Machine learning in trading: theory, models, practice and algo-trading - page 119
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
) No one is arguing that there are dependencies! They argue about profitable strategies.
A simple tree will give 65-70% of correct candlestick color definitions - you can't use it. Even in binary trading it is too small a preponderance.
It is statistically more difficult to make money on binary than on candlestick color. The accuracy of 65%-70% in binary is just endless jerking its balance up and down. And for candlestick color for timeframes starting from H1 it is already profitable.
For some reason I try to use D1, my experience shows that only >55% accuracy is enough for minimal profit, spread influence is very small there.
Regarding my experiment update.
I've been struggling to find a way to get good "out-of-sample" values from "in-sample" data. But the committee gathered on my heuristics gave me poor results. About 10% per year at a 50% drawdown.
I see the next steps as this:
Until communication. Psst.
1) Earning on binary is statistically more difficult than on a candlestick color. The accuracy of 65%-70% on binary is just endless jerking your balance up and down.
2) And for candlestick color for timeframes starting from H1 it's already a profit.
3) For several reasons I am studying the work with D1, the experience show that for a minimum profit there is enough accuracy of only > 55%, the impact of the spread is very small.
1) Do not believe the children. 65-70% is in his pink snot. It is a fitting on the history. There are no such numbers on reality. It's all in the dream of an addict. You have to communicate at a higher level to say that the model honestly gives out well say 55% without taking into account the spread.
2) 57-60% is enough for H1 to bypass the spread and come out in profit (see here, already calculated:https://c.mql5.com/1/37/teaser2.JPG)
3) On these horizons 53% already counts. But the higher the horizon, the lower the accuracy.
Making money on binary is statistically more difficult than on candlestick color.
How's that?
The candlestick color prediction is binary trading. If you have the opening price of the candlestick color and the candlestick color prediction - this is the classic "above-below" binary.
What difference does the timeframe for the binary matter? The only thing that matters for a binary is the overbetting. If you have 70% prediction accuracy and the payout is 80% of the bet - go for the millions.
1) Don't believe the kids. 65-70% is in his pink snot. Adjustment on history. There are no such numbers in reality. It's all in the dream of an addict. They should communicate at a higher level to say that the model give out well say 55% without taking into account the spread.
2) For H1 57-60% is enough to bypass the spread and come out in profit (see here, already calculated:https://c.mql5.com/1/37/teaser2.JPG)
3) On such horizons 53% is already set off. But the higher the horizon, the lower the accuracy.
Fitting on history for a simple classification tree?
No more questions for the "schpecialist".....
Making money on binary is statistically more difficult than on candlestick color. The accuracy of 65%-70% in binary is just endless jerking your balance up and down.
How's that?
Trade on the exchange, and the spread will help you!
Fitting on history for a simple classification tree?
No more questions for the "schpecialist".....
Keep fitting with your tree. Draw in your imagination 70% accuracy.
Then don't forget to tell me how much money you lost. "Specialists" will appreciate it.