Topic: Goomboo's Journal (Read 281467 times)

The price has been unmoving for quite a while now. That could be it.

I wonder what happened with Goomboo

In this post I have found a gem that should be in the awareness of every single bitcoin trader out there in Bitcointalk.

After reading the whole thread I sincerely thank you for sharing this piece with all the community.

Hi Goomboo,
Sorry if this question was answered already.

Do you use a stop loss other than EMA crossing in the downtrend direction?

If you do:
- how do you calculate your stop loss price (e.g. 1 x ATR or other way).
- how do you reenter the market if the market continues going upwards?

It is a little bit offtopic, but I am really amused by finding this text in the wiki paper you have recommended on the efficient market hypothesis (EMH):

Hey Goomboo, do you have any thoughts on using MACD for bitcoin trading? If so, care to share your rough guidelines?

https://bitcointalksearch.org/topic/m.767312

and there is a table of contents on the first page https://bitcointalksearch.org/topic/goomboos-journal-60501

Can you recommend for someone who's new at this what to read?
New I mean by no experience at all

Thank you

If you're looking to see what the average financial academic says about trading and investing, here's a place to start:

• Read A Random Walk Down Wall Street
• Learn the different forms of market efficiency
• Balance market efficiency with behavioral finance

Once you've read/studied the above, you'll have a pretty good grasp of where the typical academic stands.

Oh, and yet another question from me. As I understand the cornerstone of most trading systems is the assumption that markets trend. Are you aware of any scientific research proving that markets actually do trend? I am not trying to state they are not, just want to dig a little bit deeper in the underlying theory. Can you please point into the direction where I can find something worth reading on this question?

I'd run some tests on stocks and currencies.  I have found them to trade similarly.

I think I gained better understanding, thanks for the replies, Goomboo.
And, by the way, couldn't you please share what markets act like bitcoin market does thus being suitable for system optimisation?

That's exactly what I'd do.  I use the 10/21 combination because I have found it to be effective across decades of data from other markets.  In other words, I can trust this combination to help me know when to trade most liquid asset classes.  It let's me profit from the average trend.


In my opinion, there isn't a "correct" way to do it.  I have heard reasoned arguments for both ways (indicators first followed by stop-loss or all at once).  In my opinion, all that matters at the end of the day is that you have a system that catches the average trend and a fail-safe exit to protect you from wipe-out.

Yes, you are absolutely right. But the problem is that the only way to avoid curve-fitting is optimising the system on a longer time interval. The longest time period we have in the history of bitcoin trading is 4 years from 2011 to 2014. And during this time period we have 3 huge price spikes one in june 2011, the second in april 2013 and the last in november 2013. All of this spikes are quite similar and when optimising the system on an extended time interval from 2011 to 2014 the main influencing factor on the profit gain is how the system operated on those 3 spikes. The system performance outside this 3 spikes matters much less for optimisation results. This is an arguable opinion, but if needed I can prove it with numbers for systems that trade on 1 day time frame.

So, when picking up a system based on optimisation results we are picking the spike-catcher system.
Thus, the first problem is that there is no way to avoid "a system which is optimized around a certain market event" (3 events actually, which are very common) using only bitcoin trading data. Would extending optimisation data field to other markets be a proper way to deal with this problem?


If it is true that optimisation even on the whole bitcoin market data depends strongly on the ability of the system not to sell at the bottom of a bubble crash but to wait a little and to sell on the rebound or even later, when the price stabilizes a bit, then adding a trailing stop loss can change the results dramatically, making systems that sell at the bottom of the bubble crash more profitable than slower ones due to the trailing stop loss.

So wouldn't it be logical to optimise the system with a stop loss already included rather than optimising the system without a stop loss first and then add a stop loss into a system which is already optimised without a stop loss, thus trading the system not the way it was tested?

Before answering your below questions, I feel the need to suggest that you may be curve-fitting.  If you are to a point where you are trying to find a system which catches precise market action, you are too far "in the weeds".  The market rarely repeats itself exactly so a system which is optimized around a certain market event will probably not perform well in the future.  Think general thoughts - a moving average crossover system is designed to catch an average trend and hold it through its life-cycle.  A system designed to catch a specific trend will probably not hold up in the future.



You are correct - if you do not trade the system exactly as tested, your returns will be materially different.  Incorporating "gut" into your strategy makes very little sense in that it essentially invalidates your backtest results.  I highly discourage gut decisions.



A stop-loss is a key component of trading.  It is dangerous to trade without a stop-loss.  Over the long run, you will probably encounter a trade which will completely destroy your account if you do not set stops.  I suggest you research this more and incorporate a stop-loss into your trading system.


Here's the gist of what I was saying - optimization only works up to a point.  If you continuously optimize your trading strategy, you will constantly be chasing different variables.  The markets are always changing and a system that is heavily reliant on a specific set of inputs probably won't be able to keep up.  10/21 might not be the best optimized solution, but I am confident that the indicators have longevity in that they catch the market action that I am after and have performed well in several markets for decades.

Hello everyone and thanks for the great post!
Goomboo, I would appreciate if you could answer a couple of questions

I have noticed that when comparing results of different EMA pairs and building heatcharts the main thing that matters is when the sell signal is generated. The pairs which generate sell signal at the bottom of the panic sell pit make less profit then the pairs which generate a sell signal later when the price bounces back up a little, or even when the price stabilizes after the fall.
The point where the sell signal is generated, while the bubble collapse, seems to be the most influencing factor on the result of an EMA-cross systems comparison.

As an example during the april crash in 2013 the 21-10 EMA system sold at about $70
http://bitcoincharts.com/charts/bitstampUSD#rg180zczsg2013-03-05zeg2013-05-01ztgSza1gEMAzm1g10za2gEMAzm2g21
 
while a 17-22 system sold at about $100 and outperformed the 21-10 system on my backtests on the period 20.05.2012 - 20.09.2013
http://bitcoincharts.com/charts/bitstampUSD#rg180zczsg2013-03-05zeg2013-05-15ztgSza1gEMAzm1g17za2gEMAzm2g22

https://i.imgur.com/9M527n3.png?1

Adding some common sense while closing long positions during the bubble crash can increase the profit greatly. I think that somewhere on this thread you told that if you see a flashcrash you shouldn't sit around waiting fir the lines to cross. But doing this means ruining the whole method as you can no longer expect that the system used in a way it was not optimised for will bring you to the point of it's statistical expectation.

I mean that optimising the system by 2 parameters and then making decigions upon the change of some third parameter or a gut feeling makes no sense.

I can see three ways out of this situation:

First is that you choose the best backtest performing system and stick to it no matter what. You will get what you expect from your system but you will also leave some extra profit aside.

Second is that you choose the best system by taking into account only the quality of the buy signals, assuming that you will close long positions manually.

And the third is adding a third parameter for optimisation like a trailing stop loss which can imitate a manual position closing.

So here goes the question: have you ever considered the last two options and what do you think of them?


And the second question is about re-optimising the system. Somewhere on this thread you said that the 21-10 system is not the best one according to backtesting any more, but as far as I could understand you are against re-optimising it and choosing another EMA pair due to overfitting. Did I get it correctly?

i have a question...
the EMA and most other indicators can be made smoother by increasing the period parameter. alternatively you can feed it data from larger periods.

ex:
- you could feed the EMA the closing price of every 10 days and set it to 1 period
- or you could feed the EMA the closing prices of every 1 day and set it to 10 periods

i realize now that i've written this, that setting it to 1 period disables smoothing. but how do these alternatives compare?

Thanks, glad you have learned from the thread.

1.  I trade less size and less frequently during periods of drawdown.  Assume you trade in such a way that if you are wrong, you will lose 1% of your account balance per trade.  As your account balance shrinks during drawdown, you will trade less in order to continue only losing 1% of your account balance.

2.  I do not pyramid (in or out).

Hi Goomboo. First thank you for your extremely valuable journal. I have learned so much from it and still keep learning motivated by the clarity and wisdom in trading. 

I have the following questions:

1. What kind of risk control you apply in periods of drawdown as you mention above?
2. Do you pyramid your positions in the direction of the trend?


Thank you so much for your valuable time to share your knowledge and experience on this forum.

Just observing human nature.  These guys seem to be using a quote from Warren Buffet to justify an action, but violate their own rule by what they say.  The human capacity for self-deception is amazing.

Which of the two would you rather be?

1.  I make decisions based on my untested, fickle emotions.
2.  I make decisions based upon research and facts.

I think most would either desire or claim to be number 2, but when money is on the line, we tend to revert to #1.