Operating an automated trading bot is a discipline, not a continuous task. The bot does the work; you provide oversight. The right operator routines build situational awareness without demanding constant attention. This page is the rhythm guide.
The four cadences
| Cadence | Time per session | Focus |
|---|---|---|
| Daily | 2–5 minutes | Health check — is the bot doing what I expect? |
| Weekly | 30 minutes | Structured review — analyzing performance, decisions |
| Monthly | 60 minutes | Tax archiving + capital review |
| Yearly | 2–4 hours | Full audit + accountant handoff |
Daily — the 5-minute health check
Glance at the main panel
Confirm:
- Kill switch is in expected state (usually OFF for normal operation).
- Connection-state indicators all green for every active venue.
- No urgent alerts or warnings displayed.
Skim the Live Trades panel
Quick visual scan:
- Are positions roughly where you expect?
- Any held abnormally long?
- Any unexpected positions?
Glance at the day's closed trades
Compare against your Telegram chat. Totals match? Reasonable count?If your Telegram running total looks off, investigate further. Otherwise move on.
2–5 minutes. The goal is not to make decisions every day; it’s to confirm “the bot is doing what I expect.”
Weekly — the 30-minute review
Pull the week's analytics
Dashboard’s analytics panel for the last 7 days. Compare against last week and your monthly running average.Trends to notice:
- Trade frequency stable or shifting?
- Win rate within typical range?
- Drawdown character similar to backtest expectation?
Identify the week's worst trade
Open the order detail. Was the mode behaving as designed, or was there an anomaly?
- Mode-design behavior → expected, no action.
- Anomaly (unusual stop-loss trigger, weird fill, unexpected exit) → flag for investigation.
Identify the week's best trade
Same exercise. Was the gain due to the mode hitting expected rungs, or due to an unusually large move that the mode happened to capture?Lucky big wins are different from designed wins. Distinguish.
Check the Logs panel for recurring warnings
Search for any patterns. Recurring errors are a flag.Common: occasional “rate limit” warnings during busy market periods. These are normal if rare. Concerning: recurring authentication failures, configuration mismatches.
Confirm all expected pairs have traded
A pair that hasn’t closed a trade in a week is a flag worth investigating. Possible causes:
- Capital allocation too small for min-notional.
- Mode’s buy conditions never triggered.
- Symbol moved in a way the mode doesn’t capture.
Monthly — tax archive + capital review (60 minutes)
Pull the Tax Report
Dashboard → Tax Report → select last month → export CSV/JSON.Archive to your designated tax folder (cloud drive or local with backup).
Cross-check with Telegram running total
Last Telegram trade-close of the month should show running total close to (or matching exactly) the tax report’s net P&L.Discrepancies investigate immediately — usually a data restore boundary or a manual close not reflected.
Review capital allocation
Is each mode’s allocated capital still appropriate?
- Is reserve still ~50%?
- Has any pair drifted outside the 25% rule?
- Should you scale up (after multiple consecutive non-negative months)?
- Should you scale down (if regime mismatch persists)?
Audit the venue's API key activity
Each venue’s audit log shows API key activity. Review for the month:
- All calls came from your VPS’s IP?
- Any unfamiliar activity?
- Any unusual error rates?
Plan structural changes for next month
Any deliberate changes? New venue, new mode, new pair? Document the plan; execute it deliberately at month-start, not on impulse mid-month.
Yearly — the full audit (2–4 hours)
Verify against monthly archives
Sum of 12 monthly Tax Reports should equal (or closely match) the full-year report. Investigate discrepancies.
Strategic review
Big-picture questions:
- Did the bot do what I intended this year?
- Should next year’s setup look different?
- Are my modes/pairs/venues optimized for the regime I expect?
- Should I add or retire anything?
Operator runbook review
Update your runbook with this year’s learnings:
- Any new patterns observed?
- Any incidents that taught lessons?
- Any decisions you’d make differently?
Crisis routine — when something is going wrong
Step back. Coffee. Walk.
Most “panic decisions” worsen outcomes. Take 5 minutes before doing anything.
Open the Dashboard
Look at actual positions, actual P&L, actual unrealized exposure. Not what you fear; what is.
Compare to backtest expectations
Is this within historical drawdown ranges? If yes, the bot is operating within design parameters. Hold. If beyond historical range, engage further.
Decide: kill switch or not?
If you’re genuinely worried (not just stressed): kill switch ON. Cost of false alarm: zero. Cost of buying into worsening situation: substantial.
Wait for clarity (an hour, a day)
Most market events resolve within hours. Wait for the situation to clarify before further action.
Best practices
What’s next
Monitoring
Tools and habits for spotting issues before they’re crises.
Tax export
The monthly/yearly tax routine in detail.
Updates
How to update unCoded safely.
Support
When and how to reach out for help.
Risk overview
The risk framework these routines operate within.
Troubleshooting
For when something is wrong.