- Search Energy EXCH
- 28 Results
- beachboy44 replied Jan 21, 2016
It will be interesting to see if building examples in this way can reduce the number required when using a multiple trade outcome mechanism. For larger variation spaces, most of the highly linear systems I have found have a learning period of ...
- beachboy44 replied Jan 15, 2016
You are right that the example shown here is nothing to write home about, but I'm always interested in looking for extra layers of diversification. From what I understand, having these 1H systems only trade at a specific hour was vital to their ...
- beachboy44 replied Jan 14, 2016
Wow, this opens up a whole new avenue for 1H ML systems! I suppose the first question we should ask is how much more computationally demanding this method is than building sequential examples at specific hours. We no longer have the trading hour ...
- beachboy44 replied Jan 7, 2016
I think this is a good idea! When I was experimenting with using low DMB, historically profitable price patterns as inputs, I thought it would be nice to not be restricted to only those pairs for which we have such price patterns. This gives us a ...
- beachboy44 replied Jan 6, 2016
Thank you - the more testers the merrier! In particular I am interested to know how the multiple trade outcomes affect DMB. On the one hand the search space is reduced as you no longer have to do a brute force optimization over the exit mechanism ...
- beachboy44 replied Dec 25, 2015
F4 framework - url
- beachboy44 replied Nov 8, 2015
No I haven't done any research on them, I was just wondering whether or not to add it to my to-do-list. But I think for the reason you posted, and since there are many simpler things I haven't tried yet, I will put it on the back burner for a while ...
- beachboy44 replied Nov 7, 2015
I would say it is not worth it at all considering the amount of time it takes to measure DMB for NN systems - even more so if the system is going to be traded within a large, uncorrelated portfolio. Just out of interest, have you done any ...
- beachboy44 replied Nov 6, 2015
This thread is probably a bit too advanced if you are just starting out. url has a lot of great information, and takes a very scientific approach to trading the forex market. I'm not saying that this is the only way to trade, but it may help you ...
- beachboy44 replied Nov 6, 2015
Great! It will be interesting to see whether or not my results had any underlying truth; like I said before, pretty much everything about my previous setup was suboptimal. Just in case you find yourself having to compute the geometric mean for ...
- beachboy44 replied Nov 2, 2015
Thanks for the reply. I did some research last year (before joining Asirikuy) using software designed by a friend. I used SVM, NN and various 'simple' ML models. Inputs: The inputs were derived from the currency strength calculations found in the ...
- beachboy44 replied Nov 1, 2015
I think that the 1H timeframe gives a good compromise between computation time/cost whilst still having the ability to take advantage of the daily volatility cycles in the Forex market (unlike 1D). I suppose that going down to the 30M TF may help to ...
- beachboy44 replied Oct 31, 2015
Yes, that's correct.
- beachboy44 replied Oct 24, 2015
I am not allowed to post the currency data, as it is licensed to paid members only. Unfortunately I do not have the code for the system, and I don't have the time to code it myself at the moment. algoTraderJo may be kind enough to provide the trade ...
- beachboy44 replied Oct 23, 2015
There is a plethora of H1 systems within this thread; in fact, the majority of systems do not trade on the daily timeframe. Look at the recent post (#720) which provides an M30 system. In my opinion, this is about as good as a backtest can be - a ...
- beachboy44 replied Sep 24, 2015
This is the only machine learning system that I trade/test live. You are right to be sceptical, but I can assure you that I am not trying to fool you - I'm not selling you anything I am merely showing you that it is possible to be profitable ...
- beachboy44 replied Sep 24, 2015
This is precisely why we measure data-mining bias. I have attached a screenshot of the performance of a EURUSD D1 system which has been developed using some of the techniques within this thread. image
- beachboy44 replied Sep 16, 2015
Hi SteepCurve, I don't know if you are aware of url , but it includes some great articles that describe exactly what data-mining bias (DMB) is in the context of this thread, and also how you can measure it. I probably won't be able to explain it as ...
- beachboy44 replied Aug 19, 2015
Yes, I think that storing the long/short decisions is somewhat necessary for the majority of people to be able to measure data-mining bias for ensembles, due to the computational complexity involved otherwise. Thanks for posting your idea! Also, ...
- beachboy44 replied Aug 19, 2015
When constructing a 'parameter ensemble' with 2 systems, we still train two seperate models on each bar/trading hour - we do not combine the inputs. Constructing a 'parameter ensemble' is done in the same way as the search of the original space, ...