Contents
What a bot is, mechanicallyWhat a bot can doWhat a bot cannot doWho bots suitA sensible first monthWhat a bot is, mechanically
A trading bot is a program that watches prices on an exchange and places orders when its rules say to. That is all. It watches, it decides, it places. The "watches" part is market data feeds; the "decides" part is the strategy — the rules; the "places" part is the exchange API. Between those three parts there is nothing magical, no intelligence hiding in a box.
What a bot can do
- Execute a plan without you in the chair. The plan still comes from you; the chair is just no longer yours.
- Be consistent at 4 a.m. The bot does not get tired, bored, or tempted to break its own rules mid-drawdown.
- React at candle close. Bots work on their own timeframe; a five-minute bot acts every five minutes, regardless of what a chart looks like on social media.
- Run many small experiments. With paper modes, a bot can run a plan with no funds at risk, which is the correct place to start.
What a bot cannot do
- Make a bad plan good. Automation repeats a plan faithfully, including the broken parts.
- Predict anything. Indicators describe the recent past; strategies bet on patterns. Some bets pay, none are prophecy.
- Watch the news. Unless a news feed is wired in and the strategy reads it, a bot will happily follow its rules off a cliff that a headline created.
- Stop being software. Bots crash, dependencies break, keys expire, machines restart. A bot needs a caretaker; it is infrastructure, not an appliance.
The honest one-line summary: a bot automates a plan you already have; it does not produce a plan for you.
Who bots suit
People who enjoy running their own infrastructure, who can read a bit of Python, and who want to test ideas with their own hands. If that description fits, the tools on this desk are the right kind of toy. If it does not, the hosted end of the market exists — it is just not what this desk covers.
A sensible first month
- Set the bot up on one exchange you already use, in paper mode. Nothing live, no exceptions.
- Run the shipped example strategy for a week. Do not change it yet — watch how it trades.
- Read the strategy file line by line. The goal is to predict what it will do before it does it.
- Change exactly one thing — the exit policy is a good first knob — and run another week.
- Write down, each week, what surprised you. The surprises are the actual curriculum.