The development process
AlgoLab describes the development and evaluation of strategy candidates. Historical testing, independent validation and later go-forward observations provide different evidence. Publication and service activation are separate decisions; a promising result does not promise future profit.
Generate ideas
We define a market, trading rules, time horizon and conditions to investigate. Candidate rules must use information that was available at the trading decision. Unsuccessful research does not become a service offer.
Screen Historical candidates
Historical simulations help examine trade count, returns, drawdown, behavior and costs. We retain the report’s original source, version and period. Simulated trades remain hypothetical.
Published Historical reports are selected research. Searching many rules or settings can produce an impressive surviving backtest by chance; the selected report alone does not reveal how many alternatives were tried. Evidence from data excluded from development and subsequent Model Account observations helps challenge that selection. Missing validation is not a passed test.
Check unseen data
We assess behavior on data excluded from development. Data reused to adjust or select rules is no longer an independent final test. Passing a test cannot establish an enduring edge.
Examine robustness
Trade-sequence variation, sensitivity checks and cost assumptions explore how fragile the result may be. These methods do not remove overfitting or the possibility of future losses.
Identify the version and publish Historical research
An implementation must be checked against its instruments, strategy version and execution assumptions. A backtest does not establish identical behavior across brokers or customer accounts.
Historical means a simulation on past market data. If Historical research is published in a future release, it must retain its own source, tested version, period, capital and costs. It must not be relabelled as an AutoTrade go-forward record or silently joined to one.
Build a Model Account history
An AlgoTrader strategy’s go-forward record is built from signals submitted as trading decisions occur. Uploading a previous backtest does not create an AutoTrade track record. The automatic NinjaTrader-to-AutoTrade signal publisher is not part of this version.
The current public report uses AlgoTrader model-account statistics, monthly returns, closed trades and model positions. All displayed strategy performance remains hypothetical, including models using broker-reported source fills. It is not a customer brokerage statement.
Review the evidence
Published sources retain their observation dates and disclosed costs. Missing coverage, delayed data or unavailable metrics must remain apparent. Importing data again or updating the website does not restart an evidence period.
Compare Model Account results in the Leaderboard
Published AutoTrade records can be researched before any customer authorizes AutoTrade. Source metrics retain their meanings; no official AlgoTrader Score or rank is assigned without a calibrated method and the required evidence.
Your decision
Review the available AutoTrade record, save a shortlist and choose one strategy or several. Research, account choice and personal risk settings remain separate decisions. Missing validation, a high return or display ordering does not establish suitability or portfolio performance.
Questions about this guide
Does a published research report mean AutoTrade is released?OpenClose
No. Research publication and service activation are separate. Check the current availability and actual service offer.
Does a successful Historical test guarantee future returns?OpenClose
No. Historical simulations and robustness checks cannot guarantee that future market conditions or your account execution will produce similar results.
Why is a strong backtest only part of the evidence?OpenClose
A published backtest is selected research. Trying many rules or settings can produce a strong result by chance. Review documented tests on data excluded from development and the separate Model Account history; neither guarantees future performance.