cryptotradingbook

Bots and automation

Intermediate → Advanced · ~11 min read

At a glance

  • A bot executes a rule set faster than you - it does not create the rule set.
  • Automation amplifies whatever the strategy is: edge scales, and so does flaw.
  • API keys are attack surface: trade-only keys, withdrawal permissions OFF, IP whitelist ON.
  • Most retail "bots" sold online are grid/DCA wrappers - or outright exit scams.
  • Automation without a kill switch is a liability, not a tool.

1. What a bot is - and is not

A bot is a rule executor: it places the orders your plan defines, without hesitation or fatigue. It has no edge of its own - a bad strategy automated is a bad strategy with better uptime. The honest starting question is never "which bot" but "is the rule set worth automating" - and the backtest pitfalls apply before a single line runs.

2. What bots are genuinely useful for

Execution discipline: a bot never oversizes after a loss, never chases, never "feels". DCA schedules, rebalancing rules, alerting (a price crosses your zone - you decide), order placement speed. Grid bots mechanize range-trading - and are implicitly a bet that the range holds; they accumulate the worst position exactly when the range breaks. The tool is honest; the marketing rarely is.

3. API keys: the attack surface you control

The standard breach path is not the exchange - it's the user's API key: full-permission keys in compromised servers, cloud-synced config files, or pasted into "free bot" web forms. The rules: trade-only keys (withdrawal permission OFF, always), IP whitelist ON, dedicated keys per bot, rotate on suspicion, and never - ever - enter exchange API keys into a third-party website you have not audited. A bot with withdrawal rights is a loaded gun on a nightstand.

Remember: If a "free bot" asks for API keys with withdrawal rights, it is not a bot. It is a withdrawal tool wearing a bot costume.

4. The retail bot market, honestly

Three categories: (1) Exchange-native tools (grid/DCA bots) - mechanically safe, strategy-naive, fine for learning mechanics. (2) Open frameworks (self-hosted, e.g. open-source trading bots) - powerful, but you own every bug. (3) Sold "profitable bots" - subscription signals, closed-source black boxes with claimed track records: the market where curve-fitted backtests and outright scams meet. A bot that reliably printed money would be run by its author, not sold for $99.

5. When this fails

Automation fails operationally: exchange API changes, rate limits, a stalled order loop accumulating positions, an outage during volatility. A bot without a kill switch (manual cancel-all, hard position limits, an alert when it stops heart-beating) is an unattended position generator. The journal question stays the same: is the bot following the plan - and would you have made those trades yourself?

Risk: Bots are the only trading tool that can lose money while you sleep, at scale, without asking. The kill switch is not optional equipment - it is the product.

Knowledge check

  1. What does a bot add to a strategy - and what does it not add? (Execution speed/discipline - not an edge; the rule set is the strategy.)
  2. What are the non-negotiable API key settings? (Trade-only: withdrawal OFF, IP whitelist ON, per-bot keys, rotate.)
  3. What do grid bots implicitly bet on? (That the range holds - they accumulate exposure exactly when it breaks.)
  4. Why is a "profitable bot" for sale a contradiction? (Its author would run it rather than sell it - reliable money doesn't need $99 subscriptions.)
  5. What must every automated setup have? (A kill switch: manual cancel-all, position limits, alerts.)