Hacker News Reader

Comment by valdiorn | Hacker News Reader

there are two types of technical analysis, subjective and objective.

subjective ones, like "head and shoulders" patterns, are totall bollocks. Objective ones, like MACD, EMA, RSI, etc... can be useful when applied correctly.

Unfortunately it seems like the author of the paper was working with some very naive use cases, causing him to not get very good results (I will admit, I only skimmed the paper as it didn't seem that useful to me).

I can tell you from first hand experience that technical analysis trading strategies, when applied correctly, are very, very, very effective.

View on HN

> I can tell you from first hand experience that technical analysis trading strategies, when applied correctly, are very, very, very effective.

I don't doubt that you have traded to some large financial advantage, but it's unlikely that your specific experience is systemically repeatable. In particular in justifying technical analysis with the qualifying phrase 'when applied correctly' smells a little like 'No true Scotsman'.

>Objective ones, like MACD, EMA, RSI, etc... can be useful when applied correctly.

True, they can serve as useful tools for describing market conditions. If I want to tell a computer to do something when the market is falling rapidly, then I might define "falling rapidly" as "price below the 15-second EMA" for example.

But do they have any predictive power? No.

>I can tell you from first hand experience that technical analysis trading strategies, when applied correctly, are very, very, very effective.

I doubt it. I've never met a professional trader (i.e. at a reputable firm) who uses them, and they'd be laughed out of the industry if they did - at least in all the places I've worked. I imagine the only people who 'use them successfully' actually derive their edge from elsewhere (institutional advantages, etc.) and just don't realise it.

I think it's possible they have both short-term predictive power, and for it not to be profitable to trade the signals they generate. For example, something like "price below the 15-second EMA" might mean in the short-term there is large probability of the price falling, and in the small possibility it reverse, it spikes a lot higher, making it not profitable to take a directional bet on the price. This can make those signals good for investors to decide whether to invest in a rising bull market with low yield and a tail risk of large negative return or a falling bear market with high yield and a tail risk of large positive return. There may be no objective profit, while allowing "subjective" profit in utility by matching investor's risk profile. It does sound like gd1 has more experience than I do, though, so don't trade off my perspective.

> But do they have any predictive power? No.

How would you explain the performance of algorithms like these if technical indicators didn't have any predictive power?

https://www.quantopian.com/leaderboard/5550e55f09901773d9000...

Drawdown for the S&P500 is ~7% for the backtest period, and ~0.75% for the paper trading period, while performance is ~20% for both periods. These are automated algorithms, so the only thing they can base trade decisions on are objective, technical signals.

Judging from the dramatically lower drawdown and higher performance, it'd be reasonable to assume that the trade decisions made by these algorithms are correct far more often than not, which means technical indicators must have some degree of predictive power.

I agree with you on the fact that they can't possibly be used properly by a human trader though, even when assisted with sophisticated charting tools. But this is because human traders will have trouble objectively assessing the performance of their strategies over a long period of time, and because they lack the absolute precision and discipline required to effectively apply their strategies in the first place.

Reply on news.ycombinator.com