Higher Edge within a Single Candlestick

Hi V. I just thought I would add my content throughout your post as in discussions such as this context matters.

Yep mate....in a modest way I was stating exactly the same thing. By definition a fat tail is totally unpredictable in terms of timing and extent. By relying on these events in your trading plan you can only ever treat them as wealth creation exercises as opposed to sources of cashflow. That's why to trade them the following really helps:
1. Capital backing; and
2. Diversification.

I actually do treat flash crashes also as structured events, in that the event itself is an exceptional momentum move outside of what can be defined as a normal distribution. Price action that sits in a normal distribution is what I treat as noise in terms of my trading philosophy versus the exceptions which I treat as exploitable momentum moves of more enduring nature.

Obviously I can't exploit all forms of fat tailed event however and I specifically avoid trying to trade a flash event due quite simply to the speed of the reaction required to address it. If I am already in the market and these events occur, diversification either protects me where a stop simply cannot be relied on....or if on the right side, I simply exploit them by being at the right time and place. For example I fell foul to the Swissy event while in a long trade but because only 0.25% of trade capital was applied to the trade, the slippage was only 2%.

You will find that about 80% of the systematic diversified traders in the CTA space are either diversified trend traders or more specifically momentum traders and apply exactly the same principle in their trading plan. The terms diversification don't get much breathing space on this forum....and this is actually where obtaining a definitive edge is not such a hard game like that of seeking an exploitable edge from a single instrument. Now I won't argue with others out there who think it is possible but just will say the following......An efficient liquid market by definition is one in which exploitable arbitrage is difficult to find. That is the whole purpose for markets in the first place. To provide an efficient mechanism of exchange between buyers and sellers. There is only weak alpha (or exploitable arbitrage) in an efficient market. Exploitable opportunities are greater in less efficient markets or less mature markets. Now with the rise and rise of the algo's....exploitable opportunities are fleeting instants in time and the seething mass that represents the participant structure immediately pounces on these opportunities. A single fish like us retail traders simply becomes the opportunity or quite literally the food for the balance of market participants.

Let's assume for a moment that we are not legendary traders and that there is less alpha available in the market than we otherwise might assume. How can we build this level of available alpha through real quantifiable measures? In terms of diversification this ability to magnify alpha is what is referred to as 'the free lunch' of diversification. Simply by having a diversified range of return streams (or equity curves), then provided those return streams are not highly correlated, you magnify your alpha just by being diversified as the blend of return streams produces improved risk weighted returns. Once this is achieved you can simply scale up to your desired risk weighting through position sizing. All this is through a knowledge of how to manage a portfolio in quantifiable terms. There is no 'luck or skill to this'. There is a simple recipe available that is understood by almost all portfolio managers I know....and I know a bundle of them.

So in a nutshell I exploit weak alpha from trading the fat tails, but I magnify this through diversification. I simply apply a proven recipe that is the supposed 'secret that is not shared by those 'quant' houses like renaissance technologies etc. I therefore don't need to be a legendary trader or a 'pink unicorn' who can catch a fly in his chopsticks or practice a single kick a thousand times. If you can't beat them....join them :-)

Totally agree with this mate.......however the principle I apply is that if I am a trend follower or momentum trader, then the technique I apply must be repeatable. That's why I define a trend for my purposes as price action contained within a standard deviation channel of a defined width of 1 sigma. Provided people can identify a swing high or swing low, then by defining price action in this context....everyone draws exactly the same trend. So my advice to any trend follower out there is to use a definition that is replicatable. This helps a systematic approach and steers away from a discretionary approach.

Remember that I am simply trying to stay alive during the normal market churn...and defining trends in the hope that some of those events will take off in the linear direction I have defined within strict momentum constraints. The reason for the strict momentum definition is not for profits...it is to manage risks....in fact almost 90% of my strategy design is to simply protect my capital and be able to survive whatever the market throws my way. My profits arise from a chance 'fat tailed event' in which momentum is exceptional.


V....I can't disagree with anything you have said there. That is also how I see it...but to capitalise on any definition of trend you need to define the constraints in which your definition of trend applies. When price leaves that definition, that is when your trend ends. Everyone may have their personal definition...there is no right or wrong...but the key to trading a trend is repeatability and strict definition. The way to achieve that is through systematic application. Discretionary application doesn't really cut it for me as repeatability is compromised through discretion.

The way I see things is a bit different to the norm in that I seek to understand the underlying processes at work and then set traps to catch outcomes that are favourable to me. Much like a hunter in the savannah or a native that sets fish traps. My trap is the particular configuration I have adopted for a favourable outcome outside the day to day churn (a spider web)....and this is what I build in my design and insert into a live market. It is empty when I insert it...but provided a condition or event occurs within it, then the design of the trap ensures I catch the result. In this case I am after fast momentum bursts in a linear manner that are outside the range of the normal distribution.Provided that the events inside that trap lead to positive expectancy which is assisted by it's extreme placement then I can lay them all over the savannah..and across any type of savannah.


I really like your conceptualising here V and the way you join your dots of enquiry. Context is everything. You will find a guy called alphaomega on this forum that approaches problem solving in a similar manner to you.

What I particularly like is that your model addresses all forms of possible general market condition and is scalable to any timeframe.

I actually think that the path taken or style of interpretation is less important than the logical consistency of your approach. If it agrees in a consistent way with your logic that is defined by your understanding....then that is the way to address it. The way to therefore improve on your logic and hence your system is to expand your understanding where you have more dots to connect in a logical consistent manner. It is an ongoing process of improvement. When the logic breaks down you then know that 'Houston we have a problem' and you then need to backtrack and identify the logical fallacy. An approach like this is a more informed way to know whether you are on the right path or not.....as many paths lead to Rome.

Bottom line is that I come from a pretty strong physics background and actually visualise the market activity not as a bar chart but as signals lying in random noise much like an electronics engineer manipulates sound waves. Each pattern is an equity curve in which a significant component of that equity curve represents a random chance outcome...however their are legitimate signals in the noise and when we start to see the signal to noise ratio rise....you know that something 'real' is going on. In a sense I do not trade the market using a visual chart....I actually trade equity curves and through portfolio management deliberately configure the desired result.

Thanks for sharing V :-)

C

It's important to use the correct daily candles!!!!! 10pm GMT is daily OPEN. Did you consider the correct open and close?

Admittedly I did not. However that doesn't change the statement that the markets have a 25%-50%-25% wick-body-wick distribution.

Well if you think so....

Great spirits have always encountered violent opposition from mediocre minds.

I’m curious about the result: 1/3-1/3-1/3?
k

The outcome will completely depend on the underlying distribution used.

Spot on!

I used the standard Random function of excel, hence the distribution used is uniform.

Not important. Distribution stays the same across timeframes and random opens.

I'm a dumbass. I apologize for this. However, does this mean your conclusion is that markets are not random? And if so, what edge would you attribute to the difference in distribution?

Seneca Pilot's method seemed to be highly discretionary, although he did write that he paid money to have it researched and it held up well historically. I tested it methodically myself and found it to be catastrophically bad. I also seem to recall he put an explorer up based upon the method and it tanked. If you have some thoughts on how it can be made to work I'd love to hear but it seemed like a dead end to me.

I would say that a 25-50-25 distribution is a statisticly significantl deviation from the 33-33-33 random distribution.
So my conclusion would indeed be that markets are not (completely) random.

This in itself is an important statement, I feel. Since it is impossible to be profitable in a truely random market consistenly in the long run, this statement simply means an edge can be found, and it should therefore be possible to become consistenly profitable (as long as the distribution stays 25-50-25).

However how this information can be used to find an edge for me is still unclear. Maybe others have ideas how it can be used to find the edge?

cheers.

although i regret that when i first saw the thread in 2013 that i glossed it over, only a few months ago after someone mentioned it i read the whole thread and i was able to get an overview of the ops learning process/mistakes, this time from an after-the-fact birds eye view of his learning process. ie it took months for him to learn from his mistakes and tell about it, so i saved myself this hassle

ie he vacilated between discretion and non-discretion. early in the thread you see him patting himself on the back for cutting a trade early, then months(maybe years later) he kicks himself for closing trades early

he also vacilated between pairs, when sometimes the eu or gu would not react to the lines for a while and he opted for yen pairs instead, then back again

he also was insistant on a greater than 1:1 rr per trade, and where/when to take profit is where he used too much discretion which was one of the major stresses he often talked about

he also used a stop out, re-entry strategy, which does have a high win rate, but psychologically not so great due to incurring double losses and sometimes triple losses where he would enter a third time

so for me i adapted it (so far) in my own way, picking my own start time and pairs(actually i trade two sessions per day, folding into each other). also i do not re-enter after a stop out. and most importantly, i am staying the h e double hockey stick away from usd, simply because i trade the ny session and usd and cad for that matter are just way too succeptible to volatility and erratic behaviour at this time

Be careful since not all numbers are completely random at all. For instance, the open is close[i+1]+-error. An even if you assume a complete randomness, the distributions will be extremely skewed and median < 0.3!

Perhaps a better way to simulate it would be to use long random walks. I guess in this case the average candle would be a doji.
k
https://en.m.wikipedia.org/wiki/Random_walk

Valid point. However also setting Open[i] = Close[i+1] results in a 33-33-33 distribution.

Incedently I also find that the Close closes near the open around 10% of the time (within a 5% deviation of the open), meaning that in a truely random market

"90% of the time price will close above or below the open."

whereas using a 5% deviation on actual market data, it indeed shows a 95% probability of price closing above or below the open.

Thanks.
I got the feeling that he developed the strategy partly as a way of dealing with being in a less than optimal time zone, so he could get some sleep during his night time without having to babysit the trades.
Good luck with it.

I think I’m being silly: If one would separate the averaging between the “up” (close>open) and “down” (close<open) candles, I’m not quite sure what the result would be (possibly <close-open>~sqrt(n) for up candles? Not sure about average high/low. There should be some exact results available...). And I guess this procedure would be more in line with the subject of the thread.

k

Yes, stop-order is a lot safer than limit-order ...
For every order that has clearly defined EP-SL-TP, the limit-order will always execute losing trades but not necessarily winning trades ...
While in a stop-order, we still have a chance to cancel the supposed-to-be losing trade if the SL is reached first before the EP ...

CP, do you know a Jerry Cidzik? Also goes by JC.