I built cryptobot-research to look for crypto trading strategies that could survive disciplined testing. I ended that open-ended search and shifted to a more focused task: independently testing strategies other people have published. Earlier studies used Strategy Lab and the completed cryptobot-research work; new independent studies use research-platform.
Backtest Discipline is my independent, noncommercial research publication, separate from the research software and trading runtime. The research is freely accessible. I publish individually scoped evaluations here, alongside selected cases from the initial discovery phase. I report what the evidence supports, whether the result is favorable, unfavorable or inconclusive. I aim for a new evaluation roughly every two weeks; some cases need longer.
What this site is for
Alongside the return, I document where the idea came from, which data I used, how I interpreted the rules, and what counted as success before evaluation. The research record shows concrete outcomes, Methodology explains the controls behind them, and the engineering notes pull reusable lessons out of the preserved cases.
What it is not
Backtest Discipline does not sell strategies, trading signals, subscriptions, backtests or investment advice. It is not a managed fund or a live trading bot. The research software is separate from the trading runtime, and publishing a result does not authorize a strategy to trade.
The site also does not treat historical backtest performance as evidence of guaranteed or deployable profitability. A reproduction can succeed while the underlying methodology still has limitations; a strategy can show a positive headline return and still fail the acceptance rule that was frozen before evaluation.
Evidence policy
I tie research claims to preserved evidence and publish selected frozen specifications, decision rules, evaluation outputs, hashes and later publication reports where they help readers inspect a case. These are selected artifacts, not a complete rerun package: they let readers inspect parts of the evidence and decision process, but do not by themselves supply everything needed to reproduce a run.
Those publication reports are summaries created after the relevant experiments had closed. They are not presented as original runtime artifacts, and generating them did not replay, retune, recapture, or re-evaluate the experiments.
I do not publish raw provider payloads, private machine paths, binaries, credentials, internal runner material or the complete publishing workspace.
Maintainer and publication source
Backtest Discipline is written and maintained by Mikhail Klemin. I develop the research software and conduct the evaluations published here.
The website source and editorial workspace are private. Readers can inspect the selected evidence artifacts, hashes, methodological explanations, case studies and publication reports on the site; drafts and internal development history are outside that public record.
Corrections and contact
To report an error or broken evidence link, ask a methodological question, suggest a research topic, or provide constructive feedback, email contact@backtestdiscipline.com. Please include the page and, where relevant, the evidence artifact or hash so I can check the issue against the preserved record.
Email is the current contact route. I keep the site small and static.
The research boundary
New independent evaluations use research-platform, each with its own scope and frozen test. Completed experiments stay closed: an article does not silently change their data, parameters, holdouts, or verdicts. Additional empirical work requires explicit authorization in the research project, not an editorial change to this site.