Algorithmic Quant Trading (Machine Learning + Stat-Arb) 25 replies
Forex with Machine Learning Software Project 1 reply
Machine Learning + Retail Forex = Profitable? (Quant) 1 reply
Potential new machine learning style software. 79 replies
My most recent advancements into machine learning 16 replies
DislikedHi AlgoTraderJo, I am sure you will get to it, but what is the result if you try some of these algorithms on shorter timeframes, e.g. 1 hour?Ignored
DislikedLet us make this thread a little bit more interactive Post some ideas you would want me to try to improve the machine learning results I have posted so far (predictions on the daily timeframe on the EURUSD), I will try them out and will post the results here!Ignored
DislikedReally nice analysis you're doing here algoT. Enjoying the thread a lot. Are you using a stoploss in the recent ensemble models? Are you going to try using an artificial NN as the classifier tools in future? Even though they're a pain to parameterize, they do work well when they work. Thanks.Ignored
Disliked{quote} I have proposed earlier in the thread that the cumulative return curve or drawdown could used as a secondary input to the algorithm. You can use machine learning to analyse your cumulative return and make predictions around it. If successful it could help to better filter trades and only trade when you predict a positive return or series of returns.Ignored
Trying to get an additional advantage from the trade return series for these strategies is therefore something you cannot do successfully. This does not mean that you cannot do this for other types of systems but clearly not for these strategies. Do post any other suggestions you might have!
Disliked{quote} These suggestions do not work (already tried them extensively) due to these reasons: There is no constant relationship between trade returns. Systems generally have periods of large consecutive profits/loses and then they can have periods where they are evenly distributed, etc. There is no constant or at least clear relationship between returns that you can exploit using statistical learning. At least in the machine learning strategies I have posted here. These strategies have a high reward:risk ratio, the average profit is more than twice...Ignored
DislikedWhen you add new information that contributes with predictive power you actually get better results. The graph below shows you our K-NN test using 2 bar directions as input and a second test using 18 bars as inputs. As you can see the addition of 16 new inputs adds enough information to improve results altogether. {image} It is also interesting to see how the distribution of accuracy in predictions has changed significantly and our two systems take opposite directions through several parts of the simulation. This means that we may be able to obtain...Ignored
library(tseries) dailyoanda = function(symbol="EUR/USD", start="2010-01-01", end="2010-12-31", provider="oanda", quote = c("Open", "High", "Low", "Close")) { return(get.hist.quote(instrument=symbol, quote = quote, start=start, end=end, provider=provider)) } getyears = function(symbol="EUR/USD", startyear=2000, endyear=2013, provider = "oanda", quote = c("Open", "High", "Low", "Close")) { year = startyear hist = NULL while (year <= endyear) { sy = paste(year,"-01-01", sep="") ey = paste(year,"-12-31", sep="") dat = dailyoanda(symbol, sy, ey, provider, quote) if (is.null(hist)) { hist = dat } else { hist = c(hist, dat) } year = year + 1 } return(hist) } #quote = c("Open", "High", "Low", "Close") quote = c("Close") EU = getyears("EUR/USD", 1998, 2014, "oanda", quote)
DislikedHi ATjo I'm still trying to reproduce the linear regression results. My parameters are: Currency of the account: USD Starting balance: $100000 Money Management: 1% Timezone: Europe/Paris,Berlin Price used: open price, (bid+ask)/2 Stop loss: 0.6 x ATR20. ATR uses the daily candle in GMT time and excludes the week-end candles (including the W-E changes nothing). Take profit: None Number of price return per sample: 2 Number of samples in window: 100 Strategy: If no position is open, open in the direction of the forecast. Else if the current position...Ignored