The archive
How AI trading tools, signals and automation actually work, what the evidence supports, and how to judge them — newest first.
13 articles
AI compresses the gathering stage of trading research. It does not compress the understanding, and the people who supervise it professionally say so plainly.
6 min read
Both automate. The difference is how the decision is produced: rules a person wrote and compiled, or parameters fitted to data — not who places the order.
4 min read
What an expert advisor is on MetaTrader 5, how it is attached to a chart, how it is built and compiled, and the switch that stops it trading.
3 min read
Two published studies, read in full: measured performance was middling, and every model in a fifteen-model benchmark degraded on data it had not seen.
4 min read
Seven failures that recur with AI trading tools — most of them about how the tool is used rather than how good it is.
3 min read
A scan finds candidates against conditions computed from price. It does not judge them — and the platform documentation draws that line explicitly.
3 min read
What a signal is made of, how to tell a signal from a tip, and the line a regulator draws when following one becomes someone else managing your money.
4 min read
The phrase AI trading tool covers at least three different machines. What each one does with data, what the research shows, and how to tell them apart.
6 min read
Eight questions to put to any AI trading tool, and what a non-answer to each one tells you. No scores, no rankings, no recommendations.
4 min read
An AI assistant summarises, explains and scores. The question that decides how far to trust one is who operates it and what they owe you.
4 min read
ESMA has warned investors that AI tools can produce inaccurate or misleading advice, and notes that giving investment advice in the EU requires authorisation.
3 min read
Measured model performance is middling and degrades on unseen data, and a regulator names overreliance as a risk. None of that proves humans are better.
4 min read
A backtest is a search, not a measurement. What overfitting does to a strategy, and what MetaTrader 5 says its own tester leaves out.
5 min read
