It is I (ME) that would like to get some input on something that has been bothering ME for a time.
As you will all probably be aware: when using moving averages we are ‘generally encouraged’ to use EMA’s (‘Exponential Moving Averages’) the reason being is that no data is simply ‘dropped’ from the calculation of an EMA i.e. the most recent data carries more ‘weight’ than the earlier data but the earlier data is not simply ignored. Right???
Well here’s my issue: why should we CARE about earlier (old???) data???
This is what has led me to post about this (one example anyway):
I’ve seen MANY different ‘versions’ of the RSI calculation. SOME use Wilder’s smoothing method, some simply use an EMA, and some don’t use ANY type of smoothing i.e. only the data from the most recent number of periods is used. I’ve now coded all three of these ‘versions’ and compared them. What’s interesting is that RSI calculated using Wilder’s smoothing VERY VERY seldom gets into overbought or oversold territory (so therefore there are VERY few signals given by this ‘version’). Using an EMA: slightly more signals are generated. What’s strange is that with NO smoothing: the signals appear to be the most accurate of the three!!!
Another thing that has made me ‘ponder upon the above’ is the ‘Golden Cross’ as it is known i.e. when the 50 day SIMPLE moving average crosses over the 200 day SIMPLE moving average (on the S&P 500 for example). You’ll note that they are not EMA’s but just SIMPLE moving averages that are used (and NOTED by most investors).
I realise that the use of smoothing techniques is SUPPOSED to eliminate whipsaws BUT at the cost of some VERY VERY late signals (and inevitably ANY moving average will whipsaw you ANYWAY).
I guess there is no ‘answer’ to my issue AS SUCH but some comments and inputs and thoughts on the subject would be nice.
I don’t know if this goes to what your saying but consider this:
In comparing say an EMA system with a simple MA system, one can get very close to the same entry and exit signals by varying the parameters. For example, you mention that the EMA is supposed to get you in later in the move and keep you out of more whipsaws (and potential profit). One could achieve the same results by lengthening the SMA look-back period. Conversely, if one were to shorten the EMA’s parameters, one could conceivably get more signals than a SMA system with a longer look-back period.
So for me, it’s not so much which system one uses. EMA, MA, S/R breakouts and others can all be tweeked to give similar entry/exit signals.
Someone has just pointed out to me a variable moving average that seems to be good at eliminating whipsaws when price is ranging…
to answer the question of why do we care about the past… then why not use a 1 period ma ? not trying to be a smart@ but just saying it’s kind of like a trend line isn’t it? The longer the lookback, the smoother its is and the more it shows the overall trend.
rsi seems to be more of a momentum indicator than an over bought / sold indicator.
Yeh, OK: I see your point. That said: DON’T KNOCK IT!!! LOL!!! At SOME point I HAVE tried a 1 period SIMPLE moving average of the close!!! If it were NOT for the dreaded whipsaws you’d make a fortune!!! LOL!!!
Again: this is one of THOSE threads that I’ve started that I’m not QUITE sure what I’m hoping to get out of it!!! LOL!!!
I guess I’m just finding that with most indicators that use moving averages of one type or another: they seem to get ‘bogged down’ by ‘old’ data and by the inherent smoothing effect of an EMA if used (or whatever other smoothing method is used) whereas using an MA makes them far more responsive. That being said: I know that NOT using either an EMA or Wilder’s smoothing (which is simulated by using a very L-O-N-G EMA) with ADX generates many more false signals. Regarding my RSI example: I SUPPOSE one COULD argue that if it’s correctly smoothed as per Wilder then when it INDEED is in overbought or oversold territory then there can be NO mistake that the instrument IS INDEED overbought or oversold!!! LOL!!!
As I said: I’m not QUITE sure what my objective is here. Obtaining ‘food for thought’ I guess.
just curious, you didn’t mention LWMA have you tried that? I edited my earlier post. just thinking, in a long uptrend or down trend when rsi is at extremes it doesn’t necessarily indicate over bought or sold does it? maybe it does and I’m misinterpreting.
But like any MA or set of MAs, rsi peaks when price peaks and dips when price dips, it’s just compressed in a limited vertical range.
WELL now!!! You ask a very pertinent question the answer to which addresses a very contentious issue!!! LOL!!!
To ME: RSI is far better as a momentum indicator than in indicating overbought or oversold conditions. That said: Wilder (in his book) begins the RSI chapter talking about the shortcomings of your ‘standard’ momentum indicator (you know the one i.e. CLOSE - CLOSE(n days ago)). The odd thing is that AFTER that intro: NOWHERE does he even mention that RSI being above or below 50 is meaningful i.e. the rest of the chapter is dedicated to explanations about RSI indicating overbought and oversold conditions (which he refers to as ‘tops and bottoms’) and the recognition of divergences and support and resistance etc. Nowhere after the intro does the word ‘momentum’ appear!!! Why do I say contentious? Because there are those that believe that the HIGHER or the LOWER the value of RSI the stronger the trend in that direction and these supposed overbought and oversold conditions are NO reason to start looking for the reversal of the trend!!!
I suppose what REALLY prompted me to take note of these issues is because somebody asked me to test a system that uses RSI(14) and a variation of Bollinger Bands. The documentation says to wait for RSI(14) to be in overbought or oversold territory. The problem is that if you use the ‘correct’ RSI calculation (as per Wilder) you will hardly EVER get a signal at all (because of the smoothing). HOWEVER: if you REMOVE the smoothing you get far more signals (most of which appear to be accurate). So what’s a person to do!!! Go with what you KNOW to be ‘correct’ or ‘relax your standards’ and use a modified version of RSI!!! Of course: this then got me to thinking about the different moving averages (and of course: we now have another thread on the subject)!!! LOL!!!
I installed a chain link fence once. There is a bar that goes along the top length of the fence. after getting it level it ‘looked’ crooked as heck due to the way the ground slopes. After making it ‘look’ level, it was inches lower on one end than the other. Sometimes you have to go with what works instead of what’s right.
It’s not often that I wake up and the first post that I read in the morning immediately puts a ‘smile on my dial’!!! ROFLOL!!!
That’s probably the BEST insight and advice that I’ve EVER seen or been given!!! LOL!!!
You have NO idea how close to the truth the content of your post is (as it relates to ME anyway) i.e. I’m TOLD that I’m far too ‘rigid’ when it comes to trading systems (I guess I’m like that in ‘life’ too). For instance: JUST BECAUSE a trading system will have me place an order 1 tick above the high of a bar I WILL place that order 1 tick above the high of the bar EVEN ALTHOUGH my common sense and experience tells me that it’s too close and it’s GOING to be false stop and reverse (or whatever)!!! LOL!!! I’m THINK I’m getting better at using (actually TRUSTING is a better word) a bit of my own judgement rather than adhering to a very rigid and strict set of rules but I’m certainly not there YET!!!
YOUR post is sure going to help me!!!
Regards,
Dale.
Edit: it’s really funny though isn’t it i.e. I’ve had the same experience myself MANY times. I’m one of those ‘anal’ people who will use a level to hang a picture for example. I KNOW the level cannot be wrong but the picture will somehow seem ‘skew’ when seen in the context of its surrounding. The difference between you and me (up until now anyway I HOPE)? Just because I KNOW the picture is level I will leave it that way regardless of how ‘wrong’ it ‘looks’!!! LOL!!!
Great post MattW2009. Great post. As you say: sometimes a thread such as this leads to the posting of some real ‘gems’ and believe that to be the case here now as well i.e. I’ve much ‘food for thought’.
Moving averages, in my opinion, are the ‘no nonsense’ approach to trading. I spent MONTHS this year playing around with different moving averages and I was fascinated to find that some of the most COMPLEX of trading systems (that generate signals based on calculations that would make NASA proud) generate exactly the same signals at the same time as one single moving average (length and type dependant of course).
As far as RSI is concerned: as you probably know I (and a few others) dedicated at least a year to Wilder’s trading systems and I don’t think that we ever reached agreement as to whether or not Wilder’s smoothing was INTENTIONAL or if it was just a time saving mechanism (although referring to RSI he DOES ACTUALLY MENTION that using his particular method of saving time has the added benefit of applying smoothing to the indicator).
As I said: there is a little system that somebody asked me to test and I was just a bit concerned that a Wilder smoothed RSI will almost NEVER generate a signal but, according to the creator of the said system, signals should be generated quite often during the day. I can only surmise, based on my experimentation now, that their RSI is either NOT Wilder smoothed and, at best case, uses a standard EMA to apply smoothing, or NO smoothing is applied at all. As a matter of fact: THIS VERY DAY there is post in which somebody (so eloquently) noted that ‘RSI sucks’!!! I took a look at it and noticed that THEIR RSI MUST be either Wilder smoothed or EMA smoothed because it exhbited the same issue. EVIDENTLY: not all RSI’s are created equal!!! LOL!!!
Regarding moving averages: I’m SO glad that at SOME point a few months ago I FINALLY realised the the ‘Golden Cross’ on the S&P was formed using two SIMPLE moving averages (why didn’t they just SAY SO)!!! LOL!!! For at least a year before that I used to think that either THEY (Bloomberg) were nuts OR there was something wrong with my software!!! It turns out that (because of the ‘conditioning’ that I referred to in my first post) I just ASSUMED that they were using the 50-day and 200-day EXPONENTIAL moving average (and needless to say: my ‘Golden Cross’ was FAR away from happening i.e. I’d have been a few months late)!!! LOL!!!
Anyway: thanks again everyone for all the valuable insight. It’s given me food for thought and I’m sure it will beneficial to others in time as well (but this does by NO means mean that the thread is ‘done’ i.e. there are probably as many ‘bastardisations’ of moving averages as there are traders on the planet so feel free to let us know about them).
I think I’ve found a solution to my RSI issue (the one that gave rise to my starting this thread in the first place).
Simple solution: ROUND RSI!!! The initial problem was the RSI was JUST getting to the 30/70 limits (most times missing them by decimals). This morning I had a ‘lightbuld moment’ and decided to ROUND RSI (to the nearst WHOLE number). Problem solved!!! The CORRECTLY ‘Wilder smoothed’ RSI now generates more (and reliable) signals (for the trading system on test anyway).
Just FYI!!!
This may have other applications as well of course.
Nope: not any more. This RSI ‘thingy’ is just a simple system that somebody asked to have a look at is all. Me: I’m managing to stick with two systems: one for long term trades and another to keep me occupied so as to not mess with the one for long term trades!!! LOL!!!
