For first-time bot developers (or anyone who has never been profitable)
Hard-earned advice from 13 years of trading and 10 years of building bots — edges, TA, AI strategies, Python vs Rust, and why simple EV+ systems beat holy grails.
bots · advice · trading · beginners
This is advice from us at Nexrade.
We have about 13 years of trading experience and 10 years building trading bots. We have been wrong a lot. We have also been right enough to keep going. What follows is what we wish someone had told us earlier — not a course, not a signal service, and not a promise that you will get rich.
Can you get profitable from bot trading?
Yes.
Bots are tools. Profit comes from a strategy with positive expected value, executed with discipline, after fees and spread. A bot does not invent the edge; it repeats it without sleep, fear, or FOMO — if the edge is real.
Can you find a “holy grail” of trading?
Yes — but only for a range of time, not forever.
Markets change. Regimes rotate. Liquidity, fees, and competition move. Something that prints for six months can die in six weeks. Treat every edge as temporary until proven otherwise, and keep measuring.
If someone sells a permanent holy grail that never stops working, they are selling hope, not a model.
How much can you make from bot trading?
For most people who eventually do well, think in small percentages — a few percent per day in the best stretches of aggressive systems, or more often a few percent per month on calmer, more sustainable setups.
That does not sound exciting. Compounded over a long time with controlled risk, it is how real accounts grow. Chasing “double every week” is how accounts die.
“I’m not good at math”
Math matters for logic — rates, fees, expectancy, position size — but you do not need to be a mathematician.
Average math is enough if you understand:
- What you pay per trade (fee + spread + slippage)
- How often you win vs lose
- How large wins and losses are
- That a high win rate with bad R:R can still lose money
You can learn that without advanced calculus. You cannot skip it.
Can you trust technical analysis (TA) signals?
Sometimes yes, sometimes no — roughly half the time in the long run if you treat charts as magic.
If you can see a pattern on a chart, others can see it too. When enough people act on the same idea, the market rebalances. The edge from “obvious” TA often gets competed away.
TA can still be useful as a framework for structure (levels, volatility, regime). It is dangerous as a religion. Indicators are public information, not a secret language.
How long until someone becomes profitable?
This is a hard question. For us, from the first bot to a period we would honestly call “profitable and stable,” it took about five years.
Your timeline may be shorter or longer. Skill, capital, risk control, and luck all matter. Anyone who guarantees “profitable in 30 days” is selling a story.
Can AI-generated strategies beat the market?
If you do not understand the strategy, about 99.99% of the time you will lose money.
AI can write code and invent rules. It does not automatically understand:
- How the strategy actually works
- Fees and funding
- Spread and liquidity
- Market type (spot vs futures, trending vs mean-reverting)
- What happens when the regime changes
Blindly deploying a black-box “AI strategy” is not innovation. It is gambling with extra steps. Use AI as a assistant after you can explain the edge yourself.
Which language should you choose for trading bots?
| Choice | When it fits |
|---|---|
| Rust | You truly need runtime performance, low latency, tight resource control |
| Python | Almost everyone else — readable, fast to write, easy to update and maintain |
Python is slow compared to Rust. For most retail and many semi-pro bots, speed of writing and maintaining code beats speed of execution.
Money usually does not come from shaving microseconds on a bad idea. It comes from a clear idea you can fix on a Sunday afternoon. Choose the fast language later, when you know what is worth optimizing.
Advice for newer builders
- Study how the market plumbing works — place order, cancel order, order types, market data feeds, rate limits.
- Trade demo / testnet until the pipeline is boring.
- Then use small real size — minimum meaningful order size, hard loss limits. Tuition, not heroics.
Do not start with five exchanges and twenty indicators. Start with one venue, one data path, one rule, and logs you trust.
Advice for older (experienced) builders
Do not trust indicators as edge.
You see them. We see them. Everyone with a charting app sees them. If your whole system is “RSI + MACD said so,” you are standing in a crowded trade with a pretty dashboard.
Use indicators as helpers, not as the profit engine.
What a “good” strategy means to us
Simple definition:
A good strategy can buy lower and/or sell higher than the market — after costs — over a large enough sample.
Direction (up/down) is probability. You never know when the move shows up with certainty.
So:
- Never, never, never all-in.
- When you find a strategy with EV+ (positive expected value after costs), use the law of large numbers: trade it carefully across as many markets and samples as you can manage, with size that survives variance.
That is how “printing money” actually looks in practice — not one perfect trade, but many small edges, repeated, controlled.
Why we built Nexrade
This is the small product part.
We built the library and the market-data service so we (and people like us) can:
- Trade more markets with lower resource cost
- Reduce complexity and copy-paste when adding a new venue
- Move faster from idea → code → live test
- Keep public data reliable (rate limits, proxies) while private orders stay under your keys
The stack exists to support EV+ systems at scale, not to sell a holy grail.
If the advice above is useful, the tools are optional helpers:
- Market Data API — public exchange data through a hosted proxy
- nexrade — multi-exchange Python trading library
- nexrade-cache — local Redis-compatible cache (including Windows, no WSL)
Build slowly. Size small. Measure after fees. Assume every edge can die.
That is how we still trade — and how we still write bots.
Want to try the path this note describes? Create a free API key, or return to the full list of notes.