Acceptable Accuracy Ratio for Profitable trading

Most people in FF like to see fables from from la fontaine like ''Open price %95 of time bla bla bla and ...... with no proof of existence.And indicator cartoons.
The pdf is all about probability,calculations and nothing to do with the existing perception of ''price action'' here in forex factory.If I upload the file existing crisis will stop in middle east,illegal immigration and feminine will stop in Africa, over all the pdf file will save the universe. Only joking.

Just upload it. There are serious researchers on FF too.

Alright champion, here is the file:
http://www.transfernow.net/37d7i1344jan

Thank you. I had already read the few posts you have written on FF so I suspected the PDF would be worth reading and it certainly is a useful report!

Thanks again.


It also supports comprehensively what I wrote here:

Hi.

Thank you for sharing the document.

Hi scherzi,
I'm not sure that what you ask is terribly relevant to profitable trading. The direction of tomorrow's candle is only important if your time frame is for next day only. Also, magnitude of move is more important than direction.

In my trading I really don't care what colour the next candle after entry is. I'm concerned that I'm trading with the dominant current flow, that I've picked a reasonable entry point for the pattern I'm looking at, that my stop is outside the wrong side of the pattern, and that the reasonably expected profit if it all works out OK is worth entering the trade in the first place. It doesn't matter what the next candle is, just that the trade overall makes sense to me.

Am I missing something in your question?

Daily is the timeframe I used for testing purposes, but it is not the important factor.

I was curious to know if konwing in advance the direction of the price is useful. As you mention, volatility is also another data; but I was talking just about price direction. There are many factors indeed to monitor.

Hi scherzi,

OK, you have my views then.
It certainly helps you feel more comfortable if trades immediately start going your way but my experience is the great majority of trades oscillate back and forth both sides of entry before making up their minds and decisively moving. You just have to get accustomed to fluctuating and fleeting intra-trade small profits and losses without worrying about them. To this extent, the next candle direction is irrelevant. That your method lets the profits run and keeps the losses as small as practicable is massively a more important consideration, I would say.

Hello again.

The daily prediction is not a system but a forecasting model. This is, it predicts in advance the next day's 2-day moving average direction.

Basically, to avoid overfitting what it does is taking n thousands days in the past to learn how the instrument reacts to different input values including several common technical indicators and how they change along time and other price-related input variables. Then it is tested by predicting the next one-thousand days in the future to determine its actual accuracy. And that's all. It does not tell what to do. If anyone is curious about the implementation, there are a few papers around that took a similar approach (not exactly the same). This one is an example https://arxiv.org/pdf/1605.00003.pdf. I was very focused on reducing the smoothing factor as much as possible, so at the end I got a 2-day SMA prediction that, at the end, had an average accuracy ratio of 75% for several instruments. Next day's price was discarded after correcting the algorithm because without the smoothing factor there was no prediction above 55% accuracy and that is the same as flipping a coin.

I have finally implemented a basic trading training tool using Excel to find out if a system could be developed from that information. My first thought, and I do not have almost time to test this stuff, is that the information given by the system is similar to the information I get just by observing price action by myself.

The issue is that the difference between the 2-day SMA and the price is big enough to allow prices to go in the invere direction of the SMA and still the SMA may move away from prices. So, at the end, even the 2-day SMA range is too big to be relevant enough to improve price action forecasting.

Another important lesson learnt along the process is that I have found that technical indicators have indeed relevance. The model also helps to understand price action itself for a novice.

The conclusion is that the model is not providing to me anything at this moment. I have refactored the Excel tool to just simulate end-of-day (or end-of-bar) trading so I can use actual quotes. If anyone finds it useful I have uploaded the tool together with over 400MB of real daily stock data I have extracted for Google and prepared to be used with Excel. It's very simple and only a few features are enabled so far: https://github.com/gonzsanc/Excel-Trading-Trainer. The Excel file is in the "src" folder. The link to the data (it's compressed) is on the README document.

Thank you for your insights and responses. I appreciate your interest.

75 to 85 percent. that is quite high and it's a good thing. but you also have to take into account the risk reward. the risk reward is crucial after the winrate. you can have a system that is profitable with 20% winrate but has a very high risk reward.