NQ Futures Quant Suite
Proving what a Nasdaq-only system can't do, then shipping what it can
A 30-year research programme on systematically trading the Nasdaq-100. The brief asked for 30%+ annual returns from a single instrument. The interesting result is the proof that it doesn't exist — and the strategy that does, published with its costs, its caveats and its failure modes attached.
The problem
The brief was 30%+ CAGR trading the Nasdaq alone, with drawdowns kept small. Most backtests answering a brief like that quietly flatter themselves: they ignore financing on borrowed money, assume free execution, peek at data they wouldn't have had, or tune parameters until one lucky configuration looks brilliant. The real question was not how good a number I could produce, but which numbers survive honest accounting.
What I built
The shipped configuration returns 16.1% annually against a 32.9% maximum drawdown over 30 years, with the out-of-sample period scoring better than the in-sample — the opposite of an overfitting signature. Buy-and-hold over the same window returns 13.2% but with an 82.9% drawdown, so the edge is risk containment rather than raw return. Two harder claims were proven false: 30%+ CAGR on this instrument is unreachable at the available Sharpe without ruinous volatility, and 25% CAGR inside a 20% drawdown budget is impossible — the honest frontier maximum is 12.2%. The strategy is also ported to TradingView Pine Script and published open-source alongside the research.
How it works
The strategy counts how many of the 63, 126 and 252-day returns are positive and maps that vote to a target exposure, gates leverage on the 200-day moving average, scales position size to hit a 30% volatility target, halves the leverage cap whenever equity sits more than 10% below its high-water mark, and only trades when leverage drifts far enough to be worth the cost. Every signal is lagged a full day. Financing is charged at the T-bill rate on borrowed notional and idle margin earns T-bills, so leverage costs what leverage actually costs. The configuration was then split in and out of sample, walked forward, and pushed through a Monte Carlo to find how often the drawdown budget breaks.
Highlights
- 16.1% CAGR / −32.9% max drawdown over 30 years, financing and costs modelled
- Out-of-sample Sharpe 0.75 beats in-sample 0.66 — no overfitting signature
- Proved 30%+ CAGR on a single index is unreachable, rather than claiming it
- Proved 25% CAGR under a −20% drawdown budget impossible; real frontier max is 12.2%
- Rejected after testing: shorting, leverage caps above 5x, indicator stacks, fixed R:R stops
- Ported to TradingView Pine Script, with every deviation from the Python documented
Let's build yours.
Tell me what you're trying to ship — you'll get a scoped plan and a straight answer.