Evaluating Technical Analysis with Realistic Trading Constraints
Summary
The document asks whether historical evidence can show how profitable technical analysis is in practice. It calls for evaluation across multiple years and market conditions, with returns and loss sizes considered rather than signal hit rates alone. It also highlights trading costs and capital constraints, including the effect of limited account size, risk of ruin, and markets that close overnight.
The author is concerned that isolated successful chart examples and vague claims about extra profit do not establish usable performance. Observation bias is identified as another challenge. The text does not provide a study, strategy rules, data, performance estimates, or a method for resolving these issues; it is a request for quantitative evidence and a checklist of factors that such evidence should address. It also raises the distinction between theoretical model performance and results accessible to an individual trader, without establishing how large that gap is.
Key ideas
- A technical-analysis strategy should be evaluated over multiple years and different market conditions.
- Hit rate alone is inadequate without comparing the size of gains and losses.
- Trading costs and limited capital can materially affect practical returns.
- Risk of ruin and overnight market closures are relevant implementation constraints.
- Cherry-picked examples and unqualified profit claims do not establish robust evidence.
Tags
Full text
# Is there a proper analysis of how profitable Technical Analysis is? # Is there a proper analysis of how profitable Technical Analysis is? I often get spammed with tips about hammers, fibonacci sequences, ceilings etc. Usually they come with a single example of where it worked great. Given how many people keep focus on this I suspect there is more to this than just a cherry picked example, but I wondered if there is any indication of how much more. Key factors that would ideally all be considered at once (though finding something on even one is already hard): - Historical performance over several years, sliced by various market conditions - Moving beyond likelihood (it is great if something works 80% of the time but perhaps the losses are 5x bigger than the gains in bad scenarios) - Costs of trading, thinking about cost of transactions mostly but perhaps alsoncapital. (No need to account for costs of applying the methodology or time used.) Stretch: Tax, spending or other annual effects 4.Impact of portfolio mgmt constraints (let's assume we have 10k, or even 100k usd but are not able or willing to put in millions), if you lose the farm once you are done so probably everything you analyze has 20x less impact (and somewhat higher costs) than it could as you spread the risk. Another constraint may be that markets close overnight. And of course this leaves the usual challenges such as observation bias. Hence the question: ## Is there any quantitive historical analysis that somehow can translate to real world impact? What have I found: - It is suggested advanced models can perform quite well. Presumably there are practical constraints such as speed or change to make these less practically accessible in daily life. (Not looking for recommendations to subscribe to your solution for a small fee...) - A broad search on actual impact mostly results in claims regarding the likelihood of individual decisions being right, occasionally something more relevant like 'we achieve a few % extra profit', but always these statements are unnacked and unqualified (few % of what, annual RoE?) - I have seen some successful traders buy nice cars, but not airplanes, so even if it goes well it seems an exponential return on capital is not very common.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.