The Scientific Method : MMLC (major move life cycle) v12

How do participate in this thread :
There are 2 requirements for participation.
1-Required reading of 2 books
Nature of Trends by Ray Barros
Evidence-Based Technical Analysis by David R. Aronson
2-Ability to code your own indicators, minimally to code the basic swing indicator which will form the foundation of everything else in this thread. I will not provide any of my swing indicators.

What will this thread generally compromise :
1-Key components of MMLC v12 broken down and explained in detail
2-Methodology of finding edges via Frequency Distribution and Probabilities
3-Continued research for continuous improvement

The idea is to ensure #1 and #2 can be easily replicated by other independent individuals. Why did I decide to take this approach? I realized that if #1 and #2 cannot be easily replicated by other independent individuals, it simply means a part of my trading logic is actually not logical at all and is subjective. This opens up to the possibility either ...
A-weak point which can be a blind side
B-strong point which is still stuck within the subconscious
... in which the The Scientific Method attempts to reduce down objectively, in order to improve accuracy and encourage truthfulness. Both A and B needs to be unlocked, broken down and properly dealt with.

This is also an advantage for myself. It becomes a failsafe should I be ever be physically/emotionally/mentally down, and my work would be able to succeed/outlive me and trading results will still continue with or without me. To being near 99% mechanical. Which is the scientific ideal anyway.

Basic Swing Indicator Construction (all stages + important note + corrections + clearer formula for Wicks):
Post #2 http://www.forexfactory.com/showthre...74#post8579974
Post #3 http://www.forexfactory.com/showthre...94#post8579994
Post #26 http://www.forexfactory.com/showthre...16#post8583916
Post #47 http://www.forexfactory.com/showthre...72#post8589472
Post #74 http://www.forexfactory.com/showthre...38#post8601538

6 Swing Types, Frequency Distribution and Probabilities
Post #6 http://www.forexfactory.com/showthre...28#post8580828
Post #17 http://www.forexfactory.com/showthre...19#post8583019
Post #31 http://www.forexfactory.com/showthre...64#post8585364

Example of Research Process for a potential Additional Component
http://www.forexfactory.com/showthre...57#post8594157
http://www.forexfactory.com/showthre...24#post8598424
http://www.forexfactory.com/showthre...44#post8600044

Basic Swing Indicator Construction :

A-RAW STAGE
1-On a M30 chart, draw boxes worth 7 M30 candles on a rolling basis


2-For each box, there is a 'major move' within the box. It is the highest high and the lowest low extremes of the box. Mark them with a swing. It is ok if the highest high and the lowest low extremes of the box are of the same candle.



3-Eliminate swings that do not have their end at the end of the box, but keep a marker at the opposite side on all end of box, regardless if their swings are accepted or not.





4-Hide the boxes


B-STAGE 1
1-For successive raw upswings, keep tracking until you get a raw downswing. Sum up the entire upswing by a single line. Continue the process, down->up->down...etc. Raw swings here are thinner, Stage 1 swings are thicker, hopefully you can see them


2-Hide the raw swings.

C-Stage 2
Backtrack as much as possible, code up (obvious) logic to join up these gaps. Simple thumb of rule is always maintain the lowest/highest extreme and adjust the 'wrong' swing.



D-Stage 3 (final)
There are 6 types of swings
-Down Retrace = previous swing is up and current swing is down, previous swing low NOT broken by current swing
-Up Retrace = previous swing is down and current swing is up, previous swing high NOT broken by current swing
-Down Trend = previous swing is up and current swing is down, previous swing low IS broken by current swing AND current swing's end candle did close beyond previous swing low
-Up Trend = previous swing is down and current swing is up, previous swing high IS broken by current swing AND current swing's end candle did close beyond previous swing high
-Down Reject = previous swing is up and current swing is down, previous swing low IS broken by current swing BUT current swing's end candle did not close beyond previous swing low
-Up Reject = previous swing is down and current swing is up, previous swing high IS broken by current swing BUT current swing's end candle did not close beyond previous swing high


Classify them and color code them any way you like.

Ok, I will stop here and give this thread a rest until the following happens :

Someone else is able to follow the swing construction instructions and send it to me via PM for verification. Please do not share it here as it would defeat the overall purpose. I can still post whatever I wish to share, just that without this happening, it would be difficult to continue this thread. I've stated in the 1st post that the construction of the swing is foundational. Its not quite blackmail (haha) Admittedly, this is indeed the scientific ideal : replicate-able results from independent individuals.

The reason for construction of the swing is simple. It would facilitate gathering of statistics later. If you are the one who did the construction, then the rest gets easier for you. However, if you get the indicator from someone else, you would have a harder time in the later stages. Trust me on this, the kind of statistics I intend to look at is not for anyone who wants an easy job.

*sits and waits*

Hmmmm. I was going to reply to the thread with an image, but in the spirit of the thread, I've decided to PM you instead. Cheers.

Ok now we can move on.

Browse through :
-Page 320 to 332 Out-of-Sample Testing and Walk-Forward Testing in Evidence Based Technical Analysis
-Page 83 Statistical Time and Price Windows in The Nature of Trends
-Page 88 to 93 Fibonacci Ratios in The Nature of Trends

As stated in the swing construction, D-Stage 3 (final), there are 6 types of swings. Now, lets do things one step at a time, focusing on 2 types of swings :
-Down Retrace = previous swing is up and current swing is down, previous swing low NOT broken by current swing
-Up Retrace = previous swing is down and current swing is up, previous swing high NOT broken by current swing

I have lots of problems with Barro's work, as much as I like it and hate it. I wonder why he doesn't use Frequency Distribution and relies on Fibonacci Ratios. Now, I am not a fan of Fibonacci Ratios, big time no. But why and how did he come up with these numbers plus stated the following in page 89 :
'To define which retracements are most likely to be met, I assess the current structure of the impulse wave. For exceptionally strong waves, I expect a retracement of at least 25% and no more than 30%'...and so forth

Ok then, lets do a basic test 1st to check this out :
1-Draw 200 trading days worth of swings as sample size

2-Pull out all the swing ratio lengths for the Down Retrace and Up Retrace swings
a-Down Retrace to Extreme Ratio
b-Down Retrace to Wick Ratio
c-Up Retrace to Extreme Ratio
d-Up Retrace to Wick Ratio
Edit : posted corrected screenshot of what is extreme ratio and wick ratio (edited twice)


3-Plot a b c d results in their respective frequency distribution tables. Here are my results for a and b. Lets see if you are able to get close to that. If you could, then I could explain a lot more about what I want to prove/disprove regarding Barro's work. Plus show that there are at least 3 significant edges, off several 'important ratios' one would extract from the results. I actually use these ratios for my trading and it does help a lot.


4-You may also want to try the following if you wish to perform some Out of Sample testing :
a-different periods of time but keep the same sample sizes e.g. 2010 M30 data
b-different sample sizes e.g. 50, 100 trading days
c-combination of a and b
So far I've done my own set of stats and found out that the 'important ratios' do remain consistent. That is very important.

Here's my take on it. The color scaling makes it a bit easier to see. Normally I think I would trash this type of result, so I'm curious to 1) Know where this leads and 2) See how many useful things I've thrown away in the past couple years.. X.x

Nah... don't trash it haha. I would explain in full detail after Sciurus is done. Great to see an extra head working in this, your results look good.

Ah OK. Just give me a little while to get organized here. Interesting so far.

Yup take your time to get organized, especially the coding part. The premise is that once you have all the data points and measurements for swings on hand (LG did everything on Excel, which still amazes me to this day), the rest is easier. You can actually check ANYTHING you want and see if there's an edge there. The only difference now is that I am pointing out the few key ones I had found after tons of checking and searching.

Perhaps to help you out further, what I meant by Important Data Points and Measurements, for each swing, there are key things to keep tabs on (using 3D array or better, via structs and classes, which is my preferred method):
Start Swing Price
End Swing Price
Start Swing Time
End Swing Time

Swing Length by Candle Count
Swing Length by Price (extreme)
Swing Length by Price (wick)

Swing Type (either one of the 6 types)
.... as a start. There are more but these are the ones even myself use a lot from.

Yes pete thank you. Nice work.

I think the histograms are correct but the tables for Swings2Close are not correct and are 100% similar to Swings2Extreme. Maybe the wrong table got extracted, but the histograms do tell me what is required anyway.

It looks great. Thank you. Now we have 4 ppl (including myself) on this, I would really like to wait for Sciurus.

I was chatting with LG yesterday and I admit that this thread is actually close to completion. I would cover the key components and not the additional components. I only need to cover 4 more swing types + explain the basics of how to read the frequency distribution tables, in relation to how to use them in trading.

Most of the information will be densely packed and prepackaged, with very specific instructions. After that, it would be the point when its free for all and you peeps can start toying with your own creativity on the swings. This is where additional components come in. I have my own but you too can have yours as well.

PTe will not be mentioned further in this thread because it only answers 2 questions. Which timeframe to use and what period to use for the swings. We already have an optimal answer of M30 x 7.

hi Rela, I haven't read the second book, although the first one in parts from a year back. Have sent the swings to you on PM.
-Silver

Here is my data for EU 30M. Had to revisit the swing logic as I was ending up with fewer swings over the 200 day period. This is for a slightly longer data sample to increase sample size.

Hi Sliver, glad to see you =) (hopefully you are the same sliver I knew a couple of years back, yes?) Yes the swings look good. Very good.

Interpretation of the results









--------------------------------------------------------------------------

How to read the frequency distribution tables and use them for trading (part 1)

1-Minimal Retracement zones (0.0 to 25.0)
a = This is the minimal amount the next swing has to retrace from the previous swing's extreme.
b = We assume we already have good probability of the swing ending and this is dependent on the type of swing currently in play. we would discuss more of this later after all the swing types are up.
c = Think of this as a vacuum zone, where price needs to get out of. If you are a brave soul, zoom into M5 chart and scalp against the current swing direction. At least you know where the safe limit is, so why not?

2-Retracement Success Zone (45.1 to 85.0)
a = Price has good probability to stop somewhere here when a swing retraces, implying there is a good chance it will stop here and then start to print an opposite swing. This allows 'possible trend continuation' after retrace is 'complete'.
b = Take this as a cushion zone, not a hard cushion but a soft cushion, which can be broken. Note that it can stop before this zone, but the occurrences can be statistically seen so that one makes a better and more rational decision.
c = Possible targets after retracement stops before or inside this zone will be discussed after all the swing types are up.

3-Retracement Failure Zone (85.1 to 100)
a = Price has poor probability to stop somewhere here when a swing retraces, implying its better to reverse trade from here.
b = Possible targets after retracement fails in this zone will be discussed after all the swing types are up.
c = One can also take this as another vacuum zone. If a candle closes here but the extreme of the current swing (in process of retracing) did not breach the 100, according to these statistics, 'something is wrong'. This is actually a good place to prepare exit if one is on the wrong side, especially after taking a trade in Retracement Success Zone.

One certainly do more studies regarding these zones e.g.
-sort by origin of previous swing (much like how Barros does it his way)
-test all signals given by 3C and not given by 3C (e.g. price just shoots pass the zone and did not close in there) and resulting probabilities
... but we will talk about that much later.

--------------------------------------------------------------------------

1+2+3 means we have a model for the following type of market condition :

... but this is only 1 such market condition.

Apply 1+2+3 zones this way
a-onto the previous swing and project it out to the current swing
b-onto the current swing and project it out to the future swing that has not been drawn yet
... you would get something like this, try it!


--------------------------------------------------------------------------

But we need more data from the other 4 types of swings and eventually more zones to form a complete picture, because there are other market conditions in play as well.

1-Pull out all the swing ratio lengths for the rest of the 4 swing types
a-Down Reject to Extreme Ratio
b-Down Reject to Wick Ratio
c-Up Reject to Extreme Ratio
d-Up Reject to Wick Ratio
e-Down Trend to Extreme Ratio
f-Down Trend to Wick Ratio
g-Up Trend to Extreme Ratio
h-Up Trend to Wick Ratio
... it shouldn't be a problem for anyone who has been able to keep up so far, perhaps you are already expecting it haha.

2-Post the results here. To clarify about #Post 4 (http://www.forexfactory.com/showthre...94#post8579994), what I actually meant is that please do not post the indicator, but its ok to post the results.

Hi Rel,

I'm going to challenge you a bit . I think you are easily up to the task.

Have a look in at the following percentages. Retracements are essentially a subset of swings that continue the trend of the last swing, and price that retraces to >85% is actually a sub-subset of that. Is it fair to calculate the percentage of retracements that turn at the 90% mark using to totals for all points that retraced? Price that retraced at 10% never made it to the 90% mark, no?

Taking this a step further, price can do 3 things at the 90-100% retracement mark. 1) It can continue past the last swing and not turn at all. 2) it can do a partial retrace and not break the previous lows, 3) it can retrace and break the previous lows.

Wouldn't it be nice to know the percentage of a) all swings that broke the pervious H/L, and b) the the distribution of those retracements within the prior price swing. Lets just say that price retraced with the trend and broke the prior L/H 50% of the time . IF (lets just imagine), there was a uniform chance of this retracement happening anywhere in the prior swing. Wow, wouldn't that be nice! because the implied setup is fabulous. We can always dream.

Now, I think we should be adding the chance of there not being a swing at all to the counts and then calculating percentages from that.

Yes

Yes

Yup they don't and it can be fair. You are going into to 100% probability of all swings will retrace at least 10%. I read you and I am also tempted to do that initially for this thread. That both ends of the ratios can be 'further chopped off'. To optimize the retracement statistics further. I consider optimization of such belonging to additional components, because one needs to grasp the basic key components 1st and its obvious you have no difficulty.

The basic components itself has to have some kind of an baseline edge before optimization can begin. If the basic component has no baseline edge, IMO optimization won't get it any further.

It can be done, sure. I did do some tuning of such before for all the ratios I found. But to keep it simple for the audience, using the full general 100% length instead of an optimized 90% is preferred (for now)

Yes

Yes

Or maybe either you typed it wrongly here or I am reading it wrong, isn't it for a up swing -> down retracement :
break the previous lows
and
retrace (and close above the previous low)
.. since highs and lows are formed 1st before closes do (obviously)

But thats what I wanted to post about after the other 4 swing type stats are out haha Yes its good to dream and this dream is actually realistic lol.

Yes I am doing it the opposite way, but yes, I would need to explain this later. The overall chance is important because there are various ways to calculate it out.

Hi Pete I think you got all the results off the mark. Perhaps there's something off / a bug in your logic?

To recap here :
-Down Trend = previous swing is up and current swing is down, previous swing low IS broken by current swing AND current swing's end candle did close beyond previous swing low
-Up Trend = previous swing is down and current swing is up, previous swing high IS broken by current swing AND current swing's end candle did close beyond previous swing high
-Down Reject = previous swing is up and current swing is down, previous swing low IS broken by current swing BUT current swing's end candle did not close beyond previous swing low
-Up Reject = previous swing is down and current swing is up, previous swing high IS broken by current swing BUT current swing's end candle did not close beyond previous swing high

...and what is extreme and wick ratios (http://www.forexfactory.com/attachme...8&d=1446998938)

...plus this is how my results look like :
a-Down Reject to Extreme Ratio
b-Down Reject to Wick Ratio


e-Down Trend to Extreme Ratio
f-Down Trend to Wick Ratio