- Search Forex Factory
- 412 Results (3 Threads, 409 Replies)
- algoTraderJo replied Aug 21, 2017
I rarely have time be online now (kids, ya know) but every time I do I read your posts. Reason has very few voices here and you play a very important role. I am amazed at how many times you have been able to respond patiently to threads about "this ...
- algoTraderJo replied Aug 21, 2017
How are you controlling data mining and curve fitting bias in this project?
- algoTraderJo replied Jul 21, 2017
You're on the right path mate! Keep at it!
- algoTraderJo replied Jul 21, 2017
This is the cocolizo67 F4 system using an SL with an ATR multiplier: #include "EasyTradeCWrapper.hpp" AsirikuyReturnCode runMondayFridayEA(StrategyParams* pParams) { AsirikuyReturnCode returnCode = SUCCESS; double stopLoss, atr; int ...
- algoTraderJo replied Jul 19, 2017
And with a 100 pip stop... image
- algoTraderJo replied Jul 19, 2017
I don't post much now (got two kids so time is a bit short) but I LOVE when someone does a quantitative contribution here at ForexFactory (very rare mates). Congrats to cocolizo67 for putting his work here and sharing it with the community. I ...
- algoTraderJo replied Apr 16, 2017
Kids are keeping me busy but I think I'll have time for some more ML fun this week!
- algoTraderJo replied Apr 16, 2017
This is a toughy. You want to maintain the symbol correlations present but you want to randomize all the data. The solution is not complicated, you sample and always pick the same index for the different symbols at the same time. If your symbols ...
- algoTraderJo replied Mar 29, 2017
Cool, check this recent paper out.
- algoTraderJo replied Mar 29, 2017
This is an algo I like, it has been trading for a bit more than a year now for me: image It uses a two algo linear classifier ensemble. I post the csv for this system, if anyone wants to explore it more.
- algoTraderJo replied Mar 16, 2017
If any here are using pKantuML for ML strategy mining we can post the csv files for the systems we find here and share... Anyone up for it?
- algoTraderJo replied Mar 16, 2017
1. What specific regularization? There are many ways to do this. If you have tested in F4 post the specific algo and I'll reproduce. 2. Atm I only trade systems where the performance is positive for the past 3 years and the R is above 0.9 for that ...
- algoTraderJo replied Mar 16, 2017
If you modify the I/O functions to use renkos in F4 pKantuML works as well. The only problem is when you use OHLC bars below 15M, it becomes much much much slower. When you go from 60M to 30M it's not just 2x slower but around 3x slower, from 60M to ...
- algoTraderJo replied Mar 16, 2017
Also if you have access to pKantuML you should definitely use it. I presently can backtest 60M, 30Y tests at 0.02 ms per system using my GTX 1080. That's 20K backtests per second. You can write any custom ML algo you want and just add it in the ...
- algoTraderJo replied Mar 15, 2017
I also have some results for this same setup using a K-NN (K Nearest Neighbor) algo. Will post soon.
- algoTraderJo replied Mar 15, 2017
This is what happens when you change the bars used per example with this renko based system. image There is a sweet spot for the BPE between 3 and 5, below or above that we see a performance drop.
- algoTraderJo replied Mar 14, 2017
Time differences in renko as ML inputs can lead to some improvements in results. Here is a similar result from USDJPY trading on renko bars derived from 1H OHLC bars (trading only on the 1H open). SL_ATR_MULTIPLIER=1.2 BARS_IN_FRONTIER=60 ...
- algoTraderJo replied Mar 14, 2017
How long have you been trading? If you have traded for 4-5 years and it has been working well through all this time then don't change it. If it ain't broke don't fix it. If you haven't traded for this long then you'll have to wait to see enough ...
- algoTraderJo replied Mar 13, 2017
When using renkos to train ML algos the data from the returns becomes almost meaningless as we only get info about direction from them. Bars are all the same size so all the ML is getting is which are up and which are down . What we have won by the ...
- algoTraderJo replied Mar 13, 2017
Using time differences for predictions is not difficult. While you modify the I/O functions you can change the line to use double the number of inputs like this Data<RealVector> inputs(period,RealVector(barsUsed*2)); Then modify the loop that writes ...