The illlusion of candlestick patterns

[nunrgguy]

I've been doing some demo FX trading on simulation lately and the empirical results have been interesting to say the least w.r.t. various candle formations. I haven't filled out a spreadsheet but I've gone through well over 100 hundred trades with this, which is a lot on the time frames I've been looking at.

In the main I've been testing S/R, trend lines - either drawn or MA based on weeklies and dailies. These time frames are not necessarily what I plan to trade but these formations are supposed to be more reliable so seem the logical place to start. Most of the trades I've been looking at are trend retracements/continuations and breakouts.

Everywhere you read about certain bars meaning certain things (I understand logically and psychologically what they signify) and that you should always wait for the bar to close for confirmation. This always looks great on a chart in hindsight.

But:idea: hold on...these pictures are just illusions, they are not price, there's more going on than the bar tells us. What I've found is that while there may be a slight edge with some of the formations you are in a lot of cases also increasing your risk.

Overall with my testing I've found that by following known formations I ended up with a loss, by ignoring them and just playing the levels I ended up in profit.

Increasing risk how so? Take an outside bar bouncing off support. By the time the bar has closed a lot of the move may have already happened. Lets assume for one wild moment that we may actually have NO idea where price is going to go from here because...we don't. We might be expecting it to do something because of various texts we have read but we do not KNOW.

So we place our order north of the bar and our stop just south of it, seems reasonable and logical. Most times this gives us quite a large stop, we reduce position size accordingly. Price will then either take off north (great), take off, stall and come back, either taking out a b/e stop, retracing up the candle and then continuing back north, or taking out the original stop.

What I've found (but will need to do further work on) is that if other factors are in place (especially a nice juicy round number) it seems more reliable to just play the level itself placing the order (if going long) above the level and the stop below. This way the initial stop can be placed further away from S/R for the same size stop as before, or closer to. But without our bar how do we know where to put the stop? Oh that's forever the question isn't it...

Reliability of the entry is further increased by allowing price to retrace past where the entry will be placed (maybe even our initial thoughts on stop and then place the actual stop just below this), place the order and then wait for price to come back, moving in the right direction - essentially a rejection of S/R just the same as an outside bar/pin bar etc but not looking or waiting for a particular bar shape, just watching which way price is moving.

I'm sure looking into more detail on the shorter time frames at these levels will throw more light on these trades but it seems that looking for exact patterns does a couple of things:
1. It greatly reduces potential trades -on the weeklies for example you are going to be waiting a hell of a long time for a pin bar at a level:
2. It gets us in late
3. It can mean that we have a large stop with either increased risk/smaller position size
4. It doesn't seem to increase success rate.

Am I away with the fairies here?

 

[virtuos0]

I've been doing some demo FX trading on simulation lately and the empirical results have been interesting to say the least w.r.t. various candle formations. I haven't filled out a spreadsheet but I've gone through well over 100 hundred trades with this, which is a lot on the time frames I've been looking at.

In the main I've been testing S/R, trend lines - either drawn or MA based on weeklies and dailies. These time frames are not necessarily what I plan to trade but these formations are supposed to be more reliable so seem the logical place to start. Most of the trades I've been looking at are trend retracements and breakouts.

Everywhere you read about certain bars meaning certain things (I understand logically and psychologically what they signify) and that you should always wait for the bar to close for confirmation. This always looks great on a chart in hindsight.

But:idea: hold on...these pictures are just illusions, they are not price, there's more going on than the bar tells us. What I've found is that while there may be a slight edge with some of the formations you are in a lot of cases also increasing your risk.

Overall with my testing I've found that by following known formations I ended up with a loss, by ignoring them and just playing the levels I ended up in profit.

Increasing risk how so? Take an outside bar bouncing off support. By the time the bar has closed a lot of the move may have already happened. Lets assume for one wild moment that we may actually have NO idea where price is going to go from here because...we don't. We might be expecting it to do something because of various texts we have read but we do not KNOW.

So we place our order north of the bar and our stop just south of it, seems reasonable and logical. Most times this gives us quite a large stop, we reduce position size accordingly. Price will then either take off north (great), take off, stall and come back, either taking out a b/e stop, retracing up the candle and then continuing back north, or taking out the original stop.

What I've found (but will need to do further work on) is that if other factors are in place (especially a nice juicy round number) it seems more reliable to just play the level itself placing the order (if going long) above the level and the stop below. This way the initial stop can be placed further away from S/R for the same size stop as before, or closer to. But without our bar how do we know where to put the stop? Oh that's forever the question isn't it...

Reliability of the entry is further increased by allowing price to retrace past where the entry will be placed (maybe even our initial thoughts on stop and then place the actual stop just below this), place the order and then wait for price to come back, moving in the right direction - essentially a rejection of S/R just the same as an outside bar/pin bar etc but not looking or waiting for a particular bar shape, just watching which way price is moving.

I'm sure looking into more detail on the shorter time frames at these levels will throw more light on these trades but it seems that looking for exact patterns does a couple of things:
1. It greatly reduces potential trades -on the weeklies for example you are going to be waiting a hell of a long time for a pin bar at a level:
2. It gets us in late
3. It can mean that we have a large stop with either increased risk/smaller position size
4. It doesn't seem to increase success rate.

Am I away with the fairies here?

Well I'll be damned, someone finally twigged it. Good job. Candlesticks are utterly meaningless, we break time into random segments, but price doesn't recognise these segments. If instead of 1hr, 4hr, 1day charts we had 47minute, 263minute, and 1.5day charts the exact same price sequence would be represented by completely different candles.

If you trade candles, you're trading randomly. The only reason why it might work would be if a candle also coincides with S+R. And there is no self-fulfilling prophecy about candles either because no one that moves the market gives a flying **** about inside bars etc. I used my copy of Nison's book as bog-roll long ago.

 

[montmorencyt2w]

Nunrgguy,

You might enjoy the book "Technical Analysis and the Active Trader" by Gary Norden.

As well as the standard patterns, he has doubts over S&R as well.

As a matter of interest, your profile says you have been trading 3-5 years and I wondered what methods you had been using (and with what success, if I may be so bold).

I must admit, I've had my doubts about the standard patterns for some time, although I'm not quite ready to throw out the bathwater completely, in case there is a living baby still in there somewhere. I see so many moves that the patterns simply can't explain, or don't seem to as far as I can see. It's very easy sometimes to construct patterns with hindsight. The human eye and brain is particularly good at this. Sometimes it works to our advantage; at other times it can lead us astray.

 

[GladiatorX]

Perfect Patterns and why they shouldn't work
If a triangle is obvious enough that you can see it; the rest of the world and all its traders now have the same information as you; of the existance of a triangle ...
Now if you aren't one of the smart guys who is already in and are waiting on a breakout - On the daily close - Everyone and their recently deceased dog knows that there is a triangle and its broken out... After you enter on the breakout -

Who? Is left to BUY more... Where is the future demand?... Everyone saw it... Anyone who gives a **** is already in, what possible future demand could exist ?

Now what does this tell us; a triangle breakout that is pretty creates a simple 50/50 probability situation (as usual) of price going up or down, if anything its more likely to fall; because there will the proffesionals who entered in the consolidation will exit for profits, then the buyers of the breakouts will sell = Supply.

Basically - When you are entering on a breakout of a triangle... You are entering at a place that EVERYONE in the world knows is a 'breakout', so who else is there to buy; Except traders who trade on OTHER criteria anyway; In which case you haven't done any better in assessing the current situation.

Boom, Bang, Bingo.

Same with pin-bars, same with everything that YOU can see. Unless you have better sight than all the other traders.

Lets talk about Pin-bars significantly; If they form @ support; It means (possibly) that players of the markets percieved the support as a potential bouncing area (Now remember most players are wrong and lose money) and therefore there limit orders were filled and stop orders were filled - This demand, fuels price so that it forms a pin-bar... Now once it closes... Thats it. The orders are done... The support has been tested and there were orders there that have now been executed to the extent of moving price up a little bit to form a pin...
Now what logic is there that a long is now a good position to play? Its back to a scenario of not knowing future demand...
You are entering AFTER all the orders... Your the last in; All the players SAW the level and bet everyone else sees it and its a self-fulfilling prosphosy, and right enough price bounced, as they predicted and they have a nice profit; Now you are last in... Your the last guy to play that support - there is no more SUPPORT; the support has already enter; so once again, like the triangle - You are entering at a random, insigificant place.

(Y) I hope i'm making sense.

Historical prices then only allow you to see the patterns... Thats when they are obvious; After a trend... When we are choppy - You cannot see the hundreds of flags, triangles and trendlines that are broken... But you can when occasionally a triangle breakout works (Nothing 2 do with the triangle existance) and the OH great that looks a good strategy...

The way to make money is to continually fade the idea that 'the market is doing something'; Everytime you think 'Here we go, big trend' you'll be wrong 9/10 - Because when you trade, you always think something will happen as does when you look at historical charts and only notice the big movements... Most of the time price is just choppy; Making enough losers until the trend can start with as few people in the trade as possible - Thats the ONLY way that FUTURE DEMAND could exist; If everyone is long and the market starts to go up; Who is there left to buy? So the market has to knock all these traders out for a while and during this, 20 false breakouts, 5 triangles and 6 pin-bars will fail, when everyone is out, it'll continue...

Now thats enough for one day; My puppy is crying.

 

[nunrgguy]

Nunrgguy,

You might enjoy the book "Technical Analysis and the Active Trader" by Gary Norden.

As well as the standard patterns, he has doubts over S&R as well.

As a matter of interest, your profile says you have been trading 3-5 years and I wondered what methods you had been using (and with what success, if I may be so bold).

I must admit, I've had my doubts about the standard patterns for some time, although I'm not quite ready to throw out the bathwater completely, in case there is a living baby still in there somewhere. I see so many moves that the patterns simply can't explain, or don't seem to as far as I can see. It's very easy sometimes to construct patterns with hindsight. The human eye and brain is particularly good at this. Sometimes it works to our advantage; at other times it can lead us astray.

Thanks both. I'll definitely check the book out.

I've actually been involved with the markets a lot longer than the 3-5 years but I initially started out as probably most do investing with a buy and hold approach, initially with funds, a little later with equities. Choice of equities came down to a stock screen I adapted from somewhere else and basic trend following (line chart). In the bull run from about 2003 to 2007 it made money. In the meantime I also started spread betting with small accounts - again in equities and a positional trend trading style so I'm well used to walking away from a trade for weeks/ months and not caring about it too much. I knew nothing about candles etc and just followed trend and for a good while it worked OK but the %gains weren't really worth the risk to equity as I was, as I know now, trading with too much leverage. A couple of big losses would have wiped the account out.. Then did come some drawdown and I realised I really didn't know what I was doing and was just gambling. Then came a LOT of reading and studying, usually from pdf when I was supposed to be at work so my eyes aren't too good these days.
I then started trying out some of these ideas and did worse than I was just doing something VERY basic.

After this I started looking into FX rather than equities because with small accounts SBing the spread is better, liquidity is better and you're less likely to get slipped - everyone knows this but this was all very new to me at the time. I tell you when I first looked at T2W about 5 years ago I didn't understand a word of it.

Later I built myself an Access application for a trade journal/logging, money management and forward and back testing of results (manual trading - you just plug the results into the app). So then I needed the figures to go into the app. You need positive expectancy to know that the drawdown you're going through is to be expected and that you're not just blowing your account right

This is when the fun started really as I've found with much testing that over the long haul there's so much out there that's read as gospel that sometimes works, sometimes doesn't etc. (but with the doesn't outweighing the does) I've not even really got to the stage of plugging the numbers into the app as I've just been testing things out on an ad-hoc basis. On sim you can get through a lot of entry styles etc very quickly....and get more and more confused in the process. But it's bettter than losing the house over a couple of years.

It has all lead ultimately to studying PA setups S/R etc but I don't just blindly follow or believe any received wisdom - I want to KNOW for myself whether this stuff has any chance of succeeding. Too often you read things in books and on forums which initially look great but when you look into it further turn out to be not so great. Probability too often seems to get skipped over in a lot of these texts. On any day you can find a new forum thread where someone shows their method of tradin: "Hey everyone I've won a few times this last fortnight, this is how I did it), people try it out...it works a few times, they report back , great fantastic etc. Then the thread dies....because lo and behold it doesn't work.

Then you see the people who are 'gurus' on one forum who on another forum were perhaps only a few months ago saying they were losing...

Some may say 'you've been looking at this for how long exactly??? Give up now because you'll never trade/never do it but I think too often people just start trading this stuff without doing their homework and then wonder why they run into psychological issues. How can you have the confidence to know to leave that stop WELL alone unless you have the data to back it up. Of course that data will never be 100% because it's based on past data but it's better than going in blind with no idea whatsoever and some pictures off the internet. Then, if the trade fails, it fails, if it wins it wins but you KNOW that overall your system should deliver a profit. I don't think you can trade without emotion without this knowledge.

I can sit here all week on sim testing methods with great rapidity with no emotion whatsoever involved so if something doesn't work under those circumstances it sure as hell isn't going to work when the 'should I shouldn't I, oh no it's gone the wrong way, Ii'll wait and see if it comes back' monster rears its ugly head.

Am I on the search for the Holy Grail? No I don't think so, I'm just trying to develop a method that I can trade, and the important bit is the I, and that I know will generate profit over time

 

[arabianights]

**** me talk about tldr

 

[nunrgguy]

**** me talk about tldr

Hey the guy asked a question, I condensed years into a few paras

 

[arabianights]

lol don't worry about it mate I'm just a bit tipsy

 

[montmorencyt2w]

Hey the guy asked a question, I condensed years into a few paras

Thanks for your very frank reply nunrgguy. Wasn't expecting so much actually I hope it was cathartic

Before I got into trading (or spread betting I should say before someone corrects me), I was getting into funds a bit, and it was going to make me rich, ho ho (this was in the height of the bull market, ho ho). I'm still interested in funds actually (and have money in some which certainly aren't making me rich...), and will give them more attention hopefully in the future, maybe transferring some knowledge from "trading" (s.betting) across, depending on how it works out

Edit: p.s. Don't take any notice of arabian's jibe - he's just a very naughty boy!

 

[nunrgguy]

Thanks for your very frank reply nunrgguy. Wasn't expecting so much actually I hope it was cathartic

Before I got into trading (or spread betting I should say before someone corrects me), I was getting into funds a bit, and it was going to make me rich, ho ho (this was in the height of the bull market, ho ho). I'm still interested in funds actually (and have money in some which certainly aren't making me rich...), and will give them more attention hopefully in the future, maybe transferring some knowledge from "trading" (s.betting) across, depending on how it works out

Edit: p.s. Don't take any notice of arabian's jibe - he's just a very naughty boy!

Oh I know, I took it in jest anyway. Ask me a question and, as I rarely get asked, you're bound to get a rather full and frank answer...

See?

 

[dcraig1]

Congrats to nunrgguy for honesty and willingness to face some of the inconvenient truths. And for doing some proper work in trying to get a handle on the markets.

IMHO, you are quite right about candlesticks - largely a waste of time. As pointed out, they are based on quite arbitary construction of time series for which the the only possible justification is bland statements to the effect such as "5 min charts are the industry standard" or they (substitute your TA method of choice here) work because "a lot of traders look at them". It would be interesting if somebody did a study of candle patterns in time series constructed from constant tick bars or constant volume bars. At least that would have the virtue of reflecting in some way the pace at which the market is moving. Not viable of course for fx, because it is not exchange traded in the normal sense.

Anyway to return to candles. One of the pieces of software that I have developed is a good stock scanner that handles technicals and fundamentals. And I do mean good - much better than all the crap out on the web. There is a backtesting facility in the sense of being able to scan on specific date, take the results of the query and chart an unweighted index of the selected stocks against a benchmark and do a performance comparison. Doing it has been a great education - far better than reading any number of books on trading.

The ability to scan for candle patterns is one of the features. Using the backtesting tools mentioned above, I have never been able to find any edge with common candle patterns. They just do not "work". Not only that, I have also looked at the count of the number of stocks exhibiting bullish or bearish candle patterns at any given time to see if that provided a tradeable indication of broad market direction. It should really come as no surprise that the horse had already bolted. As candle patterns are purported to "work" on any financial instrument, I consider testing on thousands of stocks to be a much better test than staring at a few historical charts of eg EUR.

The news it not all bad though - it is possible (though not easy) to select stocks on both long and short sides that are market beating. In the end it all comes down to observation and trying to reason through what you observe.

Not being fully satisfied with my screening, I also wrote some code to mark up the most common candle patterns on charts. Sometimes the human eye can pick things up that are hard to formalize. Attached is such a chart of SPY. Does this mess really mean anything?

 

[nunrgguy]

Thanks for your very frank reply nunrgguy. Wasn't expecting so much actually I hope it was cathartic

Before I got into trading (or spread betting I should say before someone corrects me), I was getting into funds a bit, and it was going to make me rich, ho ho (this was in the height of the bull market, ho ho). I'm still interested in funds actually (and have money in some which certainly aren't making me rich...), and will give them more attention hopefully in the future, maybe transferring some knowledge from "trading" (s.betting) across, depending on how it works out

Edit: p.s. Don't take any notice of arabian's jibe - he's just a very naughty boy!

Yes, funds are great aren't they:cheesy: I learnt all about them with the dot com bust. I put something like 4k in a certain fund (I had total equity of about £25k so not all was at risk) and 'made' £3k in about 3 months....WOW, 'f**k am I GOOD at picking the right fund' ....the market then crashed and it went down to about £2.5k.
Lesson learned...

 

[nunrgguy]

Congrats to nunrgguy for honesty and willingness to face some of the inconvenient truths. And for doing some proper work in trying to get a handle on the markets.

IMHO, you are quite right about candlesticks - largely a waste of time. As pointed out, they are based on quite arbitary construction of time series for which the the only possible justification is bland statements to the effect such as "5 min charts are the industry standard" or they (substitute your TA method of choice here) work because "a lot of traders look at them". It would be interesting if somebody did a study of candle patterns in time series constructed from constant tick bars or constant volume bars. At least that would have the virtue of reflecting in some way the pace at which the market is moving. Not viable of course for fx, because it is not exchange traded in the normal sense.

Anyway to return to candles. One of the pieces of software that I have developed is a good stock scanner that handles technicals and fundamentals. And I do mean good - much better than all the crap out on the web. There is a backtesting facility in the sense of being able to scan on specific date, take the results of the query and chart an unweighted index of the selected stocks against a benchmark and do a performance comparison. Doing it has been a great education - far better than reading any number of books on trading.

The ability to scan for candle patterns is one of the features. Using the backtesting tools mentioned above, I have never been able to find any edge with common candle patterns. They just do not "work". Not only that, I have also looked at the count of the number of stocks exhibiting bullish or bearish candle patterns at any given time to see if that provided a tradeable indication of broad market direction. It should really come as no surprise that the horse had already bolted. As candle patterns are purported to "work" on any financial instrument, I consider testing on thousands of stocks to be a much better test than staring at a few historical charts of eg EUR.

The news it not all bad though - it is possible (though not easy) to select stocks on both long and short sides that are market beating. In the end it all comes down to observation and trying to reason through what you observe.

Not being fully satisfied with my screening, I also wrote some code to mark up the most common candle patterns on charts. Sometimes the human eye can pick things up that are hard to formalize. Attached is such a chart of SPY. Does this mess really mean anything?

That's a great screenshot there
While the software has found all the candle patterns, it hasn't singled out the ones which coincide with perceived S/R, trend, fibs etc. which are the only ones I've been looking at manually. I don't think there's many out there who would say that the candle patterns on their own without supporting factors mean anything anyway but my testing seems to suggest they don't mean much full stop.

 

[dcraig1]

That's a great screenshot there
While the software has found all the candle patterns, it hasn't singled out the ones which coincide with perceived S/R, trend, fibs etc. which are the only ones I've been looking at manually. I don't think there's many out there who would say that the candle patterns on their own without supporting factors mean anything anyway but my testing seems to suggest they don't mean much full stop.

Yeah, I know the arguments about candles and S/R levels, bollinger bands etc etc. Problem is they are generally pretty waffly and consist of a few charts. You can always find a few charts to support any proposition. And don't get me started on fibs...

I don't dismiss S/R though - just simple horizontal lines at prior highs and lows and especially gaps. It's really quite amazing to look at how stocks are attracted to S/R levels in relation to cycles in the broad market.

One suggested trading approach:

1. Identify cycles in broad market or that there is no current cycle if the market is on a tear

2. Identify sets of strong and weak stocks

3. Short weak or long strong stocks at resistance and support respectively at the appropriate points in the broad market cycles.

Of course there is a lot more to it than than. But it is a start and is less prone to misreading the tea leaves.

This is pretty much what Grey1 was saying on the now defunct technical trader forum, though his recommendation was to enter trades based on the stocks cycle. I reckon S/R is a better proposition.

 

[poborsky]

**** me another lucky pinbar trade I had this week. I will have to give this up cos the 15 straight losing streak is coming my way.

 

[trader_dante]

**** me another lucky pinbar trade i had this week. I will have to give this up cos the 15 straight losing streak is coming my way.

lol

 

[Paul71]

**** me another lucky pinbar trade I had this week. I will have to give this up cos the 15 straight losing streak is coming my way.

Yep, you've had it now Pob, mathematics is commin' to get ya! Don't ever think for a minute that it was skill that put you into your winning trades.

 

[Black Swan]

candlesticks rock as do pin bars...

 

[Bloodhound]

Hello Nunrg,

Really interesting thread. Not sure if you've had a chance to have a look at that post I sent you yet but here is an example on cable at the moment. It's a pretty obvious resistance becomes support level, so possibly worth watching.



If you look back an hour or so you will see an M5 'pin' bar (if you zoom in). Now I look at that and ask myself, what is important that is going on there? Is it the 'pin' bar or the level it is bouncing off? What is a pin par anyway - simply to me it's a pictoral representation of a 'v' shaped bounce off a level / fractal. In other words the important bit is the level. Q.E.D., why not just trade the level itself? Buy above the level and / or sell below with stops on opposing sides. The basic fact is that no-one knows what will happen next so the entry is pretty superfluous - it is how one deals with what happens post entry that will dictate whether the trade is profitable or not. My view on candlesticks is, whilst they work very well for some people (and I have no issue with that at all), they are intrinsically linked to levels. The danger is that, the higher the timeframe candle that we trade off, the further price will probably be from the level which triggered the formation of the candle and, therefore, the higher the risk in terms of stop distances etc. That's before getting into such details such as the very formation of a candle, price acceptance and rejection as a function of time etc.

To me, it's the level that's key. The further you enter a trade from where price bounced off a level, the more possible likelihood that price will go your way, but, with this supposed increase in likelihood, comes an increase in risk as logic dictates that your stop must always be the other side of the level.

Not sure if I have explained myself particularly well and I may just be echoing what has been written earlier in your thread.

BTW you may have noticed from the chart above that I don't look at candles at all. All I am interested in is price extremes.

All the best
Rob

 

[nunrgguy]

candlesticks rock as do pin bars...

yeah candlestiks rulz te interwebz