Research, development, and operations
The same codebase that powers historical simulation drives live execution. Strategy promotion runs through four gated stages from research to live capital, and a small set of disciplined principles governs how the system is built and operated day to day.
Switching from research mode to production is a configuration change, not a rebuild. There is no translation layer and no divergence between what was tested and what runs. Look-ahead and survivorship bias are structurally impossible: all research runs against a streaming interface that enforces temporal isolation at the data boundary.
The simulator replays a market feed through an internal custodian model that accounts for realistic fill rates, bid/ask spread, and market impact. Research results are directly comparable to live performance.
The identical signal and strategy code connects to Interactive Brokers via FIX. Strategies complete a forward-test period on IBKR paper accounts before any live capital is committed. The only change is the data source and execution target.
Historical simulation with the integrity of a live track record
The architectural unity between research and live trading has a direct consequence for investors: the five-year historical simulation is not a backtest assembled in hindsight. It is a reproducible pipeline run forward from a fixed start date, governed by the same temporal isolation and realistic fill modeling that applies to every live trade. The result is a track record that reflects what would have happened, not what the model was tuned to show after the fact.
The strategy can and will evolve over time as new features, models, signals, and ideas are statistically proven out through a large backdrop of extremely granular market pricing and reference data.
These improvements flow through the same pipeline and benefit investors in both stages: early participants within SMA-isolated trading as well as future investors in the commingled fund. Progress made on the platform is cumulative. The work compounds.
Staged strategy development
A quantitative strategy is only as durable as the discipline behind it. The published backtest numbers are just the first step; the hard part is running the system for years without a small mistake compounding into a large one. The structure that follows exists so that the live performance you receive in your IBKR account tracks the research output, not because we hope it will, but because the pipeline that produces the trades is the same pipeline that produced the backtest.
Systematic trading undergoes four distinct, gated cycles, with rigorous acceptance criteria that must be met at each stage before moving on: research; historical simulation with portfolio and risk management; paper trading with shadow reconciliation; and finally, live execution.
A signal must clear each gate before it advances; sustained drift in live trading sends it back.
Research
Research is where candidate signals are identified, formalized as a precise rule set, and tested across all the historical data. At this stage we isolate alpha on a trade-level basis, independent of portfolio construction and risk management; this is foundational to establishing whether a signal is effective and sustainable. We characterize the arbitrage opportunity and where it comes from (risk arbitrage, statistical arbitrage, and so on), and whether the signal has statistically deteriorated over time.
Every candidate has to clear walk-forward validation on out-of-sample data the strategy hasn't been trained on, and an evaluation against a relevant benchmark on a risk-adjusted, net-of-cost basis, while remaining economically interpretable.
Historical simulation and portfolio / risk management
A signal that survives research moves into historical simulation, where the question shifts from "is there alpha?" to "does the alpha survive when it has to run inside a real portfolio?" Here we add deployment logistics on top of the raw signal: position sizing rules, concurrent-position limits, exposure caps at the name, sector, and factor level, margin consumption, and the way the strategy interacts with itself when multiple signals fire on the same name in a short window.
A signal that looked clean in isolation can create unmanageable concentration at full sizing, or generate more turnover than the cost model can absorb in practice. This stage exists to find that out before any real capital is involved. It's where allocation and risk management become first-class parts of the evaluation rather than afterthoughts.
Paper trading and shadow reconciliation
A strategy that survives historical simulation runs in paper-trade mode against live market data. The system places the same orders it would in a live account, the orders fill at observed prices, and the resulting performance is logged. Every paper trade is shadow-reconciled against what the historical simulation said the same trade should look like: fill price, fill timing, post-execution exposure, P&L attribution. Any divergence between paper and simulation is investigated before live capital is allowed.
Paper trading runs for at least one full cycle of the strategy's natural holding period and has to hit pre-specified performance thresholds. This stage can require several months of monitoring and investigation to make sure both the strategy and the systems are fully robust before moving to live execution.
Live execution
A strategy that clears paper trading is allowed to touch real capital, in configurable tranches. Once live, it is monitored continuously against the shadow run and the historical simulation, and any sustained drift returns it to research for re-evaluation.
All trades are tracked in real time, feeding intraday operational dashboards and internal communication channels, including Slack messaging and daily, weekly, and monthly digests of trading and signal activity. The strategy is continuously reviewed for any divergence from historical expectations and reconciled against shadow trading.
Disciplined system principles
Five principles shape how the system is built and how it runs day to day.
-
Realistic simulation environments
We model market impact and slippage on every order, interest earned on cash and charged on margin, and corporate actions such as splits, dividends, mergers, and spin-offs. Research runs on the same real-time market-data interface that drives live execution, so look-ahead is structurally impossible: research sees exactly the data live execution would have had, and no further apart.
-
Sustainable alpha generation
Every number that gates a strategy comes from out-of-sample data. Strategies are built only from point-in-time information; anything not known to the market at that moment is unavailable for features, training, or analytics. In-sample performance, which is subject to overfitting, is never used as a benchmark.
-
Configuration-driven multi-account fan-out
Every account runs from the same execution engine, with per-account behavior expressed as configuration rather than code. Separately managed accounts and commingled vehicles run side by side without forking the system. A single rule change ships to every account at once, and no account carries bespoke logic that could quietly drift.
-
Intraday reporting and communication
Every trading decision feeds real-time dashboards and notifications, with Slack and email as the primary channels. Each alert carries the reason for the trade and whether the shadow simulation took the same trade. We forward-test alongside live capital constantly, so we can spot anything off before it drifts into real misalignment.
-
Emergency switches and thresholds
A small set of conditions short-circuits normal execution and closes out all open positions. Triggers are configured at the portfolio level; when one fires, the system flattens the book at the next available prices and pages the on-call operator. The bias is toward exiting quickly and investigating from a clean position rather than holding through ambiguity.
The same pipeline that produced the backtest produces every live trade. When you're ready, see how your account fits into it, or start the conversation.