The Engine Room: How Systematic Strategies Are Built, Backtested, and Deployed

Behind every algorithmic trading decision is a process. Not a guess, not intuition, not an idea shouted across a trading floor. A disciplined, repeatable process that moves from hypothesis to tested algorithm to live capital.

Most investors never see inside this process. They see monthly statements and performance reports. Understanding how the engine works isn't about running backtests yourself—it's about distinguishing managers who've genuinely stress-tested their systems from those who've optimized for historical data and hope it persists.

It Starts With a Hypothesis

A systematic strategy begins with an observation. Academic literature documents market anomalies with theoretical justification—earnings surprises predict momentum, value outperforms growth over long periods, quality companies deliver more stable returns. Observable market patterns generate hypotheses through direct data analysis. Macro relationships suggest additional signals—rising yields associated with widening credit spreads, for instance.

AI and machine learning models contribute as signal generators, processing decades of price, volume, and macro data to identify patterns human analysts might miss. But these machine-generated signals feed the hypothesis pipeline the same way academic research does—they don't override human judgment on position sizing or risk constraints.

The best hypotheses share three characteristics: they're grounded in evidence, they have causal logic (not just "A correlates with B" but a mechanism explaining why), and they're testable with specific criteria for confirmation or rejection.

Backtesting With Discipline

This is where rigor separates serious managers from the rest.

Data quality comes first and it's deceptively difficult. Corporate actions distort price history. Survivorship bias means delisted companies vanish from datasets. Dividends must be properly adjusted. The gap between clean data and dirty data can be several hundred basis points in model performance. A manager casually assembling price data without careful quality assurance is optimizing against a fiction.

Time periods must span multiple market regimes—bull and bear markets, rising and falling rates, high and low volatility. A strategy tested only on the pre-covid, post GFC bull market of 2012-2020 will fall apart in extreme volatility of 2020 and the bear market of 2022. Credible backtests cover enough history and cycles to capture multiple drawdowns and regime changes.

Overfitting prevention is the critical juncture. The researcher has dozens of parameters to tweak and can run millions of combinations. One will look incredible against historical data. Then market conditions shift and it collapses. The defenses: use fewer parameters (simplicity reduces overfitting risk), test rigorously out-of-sample (never validate on data you used to build), use walk-forward testing (optimize on rolling windows, test on subsequent periods), and check parameter stability (if optimal values swing wildly across periods, the model is likely overfit).

The Graveyard Matters

At Blackworks Capital, roughly 80-90% of hypotheses get rejected. Of the roughly 10% that pass initial screens, more fail intensive backtesting, and of the strategies that past this phase, they still may not make it into the live portfolio. Why is that? Because a strategy needs to not only pass individual tests, it needs to make sense within the total portfolio. If a strategy passes all the tests, works well live, but is too highly correlated to other strategies already being used, it gets shelved and tracked but doesn't get included in the portfolio.  

This isn't discouraging—it's evidence of intellectual honesty. A manager claiming most of their ideas work is telling you they don't test rigorously. The graveyard of failed ideas also creates institutional knowledge: what's been explored, why it didn't work, what assumptions proved unfounded. This learning sharpens every future research cycle.

Keeping It Simple

We trade equities and ETFs—including volatility and inverse ETFs for downside protection. No options, futures, or derivatives. 

This might seem limiting. It's actually clarifying. Every position is transparent. There's no off-balance-sheet exposure, no synthetic positioning that creates nonlinear tail risk. A drawdown is a drawdown, not a cascading margin call. Daily rebalancing stays clean without unwind calculations or gamma management.

From Backtest to Live Capital

The pipeline from research to production is deliberately slow. A promising model goes through committee challenge, extended backtesting, paper trading (running orders without real capital to verify behavior), then small-scale live deployment. Only if live performance matches expectations does capital allocation scale—gradually, with monitoring at each step.

This takes 6-12 months from hypothesis to meaningful deployment. It creates frustration for researchers eager to see ideas in production. But the slowness catches problems before they reach scale.

Watching It Run

Deployment doesn't end the work—it intensifies it.

We run daily performance attribution, decomposing each day's returns: how much came from the primary signal, from hedges, from unexpected market movement. We watch for model decay—gradually declining performance that signals the market has adapted or the regime has shifted. We track live performance against backtest expectations continuously.

When divergences appear, we do root cause analysis. Is it portfolio construction? Execution timing? Regime change? Data quality in the original backtest? The answer determines whether to adjust the model, reduce its allocation, or retire it.

The key discipline is distinguishing genuine adaptation (justified by structural market changes) from performative tweaking (chasing recent results). Constant adjustment based on recent drawdowns leads to overfitting. Thoughtful recalibration based on documented regime shifts is how strategies stay robust.

Third-Party Governance

Internal discipline matters. External oversight matters equally.

The BWC Founders Fund is a Delaware LLC administered by RePool, a leading Hedge Fund administrator, for independent NAV calculation and investor accounting. We're audited annually by Spicer Jeffries LLP. These structural safeguards ensure reported values aren't subject to manager discretion—investors can verify independently. For serious investors, these layers are non-negotiable.

What to Ask a Systematic Manager

When evaluating any systematic fund, skip the best-case backtest numbers. Ask about rejection rates. Ask about out-of-sample methodology. Ask about paper trading process. Ask how they detect model decay and regime change. Ask about worst drawdowns and what changed since.

The answers reveal whether you're looking at genuine systematic discipline or sophisticated curve-fitting. Our conviction is that sustainable alpha comes from rigorous process—multi-factor diversification, independent signal validation, and the disciplined execution of objective, rules-based logic. That distinction determines whether a strategy compounds real edge or decays into historical artifacts.

Get in touch to discuss how our systematic process aligns with your investment objectives.