Topic: Goomboo's Journal - page 30. (Read 281470 times)

This thread was constructed by taking knowledge learned in the professional markets and applying it to the BTC markets.  These principles are pretty much uniform across nearly all (liquid and active) markets.

To create them in NinjaTrader:
- Export optimization results of a moving average crossover system into Excel
- Create a Pivot Table of results
- Apply conditional formatting

I didn't use the 11/21 because I am trying to avoid cherry picking (selecting the best results) and be closer to the middle of the island of green.  The markets shift through time and the underlying theory behind the charts is that in the long run, the best area of results will tend to continue to outperform and we should try and put ourselves within that area.

Yeah, here's another area where I chose to use discretion.  In an environment in which you pay relatively large commissions and the spread is huge, you're going to want to trade less or you will be eaten alive, even if you are correct in market direction.  To directly answer your question, I did test those time periods, but commissions and spread really threatened the results.

I don't particularly see the utility of an update.  That chart was created during a different environment and mimics what I have typically found to be true of other markets giving me greater confidence in the results.  If I were to re-create that chart using recent data, it would be skewed towards longer period averages due to the recent run-up in prices.

Unfortunately, not all elements of trading, testing, and optimization can be fully scripted - you've got to rely on discretion at some point and I am doing so by saying that I believe those results are a more true representation of what will work in the long term.

hi gomboo


hi do you have something for forex/precious metals

i m pming you

Depends on the source you have.

looking at the first line:

Don't quote me but it appears to be:
separated by ";"

You could obtain data to play with here:
http://bitcoincharts.com/about/markets-api/

with the format (unixtime,price,amount)

you would need to, pseudocode:


if my initial assumption is correct you would need something along these lines, unless you are able to find historical data on a daily basis instead of trade by trade.

How can I generate a fresh new BTC.last.txt like you've uploaded in this post?
- https://bitcointalksearch.org/topic/m.716004

This ?! How to backtest BTC/USD with NT? (+ why not 11/21?)

Edit:

Found the answer to backtest here: https://bitcointalksearch.org/topic/m.715563 (in this topic )

But my last question still stands.

Nice simulation :_)

I think I could include such simulation in my bot.

However it would be some signifficant work.

First have to implement "noTrade" mode, then use this mode to generate data (each cell  as I think is an effect of going through some historical data with certain ema params).

It would have even some configurable simulation boundaries ...

I thought Gooboo did post some extended charts but I cant find them now, would be interesting to have the charts updated though if anyone can do it

It looks like even better EMA pairs might be hiding off the top of that chart.

Is there any data for 1-12 vs 13-19, or even lower EMA timescales?

Great piece of advice... I think this method is simple and will make money for those with the patience to do so.  I think I looked back some ways and outside of those stable periods where they pretty much cross back over each other on a slim margin, the method is pretty much tried and true.

Somewhere in this thread is a table with back testing results for various combinations of fast and slow averages.  10,21 happens to be one of the more profitable pairs.

I've observed same; that's why I usually wait for another candle afterward to confirm the cross.

And I also do not use the numbers suggested by Goomboo.  Sometimes I even change the timeframe I'm looking at (other than just hourly all the time), depending on other factors.  But overall I have to say, the simplicity elucidated by Goomboo's system has instilled a confidence in myself to develop my own personal trading system, and to stick with the game, and I find that as long as I stick to my system (really just A system, no matter how simple/complex), I keep profiting more than I would just buying (or mining) & holding.

There is definite value in OP here and I certainly appreciate it.

I don't think that is fair to say. The advice was indeed to use these numbers, but no one is forced to follow them blindly. Actually these numbers mean absolutely different things depending on the periods one chooses to trade in (ie. hourly/daily bars), something which Goomboo only shared his preference.


I am no expert in any of the fields but as far as i get them the times they cross each other should not be looked at as future predictors, but as synthethisers of the past. As i have observed them, even if they're, at a certain moment, very prone to cross this can change and the "very likely" crossing turns out to _not_ happen. Basically they smooth over the discreet highs/lows changes that happen over time. The higher ema value more significance is given to the precedent and less likely it is to be affected in a sudden market move (to which you may then become exposed).

This is not to say that there isn't any influence if everyone start using the same numbers under the same time periods. But one may achieve very surprising results with totaly different numbers too. Ultimately there is a pair of numbers _and_ period which will be more efficient than any other. My view is, Goomboo advice is generaly very solid not because it predicts what will happen but because it will safeguard from catastrophe.