Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
Are candlesticks patterns statistically significant?
I agree with this, though the examples given specifically are more observational, unless you plan on entering a long simply because it broke above YHOD.
If I understand what the OP is essentially asking here correctly - are candlestick patterns able to provide a consistent system to make money?
I use 5min candles myself and spent about 2 years studying candlestick price action, and I agree that candles provide a lot of psychological information on what the market is attempting to do, but the key information is missing... namely volume and aggressiveness. So while some candlesticks show reversal signs like hammers for example, I feel it cannot be a consistent system without taking into account the volume on that rejection.
Each candle stick is a parcel of information. It's really an attempt to give you a rough outline of what happened within that time frame made from thousands or millions of data points(tick).
If you unbox that bar, what you get is a sine wave of sorts. Starting from open to peak point (high) trough point(low) close(last).
Given the nature of the market, you have many many frequency waves entangled into the waves of movements. My observation is that, one candle stick pattern is usually not enough to be statistically significant as a result.
Therefore, you have to provide multiple historical inputs in testing that bar. It can be in the form of multi-time frame, or simply boxing previous movements that happened along with the current movement that just formed, in order to find something statistically meaningful. I guess you can call the other inputs "context" of sorts. In my mind, the moving averages, and other indicators derived from price (and/or volume) is an attempt to gauge this "context".
With something like the ES and NQ futures, bullish hammers are more meaningful than bearish hammers. But this also depends on "context".
TL;DR
Markets tend to operate in higher dimensions than 2D.
One bar pattern from one time-frame tend to be useless without supporting information.
Say that you have an objective to describe candlestick patterns. I guess that if you would datamine different patterns, timeframes , instruments you can find statistical significant patterns. Even some that will pass the White's reality test (book Aronson). But would that mean that they will be predictive of the future returns? I guess slicing prices up in candles by time, price or volume does not give a trader some real edge. I think candlestick pattern trading is the stuff where the MM make their money.
I've done a lot of backtesting stats on this, and I find that candlesticks tend to perform differently from other types of signals. The big benefit is that they seem to be more evenly distributed. For instance, if you trade a moving average crossover your signal might get triggered over and over again in a short period of time. So such strategies often have multiple moving averages and filters in an attempt to deal with this issue. Candlestick patterns in contrast tend to be comparatively rare and isolated events.
Unfortunately, this also seems to lead more easily to overfitting. You'll test it over a year and find patterns that appear to be very predictable with outlays of over 60% with a good number of trades, but when you run it on a different year your results are completely different.
Which suggests to me that the signals are just noise, and that people are being fooled by randomness. I would note that many of the traders out there using such "price action" strategies often rely on stops larger than their targets. Such strategies can basically make money (at least in the short term) simply from taking on excess risk, and tell you nothing about how predictive the entry signal is.
I've tested countless setups like this for people, and the conclusion is always the same. They fall apart when a rigorous analysis is applied. Given how much is going on in markets all the time, it seems kind of silly to me that any kind of analysis of past price movements could tell you much of anything. Any real signal would be small, and obfuscated by the noise. Especially in highly efficient instruments like futures.
I don't know any candlesticks pattern trader in RL or OL who is making money consistently, I do know couple of guys who use them for some sort of system condition though, meaning they aren't trading every pattern but are selective and very conditional.
Personally I don't use them but I do think its not completely worthless. Different people click with different things and trade differently. If everyone had the same edge, it wouldn't be edge anymore.
So you can have a strategy that has a stop twice as large as the target, and wins 80% of the time. In such a strategy even if your entry signal was completely random you would have periods of "profitability". You wouldn't see the problems until you run across a losing streak or some sort of tail risk event. Things that if you're lucky you could avoid for a few years. Or to be more technical about it probability distributions apply to datasets with a large number of items, and if it's not a gaussian distribution it can take longer for the true properties of the distribution to emerge.
Retail investors often rely upon such anecdotal evidence. They see that someone else was successful with it, and ignore the fallacies that relying on anecdotal evidence expose you to. They themselves might not know why they were successful. They might have only been successful for that period, and lost it all later. They might have just been lucky.
Which is why when discussing a strategy we focus on empirical evidence. That is evidence that we can look at, and verify for ourselves. If we have a specific strategy and dataset to test against then everyone can run that backtest to verify. They can even try running it against other backtests. If there's any efficacy to such strategies then we should be able to produce empirical evidence that clearly shows the market inefficiency. I continue to look, but I have never seen such evidence using anything that relies solely upon past market generated data. Especially not with price action.
Yeah I can imagine that if you have a macro or other directional view a candlestick pattern can indicate some change in sentiment, and would not completely worthless only we can not prove it using statistical methods.
