rebalancing
- Monthly, back to the target weights
Crypto rotation with a defensive asset
Algorithm 01
Holds the asset with the strongest combined trend and relative-strength score, but only while a broad macro filter is positive; otherwise it moves to the defensive asset. Position size follows a volatility target, and a margin prevents switching on small differences.
- Universe
- Seven assets, including Bitcoin, Ether and tokenised gold as the defensive holding
- Rebalancing
- Evaluated daily on confirmed closes
Crypto pairwise rotation
Algorithm 02
Compares every asset with every other asset by relative trend and holds the strongest one while a Bitcoin trend gate is on; otherwise it holds cash. It only switches when a challenger is clearly stronger than the current holding.
- Universe
- Twenty large crypto assets
- Rebalancing
- Evaluated daily on confirmed closes
Backtest results
Growth of €1, from daily closes in euros and after trading costs, compared with simply buying and holding Bitcoin, the equity ETF or gold. The model portfolio is shown from 1 May 2023, when price data for its ETF begins; the two algorithms from 1 July 2020. Data until 12 November 2026.
Historical backtest: the hypothetical result of applying fixed rules to past prices. The model portfolio has been traded with real money since September 2026, but the figures shown are the model's simulated values, not account statements. Assumed costs are 0.25% per crypto trade and 0.10% per ETF trade; taxes and other costs are not included. Rules developed with knowledge of past data usually do worse afterwards, and past or simulated returns do not predict future returns. Not investment advice.
Statistics use simple daily returns over calendar days, annualised with 365 days, with a risk-free rate of zero. Bitcoin is converted to euros. The algorithms' figures come from the model's own code and match its report. The model portfolio is recalculated from those parts with the 50/40/10 weights and monthly rebalancing; the same calculation reproduces the model's own 60/30/10 result within 0.13%.
How a result is tested
Most impressive backtests are artefacts of hindsight, missing costs or luck. These are the questions a model has to answer before its results mean anything.
- Point-in-time data
- Was every input available on the day the decision would have been made?
- Look-ahead bias
- Does any calculation use a price, revision or index membership from the future?
- Walk-forward evaluation
- Does the model keep working on periods it was not fitted to?
- Robustness
- Does the result survive small changes to parameters, start dates and universe?
- Transaction costs
- What is left after fees, spreads and slippage at a realistic turnover?
- Concentration
- How much of the result comes from a few positions or a single period?
- Volatility targeting
- Is risk kept at a stated level instead of drifting with the market?
- Dynamic allocation
- When the model moves capital, is the rule fixed in advance and applied consistently?
- Portfolio accounting
- Are cash, weights, rebalancing and compounding booked correctly at every step?
Four kinds of evidence
Every published result is labelled with one of these, from furthest to closest to real trading.
- Historical backtestRules applied to past data as if they had been followed. Prone to over-fitting.
- Forward simulationFixed rules run on new data as it arrives, without real orders.
- Model portfolioAn allocation defined by a model and tracked on paper. Not personally traded unless stated.
- Executed tradesOrders actually placed in a market, including costs.
Disclosure
Everything on this page is research. Backtests and simulations describe how rules would have behaved on historical or new data; they are not records of actual trading unless labelled as executed trades, and they do not predict future results. Nothing here is an offer, a recommendation or personal investment advice.