The author is not completely wrong. Most basic indicators like MACD, RSI, MA crosses,... are worthless.
Just try to implement them in a "dumb" and emotionless algorithm and backtest them on historical data (code the algo to buy when RSI<20 and sell when RSI>80 for example) : you will get terrible results, whatever the timeframe and the input parameters. Even when they achieve decent (theoretical) profits, it's just curve fitting. Also, when someone tells you to buy because MACD moves this way or Fisher transform is doing that, just ask them what parameters they are using and why they chose these parameters. They won't be able to provide a logical answer. Most of the time they just use the "default" parameters, whatever that means
That being said, there are some good performing indicators (mostly oscillators). Especially when they are improved with numerical tricks, they can yield significant profits.
