Detect market phases
Separate exploratory periods from trend-like deterministic phases using a causal state machine.
Causal trading simulation – no real money at risk
Trade the predictable phase. Step aside from the noise.
Random Harvester® is a long-only trading simulation tool that scans price history left-to-right, detects trend-like phases, compares results against passive benchmarks, and separates part of realized gains into Banked Profit: cash the owner can treat as taken off the market-risk table.
Causal backtesting
No lookahead bias
Not technical analysis
Academic statistical methods
Long-only logic
Banked Profit tracking
Benchmark comparison
Reproducible reports
Built for quantitatively curious investors, researchers, and educators who want to examine strategy behavior before committing capital.
What it helps you do
Move from intuition to inspectable simulation.
Random Harvester® is not a prediction machine and not a technical-analysis package. It is a research environment built around academic statistical analysis, testing whether phase-based rules behave sensibly under realistic constraints.
Stay long-only
Model a disciplined approach that enters long positions and holds cash during unfavorable conditions.
Control bear behavior
Suppress new buys and tighten stop-loss handling when the detected phase environment deteriorates.
Bank harvested profit
Track the portion of realized gains converted into usable owner cash instead of leaving all wealth exposed to the next trade.
Compare against baselines
Read strategy results next to buy-and-hold and 60/40 benchmark behavior rather than in isolation.
Audit each run
Review signals, actions, cash, positions, banked profit, transaction costs, and comments step by step.
Generate reports
Produce self-documenting outputs with parameters, charts, trades, Banked Profit, drawdown, Sharpe, alpha, and commentary.
Method
Designed to avoid the usual backtest trap.
Every decision is made with information available up to the current bar. The engine does not smooth the future into the past, and it does not revise earlier phases after seeing what happened next.
01
Scan left-to-right
The price series is processed in chronological order, one pass only.
02
Classify the state
The detector moves between probing and deterministic regimes as evidence develops.
03
Harvest realized cash
After profitable exits, the skimming rule can allocate excess cash into the Banked Profit track.
04
Measure the path
Results are evaluated with returns, Banked Profit yield, drawdown, trades, and benchmark-relative statistics.
Banked Profit
A dividend-like cash track from the strategy itself.
Most backtests focus on total equity alone. Random Harvester® also tracks Banked Profit: cumulative realized cash skimmed from profitable exits, not automatically reinvested into the next position, and therefore less exposed to the uncertain future path of the market.
Shows usable owner cash
Banked Profit makes the cash extracted from the strategy visible instead of folding everything into a single equity curve.
Reduces exposure after gains
The banked portion is no longer assumed to be automatically committed to the next uncertain market phase.
Clarifies total owner wealth
Reports can show portfolio value and harvested cash together, so benchmark comparison is based on the owner's full outcome.
Aggregates across tickers
The overview combines per-ticker Banked Profit into a cumulative total, daily increment, and contributor list.
Highlights contributors
Daily banking events can show which tickers contributed to the cash harvested on that date.
Feeds the report
Reports can include harvested cash, Banked Profit concentration, and Banked Profit yield alongside classic metrics.
Outputs
What visitors should see next.
The strongest proof is the working trail left by each run: signals, portfolio state, benchmark context, and reportable metrics.
Phase timeline
Price series with phase annotations, buy/sell markers, and bear-mode state.
Portfolio graph
Strategy equity curve next to benchmark behavior and drawdown context.
Banked Profit Overview
Dividend-like owner cash, daily increments, and the tickers contributing to each banking event.
Statistics and report
Sharpe, alpha, max drawdown, trades, win rate, parameters, and generated commentary.
Evidence on request
Benchmark reports are available for serious review.
I maintain benchmark test reports for both a long-term 2016-2025 window and a current one-year view on the same ticker universe. I do not intend to publish them as headline marketing material, but they are available on request for readers who want to inspect assumptions, benchmark comparison, Banked Profit behavior, and the simulated owner-wealth outcome in more detail.
Long-term benchmark test
Historical simulation over a 2016-2025 evaluation window, compared against a constant-mix 60/40 benchmark.
Current one-year view
A more recent view on the same ticker universe, useful for seeing how the logic behaves in the latest market environment.
Simulation caveats included
The reports include explicit backtest limitations, including parameter calibration, universe selection, fixed cost assumptions, and the need for out-of-sample validation.
Who it is for
For careful market research, not trading hype.
Individual investors
Test ideas before putting money at risk and learn how a rule set behaves across different market conditions.
Analysts and quants
Use a visual front end for phase-based strategy experiments, portfolio inspection, and exportable results.
Lecturers and trainers
Demonstrate causality, drawdown, benchmarking, signal quality, and overfitting in a hands-on setting.
FAQ
The important boundaries.
Is Random Harvester a live trading platform?
No. It is currently best presented as research and simulation software. Optional IBKR-related modules exist, but the website should lead with simulation, not live execution.
Does it predict prices?
No. The premise is to detect when recent market dynamics become more structured or trend-like, then step aside when that structure fails.
Is it technical analysis?
No. The method is better described as causal, phase-based statistical analysis rather than chart-pattern or indicator-driven technical analysis.
Who executes orders?
Not me. Order execution belongs with professional traders and professional execution infrastructure. The public website should be understood as describing research and simulation work.
Can it guarantee better returns?
No. The value is in transparent testing, benchmark comparison, and risk inspection, not in guarantees.
Is Banked Profit the same as a dividend?
No. It is dividend-like from the owner's point of view, but it is generated by the strategy's realized exits rather than paid by the underlying company.
Do users need to program?
The desktop interface is intended to make the workflow inspectable, while the Python codebase remains accessible for deeper research and extension.
Demo
Interested in Random Harvester®?
I am refining the tool and exploring conversations with investors, educators, and research partners who care about transparent strategy testing.
Request a demoDirect contact: eric.van.horenbeeck@randomharvester.com