[candlesticks] Do Candlesticks Work?

[guruji 10]

I am new to the world of candlesticks and recently started to read about them. I completed a back-test on Doji as a trend-reversal signal. The results are located at:

 

http://blog.kilotrader.com/2010/05/do-candlestick-patterns-work.html

 

And they are NOT impressive

 

What did I do wrong here? Do candlestick signals work?

[Jumper 10]

Everything works and everything doesn't. Depends on the individual and how they implement it and their risk management. Context is key.

[Kiwi 16]

Context = Support and Resistance

Context = back to key mas

Context = fib retracements

 

otherwise they're amusing noise.

[brownsfan019 17]

I've said it many times before - playing find a shape is a waste of time and energy.

 

Do candlesticks all by themselves work? Probably not.

 

Do candlesticks work in conjunction with other forms of analysis? Sure do.

 

So give up playing find a shape and start to learn how to read a chart and then candlesticks can become a friendly addition.

 

Also, you should note that candlesticks become less reliable the lower the timeframe. When Steve Nison was not a referral whore and actually writing meaningful work, those charts were daily or LONGER. In other words, you will need a lot more than a random doji to signal a reversal point.

 

Good luck, it's not quite as simple as finding shapes.

[TheDude 10]

May be a slight detour but.....

 

I'd be interested in other peoples thoughts on this idea - which BrownsFan kind of eludes to. Time bars/candles patterns will often use the close as part of that pattern. It is certainly an important element in candle shape. A close on a intraday bar is pretty much random in my opinion. Therefore, patterns relying on a close (like candles) are doomed I tell you. Dooooomed. They may have more meaning on a daily bar because this price represents the price people are happy to hold overnight at, have their positions settled against etc. But in an intraday it isnt significant at all - which is why Browns correctly (IMO) states shape hunting is a fools errand.

[guruji 10]

OK... back to the topic of discussion. I completed backtesting the doji, bullish-engulfing pattern, bullish harami and the hammer. These tests were conducted over daily candles. The back test data was SP500 stocks over the last 15 years.

 

I found that over the last 15 years, the doji reversed ~52% of the times and that it is better than the other three patterns

 

Details are available at : Quantifying Technical Analysis: Summary of Candlestick Backtesting

[james_gsx 16]

May be a slight detour but.....

 

I'd be interested in other peoples thoughts on this idea - which BrownsFan kind of eludes to. Time bars/candles patterns will often use the close as part of that pattern. It is certainly an important element in candle shape. A close on a intraday bar is pretty much random in my opinion. Therefore, patterns relying on a close (like candles) are doomed I tell you. Dooooomed. They may have more meaning on a daily bar because this price represents the price people are happy to hold overnight at, have their positions settled against etc. But in an intraday it isnt significant at all - which is why Browns correctly (IMO) states shape hunting is a fools errand.

 

I think I know what you're saying, and I agree to an extent. But to add to what Brownsfan said, it's not really find a shape - even on daily data. In reality, the more you watch charts and candles, bars, etc, you really just get a feel for price movement. Therefore, the doji or single bar is meaningless in and of itself (and like you said, closes seem to be random), but understanding how price reacts around certain levels is very beneficial to some.

 

A better idea, might be to look at what happens around the dojis a program finds. Are there certain patterns, not just technical. Is there increased buying/selling pressure in the tick, etc. There are many ways to analyze a market, some more data intensive than others, but none of it is really wrong or right.

[guruji 10]

I am not sure pattern hunting is meaningless. A pattern represents a market behavior. It is not predictor of what will happen. And because the human brain is better than the computer at spotting patterns, people try to find it.

 

A doji, by definition, represents a Close near the Open. When it happens after a few down days it probably means that the buyers are fighting the downturn and find the price attractive enough to jump in. It MAY mean a price reversal is ready to happen.

 

Also, I agree that looking at one signal in isolation will not work and we need to look at what else is happening. But it is also important to look at one particular signal/pattern in isolation to understand the underlying strength of that particular signal.

 

At Quantifying Technical Analysis: Summary of Candlestick Backtesting, I completed testing four candlesticks patterns and the results are summarized. In the past 15 years, the doji successfully reversed the price trend the following day in 52% of the opportunities. That, I believe is a good representation of the doji candlestick

[brownsfan019 17]

I am new to the world of candlesticks and recently started to read about them. I completed a back-test on Doji as a trend-reversal signal.

 

And they are NOT impressive

 

 

I completed testing four candlesticks patterns and the results are summarized. In the past 15 years, the doji successfully reversed the price trend the following day in 52% of the opportunities. That, I believe is a good representation of the doji candlestick

 

 

So one day you hate doji's and the next you like them? Are you really doing any real work here or just trying to drive traffic to your blog?

[brownsfan019 17]

May be a slight detour but.....

 

I'd be interested in other peoples thoughts on this idea - which BrownsFan kind of eludes to. Time bars/candles patterns will often use the close as part of that pattern. It is certainly an important element in candle shape. A close on a intraday bar is pretty much random in my opinion. Therefore, patterns relying on a close (like candles) are doomed I tell you. Dooooomed. They may have more meaning on a daily bar because this price represents the price people are happy to hold overnight at, have their positions settled against etc. But in an intraday it isnt significant at all - which is why Browns correctly (IMO) states shape hunting is a fools errand.

 

Intraday trading is a unique animal of it's own and I do not believe that candlestick patterns as the only trading mechanism to be of any use. I agree that the close on intraday bars is not of much use.

 

I think I know what you're saying, and I agree to an extent. But to add to what Brownsfan said, it's not really find a shape - even on daily data. In reality, the more you watch charts and candles, bars, etc, you really just get a feel for price movement. Therefore, the doji or single bar is meaningless in and of itself (and like you said, closes seem to be random), but understanding how price reacts around certain levels is very beneficial to some.

 

A better idea, might be to look at what happens around the dojis a program finds. Are there certain patterns, not just technical. Is there increased buying/selling pressure in the tick, etc. There are many ways to analyze a market, some more data intensive than others, but none of it is really wrong or right.

 

Well said! A doji by itself, even on daily or longer timeframe charts if looked at in a vacuum simply symbolizes that on that day, neither bears or bulls won. It does not signify a reversal or continuation when looked at in a vacuum.

 

The OP is only looking at doji's for reversals but I would argue they can also be tremendous continuation signals as well...

 

[guruji 10]

 

So one day you hate doji's and the next you like them? Are you really doing any real work here or just trying to drive traffic to your blog?

 

No, I don't hate or love the doji. I have not yet taken a position on the "goodness" of doji / candlesticks. I am trying to figure this out. Of the 4 signals that I back-tested, the doji after 5 down days is the showing the best next day reversal rates. I don't think it is enough to make a good system.

 

Yesterday I completed a parametric study that looks at the height of the doji candle body in relation to the total length of the candle. Conventional wisdom tells us that for a doji this needs to as small as possible (open=close). However, I was surprised by the results. In fact the larger the (candle body/candle length) ratio, the better my reversal rates were. Results are posted on my blog at : Quantifying Technical Analysis: Doji : A parametric Study. Also, I found no instances of a "true" doji (close=open) after five down days. And if were to reduce this ratio from 15% to 5%, my reversal rates go down from ~52% to ~35%.

 

 

PS : The goal of this post is to get a debate going and to learn of the other people's work. My work is "real" and I am back testing on Wealth-Lab with Sp500 stock for the last 15 years.

Edited May 20, 2010 by guruji

[Kiwi 16]

I learned how "poor" and how good candle patterns could be by writing Sierra Chart paintbars to show them.

 

The point about a pattern is that it is showing something. A pin bar or a BUOB say show price rejection. An inside bar or pair of IBs shows indecision, pause, potential reversal. And the bars that follow confirm etc the pattern.

 

The issue is that although the pattern can be valid in that it shows you a price is rejected ... as you can imagine rejecting a price in the middle of nowhere is quite probably meaningless ... a temporary overextension returning to value say with little chance of profiting much from it. But a rejection of "key" prices might be meaningful.

 

So testing a pattern in isolation of context (see my last post) is likely to be an unsatisfying experience for the tester.

[brownsfan019 17]

As Kiwi said, everything is in context. To simply tell the program to find a shape may or may not work but I would venture a guess that the user will be disappointed.

 

Trading in it's simplest form is buying support and selling resistance, however you define support and resistance. Once you define your s/r then see how doji's or whatever formation you want looks at your s/r levels.

[stpips89 10]

I have found that Nial Fuller's usage of candlesticks is very clear and straightforward.

[BlowFish 10]

Several random things occur to me (that's the trouble with hangovers).

 

They are not that effective at illumination though will do at a pinch if the power is out. They do have a rather nice ambience however. The missus likes them she has about as many as the Vatican.

 

What do you mean by 'work'. (That was a serious question) seems to me you are expecting something from them that they probably do not 'work' very well at. That does not mean that they might not work quite well at something else. This seems pretty important to establish if you want your question answered. Depending on exactly what you mean by 'work' I think the short answer may very well be no.

 

Seems to me they are eminently more suitable as triggers rather than signals?

 

One good thing about doing things programatically is you have to clearly define what they are (if the days close is a point away from the open is it still a doji for example, are long legged varieties special cases etc.)? What it does not do is to really make you think about what they really are. You don't want to skip that bit.

 

Just because something does not 'work' (depending on what you mean by work) does not mean that you can not glean useful information from the test.

 

Seems to me that the people that get value from them are not bogged down in candlestick dogma but simply use them as a visual aid to help them look at what is behind them.

[guruji 10]

I am trying to find out if candlestick patterns are any good at picking out short term trend-reversals.

 

Based on what I have done so far, they can pick next day trend reversal with a success rate of 45-52%

[Tams 20]

if you use MultiCharts/TradeStation (or EasyLanguage compatible program) ...

here's a program to help your candle stick analysis

 

http://www.traderslaboratory.com/forums/f46/candle-pattern-5736.html

 

 

89.68

[BlowFish 10]

I am trying to find out if candlestick patterns are any good at picking out short term trend-reversals.

 

Based on what I have done so far, they can pick next day trend reversal with a success rate of 45-52%

 

45% chance of identifying the start of a trend sounds all right to me. I am pretty sure one could make that 'work' in actual trading.

 

If that really is the extent of your question you have answered it right there. There really is not much more to discuss. Of course there are far more interesting questions one could ask (many have been eluded to) e.g. how can I use context to improve these figures, can one trade these reversals successfully? how? etc. etc.

[Kiwi 16]

stpips.

 

2 out of 3 of your posts have been promotions for Nail Fulcher's site (don't go there newbies ... there are better places to waste your money).

 

Stop posting about him or you will be banned. Also, feel free NOT to correct the spelling ... that would be a further spamming of Traders Lab.

[wrbtrader 10]

I am trying to find out if candlestick patterns are any good at picking out short term trend-reversals.

 

Based on what I have done so far, they can pick next day trend reversal with a success rate of 45-52%

 

Hi guruji,

 

I've been studying Japanese Candlesticks since Nison and many others released their first books on the topic along with coding, backtesting, trading, writing topics about anything involving them...many know my former user name as NihabaAshi.

 

I have never met a profitable trader that uses Japanese Candlesticks all by itself and nothing else.

 

Therefore, I'm always scratching my head why traders waste valuable time in testing candlestick patterns as such (all by themselves). In contrast, profitable traders that use candlestick patterns are using them secondary or merged to a primary trading system (e.g. s/r analysis with candlestick analysis, intermarket analysis with candlestick analysis).

 

Also, I can tell via your replies or lack of reply when traders advice to use them in context...you continue mentioning testing them out of context as if within context is not important. Therefore, I will state what others have tried to tell you already...you must understand the price action prior to the appearance of any candlestick pattern signal to get within context. That means you should have already determined if the price action is bullish, bearish, price noise or whatever prior to your defined candlestick pattern...then only using the candlestick pattern to open a position when the reward is in your favor.

 

Your backtesting rules at your blog are not within context. However, that's just one problem you have and there are others. For example, base upon what you show in the price action of your testing rules, you don't understand the doji via stating at your blog that conventional wisdom states that its a trend reversal signal. It is not. However, if you meant to say that the interval(s) after the doji can sometimes involve the doji as part of another candlestick reversal patterns...that's true.

 

* The type of intervals that occur prior to the doji will often determine the strength of the indecisiveness of a doji and indecision is never a trade signal.

 

* Trade management after entry has a great impact on the performance results. I've done extensive testing in this particular area and have shown that traders taking the exact same candlestick patterns in the same trading instrument at the exact same time can have results vary from 44% - 73% depending upon the trade management rules. Those that had higher (better) trade results had a stronger understanding of the price action in comparison to those that did not.

 

Simply, you're going to have to make a decision...continue testing candlestick patterns the way most losing traders use them or test them the way profitable traders use them. If you decide on the latter...you're going to need to learn a lot more than just candlestick patterns...involving more time, more work and less simplicity. However, if you decide on the former (continue testing them as stand alone method)...at least understand the candlestick pattern you're testing and its sub-groups because your blog information implies you're still thinking inside the box (conventional wisdom).

 

By the way, I stopped coding a long time ago but it was useful in that I was able to discover many variations or sub-groups of each candlestick pattern (e.g. there are 15 different types of bullish white hammer patterns and most are not reliable except a few)...reliability determined within context of the overall price action. Therefore, your testing only reveals what some of us veteran traders already know...don't test them the way losing traders apply them.

Edited May 24, 2010 by wrbtrader

[windsurfer 10]

Thank you, guruji.

 

You seem to be criticized for what you did not do (e.g. testing in "context"). However, I appreciate what you did do- test dojis and 3 other patterns in isolation. I recall another study many years ago that came to the same conclusion- that candlesticks in isolation give no trading advantage. This is very useful information. Thank you for sharing.

[guruji 10]

Thanks for your comments and taking the time to go thru what I did. Your comments are valuable and will probably guide me forward. A few comments :

 

 

- My goal is not build a trading system based on candlestick based trading system but understand if they mean some thing. I think this exercise is valuable to build a good and robust trading system. There are infinite number of primary/secondary signals that people use. To test each combination of these signals would be a huge effort.

 

- I am using a doji after 5 down days as a trend reversal pattern and not in isolation

 

- Yes, trade management is key. My success rates change quite a lot if I allow my trades to last longer than the next day. I can even make them profitable. But I would like to find a primary signal that gives me 65-70% success rate before I go into optimizing trade management. (may be I won't get there... or... i will)

 

 

Hi guruji,

 

I've been studying Japanese Candlesticks since Nison and many others released their first books on the topic along with coding, backtesting, trading, writing topics about anything involving them...many know my former user name as NihabaAshi.

 

I have never met a profitable trader that uses Japanese Candlesticks all by itself and nothing else.

 

Therefore, I'm always scratching my head why traders waste valuable time in testing candlestick patterns as such (all by themselves). In contrast, profitable traders that use candlestick patterns are using them secondary or merged to a primary trading system (e.g. s/r analysis with candlestick analysis, intermarket analysis with candlestick analysis).

 

Also, I can tell via your replies or lack of reply when traders advice to use them in context...you continue mentioning testing them out of context as if within context is not important. Therefore, I will state what others have tried to tell you already...you must understand the price action prior to the appearance of any candlestick pattern signal to get within context. That means you should have already determined if the price action is bullish, bearish, price noise or whatever prior to your defined candlestick pattern...then only using the candlestick pattern to open a position when the reward is in your favor.

 

Your backtesting rules at your blog are not within context. However, that's just one problem you have and there are others. For example, base upon what you show in the price action of your testing rules, you don't understand the doji via stating at your blog that conventional wisdom states that its a trend reversal signal. It is not. However, if you meant to say that the interval(s) after the doji can sometimes involve the doji as part of another candlestick reversal patterns...that's true.

 

* The type of intervals that occur prior to the doji will often determine the strength of the indecisiveness of a doji and indecision is never a trade signal.

 

* Trade management after entry has a great impact on the performance results. I've done extensive testing in this particular area and have shown that traders taking the exact same candlestick patterns in the same trading instrument at the exact same time can have results vary from 44% - 73% depending upon the trade management rules. Those that had higher (better) trade results had a stronger understanding of the price action in comparison to those that did not.

 

Simply, you're going to have to make a decision...continue testing candlestick patterns the way most losing traders use them or test them the way profitable traders use them. If you decide on the latter...you're going to need to learn a lot more than just candlestick patterns...involving more time, more work and less simplicity. However, if you decide on the former (continue testing them as stand alone method)...at least understand the candlestick pattern you're testing and its sub-groups because your blog information implies you're still thinking inside the box (conventional wisdom).

 

By the way, I stopped coding a long time ago but it was useful in that I was able to discover many variations or sub-groups of each candlestick pattern (e.g. there are 15 different types of bullish white hammer patterns and most are not reliable except a few)...reliability determined within context of the overall price action. Therefore, your testing only reveals what some of us veteran traders already know...don't test them the way losing traders apply them.

[estate1997 10]

I heard a lot of you say that intraday bars are of no use. Set your chart to Kase 8.0 and your Keltner channel to 4 and .55. Does it tell you anything?

Workyboy

[RichardTodd 10]

I heard a lot of you say that intraday bars are of no use. Set your chart to Kase 8.0 and your Keltner channel to 4 and .55. Does it tell you anything?

 

It told me: "Be sure to drink your Ovaltine." Kind of a let-down. Maybe my chart colors aren't the profitable ones, or something.

[wrbtrader 10]

Thanks for your comments and taking the time to go thru what I did. Your comments are valuable and will probably guide me forward. A few comments :

 

 

- My goal is not build a trading system based on candlestick based trading system but understand if they mean some thing. I think this exercise is valuable to build a good and robust trading system. There are infinite number of primary/secondary signals that people use. To test each combination of these signals would be a huge effort.

 

- I am using a doji after 5 down days as a trend reversal pattern and not in isolation

 

- Yes, trade management is key. My success rates change quite a lot if I allow my trades to last longer than the next day. I can even make them profitable. But I would like to find a primary signal that gives me 65-70% success rate before I go into optimizing trade management. (may be I won't get there... or... i will)

 

Hi,

 

Thanks in replying.

 

There are hundreds of documented testing performed on candlestick patterns in isolation and your blog results is just another one. You're doing nothing new in comparison to all the prior blogs, websites, magazine articles, forum discussions et cetera. However, that's not a concern because I do understand that you're trying to see for yourself via your own data instead of listening to others that have done the same before you.

 

There are infinite number of primary/secondary signals that people use. To test each combination of these signals would be a huge effort.

 

That's correct and you don't need to test them all because that's impossible...just start with one. Like I said to you in my prior reply...you're going to need to sit down and make a decision. You can continue testing candlestick patterns the way losing traders use them or test them the way profitable traders use them. Choosing the former instead of the latter as a valuable exercise to build a good and robust trading system doesn't have merit. That's ok for simplicity (computer codes) or discussion but not ok as an exercise to building a robust trade method. A robust trade method involves a complete trading plan and candlestick patterns is not the only chapter in a profitable trader's book.

 

I know...I've stood in the past where you're at right now via trying to approach trading via pure science (computer codes). However, if you're going to be a profitable trader...I recommend you make the effort and concentrate your efforts on testing candlestick patterns the way profitable traders use them instead of the way losing traders using them unless the purpose of you're testing is to show how losing traders are using Japanese Candlesticks.

 

I am using a doji after 5 down days as a trend reversal pattern and not in isolation

 

That price action definition is not within context. Here's a example (hypothetical) of a candlestick pattern within context (understanding the price action). The S&P 500 breaks below a key support area and begins making lower highs with failed counter-thrusts in between those lower highs along with other key markets doing the same. In addition, the industry group itself involving your selected stock is getting hit with negative news and then 5 consecutive down days appear. This is an example of within context because you understand the price action, able to explain it. Thus, to be looking at bullish candlestick patterns via a code only...you'll be ignoring the "within context" information I just revealed...you're testing candlestick patterns in isolation. (I have stood in the past where you are at right now...I know exactly what you're trying to do).:hmmmm:

 

Further, if you truly understood the price action and are not using candlesticks in isolation via some analytical computer code...you should have designed a code to look for and test the reliability of dojis as a bearish continuation signal after 5 consecutive down days and then compare the results to using dojis as a bullish reversal signals (I'm going to give a little via pretending that a doji is a reversal signal even though it's an indecision price action)...thinking outside the box.

 

Reminder - each candlestick pattern has different sub-groups.

 

Here's a real example of within context via your GNW Financial of October 2008 chart at your blog instead of my hypothetical explanation above. Your code involves a handful of intervals only and your posted chart of GNW is a financial stock during a time period everybody was running scared in the overall financial crisis along with the fact that GNW had fallen from $25 price range to $3.50 price range...your code explanation with zoom-in chart is an example of using candlestick pattern in isolation while ignoring the facts I just explained (that within context information). Therefore, you're trying to let a code define the price action instead of you understanding the price action prior to the appearance of any candlestick pattern. Should you been looking or testing for bearish continuation signals or bullish reversal signals or both (thinking outside the box)...comparing the reliability information of each for "within context" situation I just explained about GNW ?

 

Lets now talk about the AIV chart at your blog because it represents a Bullish Hammer "continuation" pattern. Instead, you in error called it a Bullish Hammer "reversal" pattern. There are several subgroups of hammer patterns called continuation hammers and reversal hammers. Your AIV chart falls under the continuation sub-group category that can be broken down further into different variations because I took a closer look at AIV chart via my own data provider with many months of price action in comparison to you zoom-in chart of a few intervals in isolation. AIV is clearly in an uptrend with obvious support levels when a small retracement appears involving that hammer line. Had that hammer line appeared below any key support areas along with the uptrend having a key change in supply/demand...that's when you can designate a hammer pattern as a "reversal" instead of as a "continuation" (this is something you aren't going to read in any books nor learn from codes of losing traders trying to build a robust system).

 

Like I said, there are problems with your candlestick pattern identification but that's ok because you did say you are new at this and you're obviously approaching this via codes only (in isolation) as seen by those that have done in the past what you're currently doing. Therefore, I think a good exercise for you and the readers of your blog is for you to show zoom-out charts that show many prior months of price action before your candlestick patterns along with those zoom-in charts that only show a handful of intervals in isolation without any "within context" information. The zoom-out charts will encourage your readers to seek on their own the "within context" information that may agree or disagree with your analysis.

 

Mark

Edited May 27, 2010 by wrbtrader