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

Hey, I think you caught most of what I was trying to say. My native tongue is english, but I'm often not very good at using it.

No probs... we all learn =)

Thank you, yup same Silver, have been doing stuff in the background

I realized I should correct this to :
-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 wick price is 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 wick price is 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 wick price is not 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 wick price is not beyond previous swing high

Wick price :
current swing = up + swing's end candle = up -> use close price
current swing = down + swing's end candle = down -> use close price
current swing = up + swing's end candle = down -> use open price
current swing = down + swing's end candle = up -> use open price

Hope this is clearer.

Oh boy. Not having the wick price correct really threw me in circles!






Lets hope this one is correct =o

Correct.

Okey Dokey

Looks good. Now we have to wait for Pete.

No probs Pete.... I would continue. I am actually down to my 'final post' for this thread, so I am sure you are able to catch up irregardless.

Please refer to the following
Pages 8 to 23 Barros Swings and Trend Definition and 83-84 Statistical Time and Price Windows, The Nature of Trends
Page 64 to 66 How Vague Evaluation Criteria Contribute to the Confirmation Bias, Evidence Based Technical Analysis
(should had posted this earlier) The whole of Chapter 5 Statistical Analysis, Evidence Based Technical Analysis

4-Breakout Zone
This is a little harder to read, but I will do my best to explain.











The results for Trend Extreme and Wicks Extreme spreads out and there are almost no significant bumps. Compare this to Retrace Extremes. We can't apply the same methodology in this case, but this is where knowing the 68-95-99.7 rule for standard deviation can come in to provide some order.

Personally, I am willing to accept 100.1 to 120.0 as the breakout zone because :
a-100.01 to 120.0 covers more than 68% (1 std dev) for Rejects Extremes, around 75%. This means most Rejects don't push out too far. Anything beyond 120.01 or more, its 25% chance its a Reject.

b-100.01 to 120.0 covers around 10% to 15% for Trend Extremes. This means remaining 90% to 85% of Trends will go beyond 120.01 or more. This covers more than inverse 68% (1 std dev). Note that this area has a significant vacuum around 110.01 to 120.0. Its like price does not want to stop there and will continue on a a bit more or a lot more.

c-100.01 to 120.0 covers around 20% to 25% for Trend Wicks. This means remaining 80% to 75% of Trends will close beyond 120.01 or more. This still covers more than inverse 68% (1 std dev). Note that this area has a small bump around 100.01 to 115.0.

d-I am not too worried about minor bumps at further up at 130.01 to 135 (for Extreme) and 125.01 to 130 (for Wick). The main point is that the overall distribution beyond 120.0 is so unpredictable, that it is actually part of its strength. At least we have a good idea of where to start but not where to end.

e-To find it where it potentially can end, this means we can no longer use Frequency Distributions. I've also tried what Barros did and I didn't like the results. How and what can be properly done, I might or might not want to include in this thread. It is not the holy grail but yes perhaps close to it. This will be in the realm of Additional Components of MMLC v12.

f-Notice that for Reject Wicks, price came back about 50% of time around the 85% to 100% area. Yes, the overall sample size for Rejects is small, but important as we want to find a way distinguish true breaks and failed breaks of price.

g-Its possible to accept 100.1 to 125.0 as the breakout zone, but frankly I don't quite see the difference.

Additional Comments :
a-Coincidentally, Barros uses 120% as maximum extension for a sideways trend (NOT book page 9).

b-I had a difficult time determining the true breakout level. Hence why Reject type of swings are created. You may want to laugh at page 65 if you like, which I did =D. Its true that 'When an uptrend does not materialize it should be counted as a prediction error.' If the Reject type of swings are not created, then I would never had known the extend of prediction errors for Trend types. The older version of Trend types has results for Reject type mixed in it.

c-'Objective pattern definitions and clearly predefined evaluation criteria would prevent such after-the-fact evidence fiddling.' Knowing and reducing the prediction errors with proper classification of Rejects helped to improve the probability of Trend types better. I don't just rename failed patterns without proper justification. I improve the pattern definition further, this time by using the Close as a filter.

d-'Another way prediction errors are obscured is when forecasts lack a clear endpoint'. There are definitely still prediction errors for Trend types, e.g. breakouts that broke 120.0 but stopped at 125.0. We can see that very clearly from the data. I've accepted these probabilities but continue to work and improve on them. I did find something of significance regarding this recently.

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

Now we have a model for this particular market condition :


And apply this zone as well :


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

Overall Population of Swing Types
Sciurus asked about the Overall Population of Swing Types. (http://www.forexfactory.com/showthre...83#post8583083) It is indeed important. There are 2 main ways to get this done;
a-population sample sorted by swing types + b-swing type to swing type statistics





c-LG commented to me that it is close to 50/50 coin flip but I would say we have a smarter and more rationed way of dealing with this 50/50. The 50/50 we are seeing here is skewed on both sides; the key swing ratio levels from statistical analysis tells us that.

d-There was someone who PMed me sometime back and we talked about edges of using ZZs. We both agree about one thing; up swings do begot up swings later and likewise for down swings. The main problems are always finding the initial swing + knowing if it would continue or not. My answer to this is always the swing ratios, but swing type to swing type statistics do play a role here. IMO topping that up with some kind of momentum indicator (e.g. 15 Period EMA on M30) or MTF strategy can stack up the odds onto your favor.

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

Debrief
Discussion of key components of MMLC v12 is now completed. This thread has indeed fulfilled its purpose and objectives within a very short time (Nov 7 to - 11 Nov, that's 4-5 days wow). For those of you who took the time to code up, congratulations! You can now implement this as a Core Strategy and perform Primary Analysis for your own trading. IMO why is this very suitable to be a Core Strategy is because it uses straight out hard price/time points directly from the chart.

Trading asks us to successfully trade from point A to point B. We now have something that assists to help us answer that question and perform the role. You now also have an indicator that has the ability, through common sense statistics, to tell you if the overall market structure, in turn affecting the swing structure has changed. So far as I know, it hasn't changed for years.

Whats even more fantastic now is that tools and skills are on your ends hand to perform continued research for continuous improvement. It would be beyond what this thread's intended original purpose, but I don't mind you guys (LG, Sciurus, pete7) further the discussion in this thread regarding Additional Components. This is where you can start to be creative and start optimizing around. I do have a few Additional Components still on beta/staging at this point, but all of it looks statistically very good and should soon pass onto production level soon.

For anyone who is late upon this thread but would like to participate, do not fret. The instructions of swing construction and 100% of whatever you need to get started are well condensed within the first 2 pages of this thread.

Personally, I can't tell you how many times my current work has 'saved' me, from really horrible trading mistakes. With a workable and near predictable model of price movement, I am now trading with a lot of confidence and no longer fret with impatience. I take things 1 swing at a time, taking the most probably route available and do not worry about missing out on opportunities; its everywhere, from swing to swing.



Green pips everyone. To your own health and mine, green pips.

But I'm bad at being creative ):

To me the last picture you provided summarizes the thread quite nicely, and my take-away is simply this: those marked areas show the most likely areas "of interest" or action zones. Would you agree?

My issue remains though how exactly one goes about trading this. The first area (min retrace) is only of real trading use if you can project where the swing end will be. Picking random spots and looking for 30% doesn't seem like a great idea to me..
The second area seems more "coin flippy" to me. It just seems like a balance between potential R:R and win rate that ends up being roughly the same. There's no "this spot is best" (that would be too easy though..). The end (85, 90+ retrace) also has the R:R issue in my view. The probability that the swing will break is high, but to be fair to the stat, you can't leave the position until the next swing has printed, and who knows how bad the floating loss will be at that point (I think this is what you mean by finding the "failure stats")

I guess I always expected the edge to be more "obvious" though I suppose also, however, that this obvious and clear entry/exit system can only be found in the most refined areas of the model. Historically I've always checked areas of interest against the outcome (success or failure) and I generally don't get good results. Perhaps I should instead be trying to forecast probabilities for additional structures, rather than prices?
Not sure if any of this makes sense (who knows if it'll even make sense to me when I wake up..). I'll have sometime tomorrow (15 hours or so) to think about it some more..

But you are a genius lol.

Yes.

No of coz picking random spots for price to stop and then retrace minimally would be a bad idea. Yes, if you can project where the current swing end would be, that is the 1st thing to look for, then the minimal retrace comes 2nd. There is no other way around this, one needs to reduce risk first.

If the current swing is a retrace type, then one can more or less spot where it can stop or at least struggle to get through.
If the current swing is a reject type, then one can know but with less certainty, where it can struggle and reverse.
If the current swing is a trend type, then one has to use other methods to find that final swing end.

Price needs time, momentum and space to get through quite a number of 'barriers' before it can take off.

Not quite, but at least knowing that price needs to struggle through this area in some manner, smartly forces the trader to devote screen time when price reaches here. This is part of the edge I am talking about. One cannot do screen time all the time, but this is one key area I strongly encourage traders to watch and learn.

You don't have to force yourself not to leave the position until the next swing has printed. This is because the next swing, even says it does turns against you, it needs to get through its own number of 'barriers', before you have to truly call it quits. This is very true for squeeze market conditions (e.g. UR -> DR -> UR -> ?). This is also why I project the zones into the future. And yes finding the failure stats is important, which belongs to additional componments. The results are surprisingly good.

Yes, more work needs to be done, but what we now have is a very good start.

Yes and especially additional structures with specific rules. The structure gives context to the price and vice versa.

But thats why I record down things haha....

Sorry Rela, given I have been out a while I'm taking more than ever to code the stats and distributions. But what you are saying makes sense overall. I will get my stats and numbers in order shortly to see for myself.
In the meanwhile one other thing I wanted to check on was the 30x7 that you put down at the start. Have you discussed on this in the past, why 7 ? Can you point me to any thread to read please on this so I can do my own stats out to verify and learn.

I discussed a part of it here (http://www.forexfactory.com/showthre...14#post8170914), and later refined the PTe work to the point M30 x 7 is really good.

Rel, thanks for all of your contributions. It will take some time for all of it to sink in.

One thing I can say is that I just realized why there are so few retraces <10%. Its because in most cases a zigzag indicator wouldn't register a swing at 10% retracement unless there was significant sideways movement (flagging). There are actually plenty of 10% retracements that occur, but they never get registered as a swing. That in itself might lead to some interesting setups, because if price is flagging between 0-10% retracement, then a reversal might be in the works. This makes sense to me in terms of market structure.

I always enjoy number crunching if it will be useful at the end is a different story but it's just in my blood . Are we all doing this by programming our individual mt4 indicator? I am currently at the back tracing step and will most likely catch up during the weekend when I got a little bit more time.
Interestingly this reminds me of the work gg has done at this forum and I will give it a shot and combine it with the fractal indicator to see what insight can be gained. Oh this will be an exciting weekend.

Cheers

in my understanding, the swing takes time or candle count to develop. there are truely a lot of retracements happened within the swing , even within one single candle, not only 0-10% retrace ,maybe 30%,50%... even more...
in a candle stick chart, the swing formation is combined with price level plus time/candle count.
just my humble opinion...

They look great. Thanks Pete.

I would use the candle's open-close direction. If the candle is a neutral candle, then I would use midpoint-close. Yes this sort of odd thing does happen and I do account for it. Good question.

Yes. I run my PTe statistical indicator every 24 hours. I won't explain too much about PTe because it would take me quite a lot to explain its full logic and break down the basic construction (like how I did for MMLC v12). Maybe I might do it in future, I don't know. It isn't in my priority to do this at the moment.

I planned to do this once every 3 months, but it turned out I would just check it whenever I 'felt like it', because I could wiz up the stats in almost an instant anyway with a couple of clicks. I had been doing this sort of thing for a couple of years so far.

Someone I know who is a full time trader advised me to perform Out of Sample testing. It turned out to be very good advice, and it gave me more confidence to be sure that the structure that we now know is unlikely to change. Even if it does, the change would not be too significant (e.g. high vs low volatility periods of the year) and should not affect overall results too much.

What's more interesting is that you are using 2010 data and I don't really see too much of a difference between the results the others obtained.

Thank you.

Yes because there is a time (M30 x 7) qualifier and price (has to breakout of the box as a leading candle) qualifier for it.

This filter cuts out a lot of the 'noise', but it does not mean the 'noise' isn't useful; a good lot of it is indeed useful. Yet there's also totally useless noise that we really don't wish to devote screentime to, so what can we really do about this? How can we qualify and quantify this?

I understand what you meant because its part of what I am doing now. I don't use 10% retracements but another qualifier. But to rephrase it to your idea : Which are the 10% retracements that are important to look at and which aren't? Which are the ones that will lead up to reverses and which are the ones that will fail and become continuation of swing line?

This is where additional components come in. One such is MTF setup, which overall seems what you are trying to talk about. I've never agreed with classical MTF setup, but still accept and understand its concepts. Recently I've found a logical and statistical based solution to this personal problem of mine. The results are pretty good so far. I don't intent to post it because it will require further investigation, plus I don't wish to post up half-done work that is not presentable.

It did become very useful on a practical level, because now I have a reasonable model of price movement. Not manually drawn or eyeballed on the chart, but properly quantified and qualified. At least this has helped me better visualize and almost anticipate future price movements. Do what works for you =)

Yes.

Give it a shot to see it for yourself. You should have more than enough instruction to get something up and working.