Investment Research Technology
Research software,
honestly reported.
Proprietary software for building, testing, and monitoring quantitative investment strategies — including a register that keeps the results that did not work.
Please read first
This page describes software. It is not investment advice, not a recommendation to buy or sell anything, and not an offer of automated investment management.
Zach Edwards LLC is not a registered investment adviser. All strategies described here run on simulated (paper) accounts only. Investing involves risk, including loss of principal. Past or simulated performance does not predict future results.
Research and strategy architecture
How a strategy gets built and tested.
Hypothesis
A rule is written down before it is tested, in full, including the universe it applies to and the dates it covers.
Testing without hindsight
The hardest part of this work is not building a strategy that looks good on history — it is building one that was not chosen because it looked good on that history. Universes are selected without knowledge of which performed well.
Benchmarking against something dumb
Every strategy is compared to a deliberately naive alternative — typically equal-weighting the same holdings. A strategy that cannot beat the dumb version is recorded as not beating it.
Ablation
Each component is removed in turn to establish whether it contributes anything. Components that do not earn their place are removed from the strategy, not left in for appearances.
Risk management and review
Controls that are tested, not asserted.
A risk control that has never been shown to change an outcome is decoration. Ours are measured, and removed when measurement says they do nothing.
- Position and concentration limits enforced in code, not policy
- A long-term trend filter governing market exposure
- Every order validated against the account's own rules before submission
- Paper and live paths kept strictly separate
- Changes recorded with the evidence that justified them
Example of the above in practice: three stop-loss style controls were removed in October 2026 after testing showed they reduced returns without reducing drawdown.
The software
Portfolio monitoring and analysis.
What the system does, in plain terms. Interface captures and any numeric figures are deliberately withheld from this page until the investment-advertising review is complete.
Portfolio monitoring
- Current holdings and position sizes
- Exposure against the strategy's own limits
- Drift between target and actual allocation
- Every order checked against the account's rules before it is submitted
Strategy analysis
- Backtests over explicitly stated windows
- Side-by-side comparison with a naive equal-weight benchmark
- Ablation runs showing what each component contributes
- Maximum drawdown across the window, not just the headline return
Performance methodology and evidence
One register. Everything in it, including the failures.
Published results are selected after the fact almost everywhere. The defence against that is a register that records every strategy tested — so nothing can be quietly dropped for being inconvenient.
| Field | What it records |
|---|---|
| Strategy ID & name | Stable identifier so a result cannot be silently renamed |
| Start / end dates | The exact window tested |
| Classification | Simulated, paper, or live account |
| Capital & cash flows | Starting capital and every deposit or withdrawal |
| Source records | The underlying data the result was computed from |
| Benchmark & method | What it was compared against and how the return was calculated |
| Costs | Fees, expenses, and slippage assumptions |
| Returns | Gross and net, as applicable |
| Maximum drawdown | Worst peak-to-trough decline over the window |
| Verification date | When the entry was last checked |
What the register currently says — including the unflattering part
Paper onlyTwo verification passes were completed in October 2026 on simulated accounts. Their headline finding is not favourable to our own strategy, and it stays on this page for that reason:
- On universes chosen without hindsight, the strategy did not outperform simply equal-weighting the same holdings
- Its drawdowns were lower than the benchmark in every test performed
- Three stop-loss and drawdown-breaker components were measured to add nothing, and were removed rather than retained
Simulated results. Numeric figures are withheld from this page pending review. If a future version of this page shows only favourable results, that is a defect — ask about it.
Important distinction
Research software is not investment advice.
What this is: software that researches, tests, and monitors investment strategies, built and operated by Zach Edwards LLC for its own research.
What this is not: a recommendation to buy or sell any security, a managed account service, a signal subscription, or an offer to invest your money. Zach Edwards LLC does not hold investment-adviser registration and does not manage client funds.
If and when that changes, it will be stated here explicitly, with the registration it rests on. Until then, treat this page as an engineering portfolio entry.