*Humans seek patterns where there are none, and make up stories to paint meaning over the noise.*

--------

Originally, I wrote a post on Bitcoin Technical Analysis. Perhaps I did it because TA posts earn a lot of money on Steemit, and I wanted to chip in. 

But then I've asked myself:
- Is this objective? Does it make sense?
- Am I fooling myself?
- Does this actually help my Steemit followers?

I have scrapped that post and debunked my own TA for your intellectual pleasure.

## Bitcoin & Technical Analysis
I have been cautious of, and turned bearish on Bitcoin in the past couple of weeks. Fortunately, I have sold most of my Bitcoin holdings prior to the current crash, which removes most of the emotion from my analysis.

In the absence of bias, interesting questions start to emerge. 

### Looking at the trendlines
![](http://i.imgur.com/74s6aPb.png)
*Is Bitcoin headed for a re-test of its long-term bullish trendline?*

My original post contained analysis, based on the market structure. In other words, I overlaid the chart with some pretty bold and dotted lines, and said, lo and behold, here are the trendlines.

![](http://i.imgur.com/CLlfvtI.png)
- The short dotted line is a short term trend, that was 'broken'.
- The second dotted line is a long term trend, lasting almost a year.
- The third and filled line is a long term trend, going all the way back to 2011.

What if the reason why we tend to hit support and resistance on arbitrary lines on the chart, is not because those lines are some kind of magical lines in the sand, but because people's belief in them made them so?

**Now my fellow reader, let me ask you a question:**
If I'm a chartist with a big following within a trading community, and a big portion of my followers see my 'Technical Analysis', does it make it more real?
In other words, will showing you these patterns plant a tiny seed in your mind, germinating into an idea that you think is your own. 
If the idea propagates trough the crowd, can it become a self-fulfilling prophecy? 


### The 'bearish' pattern
![](http://i.imgur.com/ku2EUSH.png)
Now these 3 patterns are not exactly the same by % similarity, however they do tell a similar narrative in all cases. First, we have a strong rally, followed by a small correction, a second attempt to rally that fizzles out with a lower high, some sideways to down, and after that, traders start giving up, exiting the trade, and then, the pain comes.

Before I go any further, I would like to make it very clear, that I do NOT believe this particular pattern has any actual predictive power. If it had, it would be magical.

For starters, it is not statistically significant.

-----------

**If you find this part boring, you can skip it.**
I have done some pattern analysis earlier this year, exploring the possibility of mining historic price action for fractal-ish patterns, and their 'predictive power'.
Although the research looked promising at the beginning, it soon turned out to be a dead end.

The problem is, that the larger scale timelines (30min+) hadn't produced enough similar patterns to make their analysis statistically significant. We need to have a good sample size if we don't want to get fooled by randomness, and 3 samples are not enough. Actually, 3 is far, far from enough.

The formula to calculate the required sample size I've used is as follows:
```
sample_size = (z-score^2 * p * (1-p)) / m_error^2
```
Z-score is our confidence of the outcome, p is proportionality of the outcome, and the margin of error (m_error) tells us % wise how much error can we tolerate.
Suppose we take 50% (0.5) as proportionality (the trade could be either winner or a loser, basically an unbiased coinflip). 

Suppose our desired confidence level is 90%, and our margin of error is 10%, meaning that the desired outcome is going to be +-10% accurate, 90% of the time in a real world. Sample size required for this scenario is:
```
sample_size = (1.645^2 * 0.5 * (1-0.5)) / 0.1^2
sample_size = 67.651
```

Only the short timelines (10min or faster candles) have produced enough similar patterns (with similarity treshold being 80% or more), to have statistically significant results.
I have combined similar patterns and looked at their outcomes. To avoid lookahead bias, I used walk-forward analysis, and I've ran the whole thing trough several montecarlo simulations to look at different variables. 

Long story short, the short term timelines produced positive outcomes on paper. In general, detecting a fractal, and then trading based on its past performance suggested a very slight edge in the market. HOWEVER, these patterns were very, very noisy. To act on every signal would mean to perform hundreds or even thousands of barely profitable trades per year. Do you see the problem?

![](http://i.imgur.com/fsjyjcE.png)
*(screenshot of my trading simulator)*

When I plugged the strategy into a more realistic trading simulator, it quickly became obvious that the slight edge would produce far less profit than what the trading fees and slippage accounted for, thus making our strategy lose money consistently over time.

And this is how my idea of using patterns to trade Bitcoins turned out to be worthless.
**==End of the boring part.==**

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

**So if the pattern is not statistically significant, you may be wondering, why I am I even talking about it?**

Good question.

I've stated before, that the the way I use TA and charts in general, is NOT to look for things with 'predictive power', but to look for things that could be used by other traders as a narrative. Things that are obvious, bold, standing out, hard to miss.

> Humans are emotional and irrational creatures by default, and most retail 'traders' lack this awareness. They seek patterns where there are none, and make up stories to paint meaning over the noise.

**Now you may ask, how do I know if other people see what I see? How do I know they are looking at this exact pattern?**
I don't. Here is where it gets subjective. I had a deja-vu. I've been in the market trough both historic 'patterns' in the chart, and this time it feels oddly similar in some unexplainable way.
If I get this sensation, perhaps some other people do as well. Or perhaps I'm just delusional.

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

To be continued...


### Addendum
Due to the recent events and the impact they had on me, I could not finish this post yesterday. It looks like I did it again...

I have sold my coins at $1100 on MtGox in 2013.
I have sold my coins at $650+ on BitFinex in 2016.

It looks like I'm an expert at losing all my money. 
I had all my savings on Bitfinex, mostly in USD lending: $50,400 gone.

Oh, and the pattern is still replaying:
![](http://i.imgur.com/dKYn35J.png)