Quantellect Insider Net Flow
STRATEGYEVALUATED RESULTSIts DEVELOPMENT BACKTEST measured a net equity gain of +$102,604, a 62.0% win rate, 1.32 profit factor, and 57.2% maximum drawdown. Backtests describe historical results and do not predict future results.
Quantellect Insider Net Flow
How the strategy works
Quantellect Insider Net Flow compares qualifying insider purchase and sale dollars over a 120-day window. A stock must meet minimum total activity and purchase-value thresholds, along with the configured minimum share of activity represented by purchases. The default score ranks eligible names by that purchase ratio. At each monthly review close, the ten highest-ranked stocks become an equal-weight target basket whose trades are modelled at the next session's open. The portfolio uses whole split-adjusted shares, remains long-only, and can hold positions overnight through later reviews. It uses no leverage or stop-loss. Qualified public-availability dates control signals; trading-plan status may be incomplete, while dividends and cash interest are excluded.
- Signal: Ranks the buy share of two-sided open-market dollars over 120 filing days; at least $250,000 total, $50,000 purchases and 50% buy share; top ten.
- Parameters: {"top_n":10,"lookback_days":120,"min_total_usd":250000,"min_sell_usd":0,"min_buy_usd":50000,"min_buy_ratio":0.5,"ranking":"ratio"}
- Public availability: Form 4 filing day, available at end of that day; execute next session open. Reconstructed aggregate windows use filing dates, never undisclosed transaction dates.
- Our engine: Independent event selector feeds the shared-cash portfolio engine; 548 admitted histories, monthly last-session signals, next open, no terminal liquidation.
Intended use and limitations
Rules-based historical research for comparing a pinned strategy under a common portfolio model.
Pinned class defaults, not separately optimized shipping configurations. Fixed admitted universe is selected using all declared discovery trials; survivor, complete-history and outcome-dependent universe bias remain. No sealed holdout. Source events and identity mappings are current-knowledge revisions; public-date gating does not reconstruct historical vendor vintages. Stable first/last native filing symbols bridge issuer identity; changed or ambiguous mappings and amendments without a version replay are excluded. Intermediate symbol/role corrections remain a historical-vintage limitation.
Development backtest; historical periods were not kept untouched for a final independent test.
Split-adjusted price return excludes dividends, cash interest, taxes, market impact, and forced terminal liquidation.
Insider trading-plan status can be unknown.
Clone
Clone isn't available yet. When it is, you'll be able to make a private version of this strategy to change and test.
Auto-Trade
Auto-Trade isn't available for this strategy yet.
Backtests describe historical results and do not predict future results. These DEVELOPMENT BACKTEST results are hypothetical, not actual trading, and do not promise similar future results.
This disclosure is published by Quantellect Technologies Inc. for quantellect.ai.
Simulated results
Backtests, paper trading, research portfolios, estimated returns from public disclosures, scenario analysis, and examples generated by automated systems are hypothetical. Any result claimed to be actual trading must be separately identified and substantiated as such.
A hypothetical result is a calculation based on a model and assumptions. It does not show that a real account placed the trades, obtained the assumed prices, paid the assumed costs, or earned the displayed return. No representation is made that a user will achieve similar profits or losses.
Our research engine
Quantellect uses an internally developed backtesting engine. A published result should identify the engine version and run identifier used for that result. Where a result on this site does not show that information, it is a summary presentation and carries additional uncertainty; treat it as an illustration of a research rule, not as a reproducible record.
Different engines and different configurations of the same engine can treat order timing, price matching, partial fills, account balances, margin, fees, and event sequencing differently. Agreement between two runs cannot establish that their common assumptions are correct.
Hindsight and selection
Strategies are often designed after examining historical data. Selecting instruments, signals, parameters, start dates, or examples after viewing outcomes can make results look more favorable than a strategy would have looked in advance.
Repeated experimentation increases the chance of finding apparent success by luck. Performance can change materially when tested on a different period, instrument universe, market regime, or dataset. A successful evaluation period does not eliminate these risks.
Data availability and revisions
A simulation may use data that was revised, corrected, or unavailable when a historical trade would have occurred. Reporting date, transaction date, filing acceptance time, public release time, provider availability, and Quantellect ingestion time are distinct.
For a strategy based on filings, estimates, or economic releases, the availability assumption is especially important. Corporate action adjustments, delisted securities, stale instruments, symbol changes, missing observations, and the treatment of failed companies can materially affect performance. A current database is not necessarily a reliable reconstruction of what was knowable in the past.
Execution and costs
Historical prices do not guarantee executable prices. A model may not fully reflect bid and ask spreads, queue position, routing, latency, slippage, transaction fees, commissions, exchange fees, market impact, price limits, trading halts, rejected orders, or limited available volume.
A bar can contain a high and a low without establishing the order in which they occurred. An assumed entry, stop, and target in the same bar can therefore depend on an artificial sequencing rule. Results also depend on whether a signal can trade on the same observation used to generate it.
Unless the accompanying assumptions expressly include them, do not assume results include taxes, financing, stock borrowing fees, locate availability, recall risk, data costs, platform costs, or the value of the time needed to operate a strategy.
Options, futures, leverage, and short selling
Options simulations depend on contract selection, quote quality, spreads, assignment and exercise assumptions, expiration treatment, and volatility inputs. A quoted midpoint or theoretical valuation is not proof that an order could be filled.
Futures results depend on contract selection, roll treatment, continuous contract adjustments, expiry, settlement, margin, funding, and trading session definitions. Synthetic continuous prices may not correspond to an instrument that could have been bought or sold.
Leverage, margin calls, gaps, and short positions can create losses beyond initially committed funds. Simulated accounts may fail to capture the practical constraints and financial pressure of such losses.
Performance statistics and comparisons
A return depends on starting capital, leverage, reinvestment, cash flows, fees, exposure, currency, and the measurement period. Annualizing a short period can exaggerate its apparent significance. Win rate alone does not describe the size of gains and losses. A favorable average can conceal large drawdowns, illiquidity, or a few exceptional observations.
A benchmark is useful only with a clear comparison basis, including dates, dividends, currency, costs, and exposure. A score, confidence value, or ranking is not a calibrated probability unless its methodology expressly states and supports that interpretation.
Public disclosure estimates
Returns attributed to politicians, insiders, institutions, or other reported investors may be estimates based on incomplete disclosures. Reported value ranges, unknown trade prices, delayed disclosure, omitted positions, transfers, amendments, and unreported costs can materially change the result.
A portfolio reconstructed from a filing is not the filer's complete account record. A displayed estimate is not an audited personal return, proof of current ownership, or evidence that the person caused a market movement.
What to look for with each result
Read a result with its strategy description and version, engine and run identifier, instrument universe, evaluation dates, data sources and availability assumptions, starting capital, sizing and leverage, execution rules, costs included and omitted, benchmark, and material limitations, wherever those are shown. If assumptions or source information are missing, the result has additional uncertainty.
A presentation that distinguishes the period used to develop the strategy from any separate evaluation period, and explains material selection, optimization, or data exclusions, is more informative than one that does not.
Changes and corrections
Research methods, data, and software can change. A result reproduced after a correction can differ from an earlier version. Refer to the publication date and run information where shown, and report suspected problems to [email protected].
Use these results to examine research hypotheses. They do not establish suitability, assure future performance, or replace independent evaluation and appropriate professional advice. Rights and obligations that cannot lawfully be excluded remain unaffected.

