How Crypto Day Traders Automate Strategy Improvement
For Crypto day traders with an existing manual strategy · Based on Lewis Jackson Self-Improving Trading Agent Framework
// TL;DR
If you're a crypto day trader with an existing strategy and trade history but no systematic way to improve, this framework converts your manual approach into an autonomous self-improving agent. Feed your strategy file and trade history into the oneshot prompt, define your success metrics (e.g., monthly return target, Sharpe floor) and failure thresholds (e.g., max drawdown), and Hermes takes over the improvement loop — testing one variable change per week and promoting winners as new baselines. Your strategy compounds learning while you sleep.
Why do most manual crypto trading strategies stop improving?
Manual crypto traders hit a ceiling because improvement is ad hoc. You might tweak your stop-loss after a bad week, adjust position sizing after a big win, and change your entry signal all at once — then have no idea which change actually helped. The Lewis Jackson Self-Improving Trading Agent Framework solves this by applying the scientific method to your strategy iteration: one variable change per cycle, measured against clearly defined success and failure thresholds.
If you have 30-50+ historical trades and a documented strategy (entry/exit rules, position limits, slippage tolerance), you already have what you need to start.
How do I convert my existing strategy into a self-improving agent?
During the oneshot prompt's Phase 2, choose path A — point the agent at your existing strategy file. The onboarding agent extracts your current parameters: position limits, entry signals, slippage tolerance, gas reserves, and more. In Phase 3, your trade history (wins and losses) is converted into a Hermes-readable ledger.
Define your success criteria specifically. Instead of "I want to be profitable," set targets like: 15% monthly return, Sharpe ≥ 1.2, maximum drawdown ≤ 12%. Your failure definition must be equally specific: two consecutive months below 5% return, or a single 30-day drawdown exceeding 15%.
Hermes uses these definitions as the polarity for every improvement decision.
What does a typical improvement cycle look like for a day trader?
After the first read-only cycle (which you must review before enabling live mode), Hermes might identify that your position sizing on losing trades was 40% larger than on winning trades. It proposes reducing max position size by 20% as the sole change for Cycle 2. You approve.
Cycle 2 runs with this single change. If results move toward your success definition, the reduced position size becomes the new baseline. If not, Hermes forms a new hypothesis — maybe testing a tighter trailing stop instead. Each week, one variable, one test, one clear signal.
Over 12 cycles (three months), this produces a strategy that has been systematically refined 12 times with clean attribution at each step.
How do I avoid the biggest mistake day traders make with this framework?
The number one pitfall is changing multiple variables between cycles. If you manually adjust your entry signal AND your position sizing between Hermes reviews, the learning signal is corrupted. You won't know which change drove the result. Discipline yourself to let Hermes control the iteration pace: one variable per week.
The second biggest mistake is skipping the read-only cycle review. Your first Hermes output validates that it correctly parsed your strategy and goals. Skipping it risks live capital on a misconfigured loop.
What's the fastest way to get started?
1. Document your current strategy's entry/exit rules, position limits, and target asset
2. Define your success and failure thresholds in writing
3. Retrieve the latest oneshot prompt from the 01 Systems community
4. Open Claude Code, paste the prompt, and follow the guided onboarding
5. Authenticate Railway for 24/7 hosting
6. Review the first read-only Hermes cycle before enabling live mode
Your strategy stops being static the moment Hermes starts its first improvement cycle.
// FREQUENTLY ASKED QUESTIONS
Can I use my TradingView strategy with this framework?
Yes, if your TradingView strategy can be exported as a documented set of rules (entry/exit conditions, position sizing, indicators used). During Phase 2, point the agent at your strategy documentation file. The agent extracts parameters from it. Pine Script files can serve as source material, but the agent needs clear rule definitions, not just chart-based visual setups.
How much trade history do I need before starting?
The framework works best with at least 30-50 historical trades — enough for Hermes to identify patterns in wins versus losses. More history gives Hermes richer data for its initial analysis. If you have fewer trades, the framework still works but the first few improvement cycles may produce weaker hypotheses. The read-only first cycle helps you verify Hermes has enough signal to work with.
Will this work for scalping strategies with very short hold times?
Yes, the framework is agnostic to hold time. Scalping strategies generate many trades per week, which actually gives Hermes more data per improvement cycle. The key requirement is that your success and failure definitions are measurable and your data feed is accurate enough to capture rapid entries and exits. Slippage tolerance becomes especially important for scalping and should be explicitly defined during Phase 2.