Machine Learning Trading Algos

Hi All,

To introduce myself, I have been trading since early 2000 and have gone through several phases. I've traded various styles and methodologies some with success and some without. This could be partly due to my ignorance and inability to understand that particular style of trading. Overall my experience has taught me that no system is good forever and always needs tweaking. But the tweaking takes time and requires additional learning curve. Trading part time means this learning comes at a cost, i.e you are not making money when learning. Once you have learnt something new, applied in demo and then applied in live, its time to tweak again.
This constant cycle is something which I have been struggling with due to time and effort. And I believe this is the reason why most traders fail.

I am a programmer and an IT person, so I always wanted to automate my trading strategies. Some worked and some didn't. I started experimenting with machine learning, AI and stocks. I have been working on some algorithms for last couple of years which have proved to be useful and I am now spending more time in this field for the simple reason, if I can teach a computer to learn then I don't have to. In simple words, the constant cycle of learning I explained above is now done in hours rather than in days/months.

I have data dating back to 1993 across various pairs and other data points. Overall I have data which is going into multiple TB that is used for predictions and modelling. I want to start this journal to see how I progress through this latest machine learning (or how I don't progress). I will not be spoon feeding the traders here but rather sharing my thoughts on latest output from my models. I hope the output of these models can help supplement your own trading models.

These models are not 100% right and I will express the outcomes in probabilities (percentages) which help confirm your own trade setup.

I will not be posting exact entry/exit criteria but rather that is for you to decide. I do not want to give away the whole trading setup in first post, but as I progress in making the models more mature I will be posting more details.

I will aim to post updates at the start of new day which is supposed to be 5PM ET (10 PM UK TIME). Trades should be set ahead of next day.


Keywords to keep in mind for current EXP1_V1 model:

Buy at Pivot - Try buying at 50% Fib of yesterday range. Expected day close to be higher than open.

Sell at Pivot - Try selling at 50% Fib of yesterday range. Expected day close to be lower than open.

Sell Above Pivot - Try selling above pivot. Expected day close to be lower than open.

BUY Below Pivot - Try buying below pivot. Expected day close to be higher than open.

Some keyword explanation:

1. All numbers are expressed as percentages (100% being max).
2. Date - the actual day being predicated.
3. Profiled Days - The higher the count the more probability of meeting the expected outcomes.
4. 2x D-high and D-high are some of my support/resistance equations that have been adapted over the years. I will explain these in details later on.
5. Currently trading model is optimised for hitting 50% Fib of yesterday's range or as I call it pivot.
6. Close high - today's close is likely to be higher than open.
7. Close low - today's close is likely to be lower than open
8. Rest will come in later posts.

I agree with your supposition that markets change and tweaking needs to happen but 1 constant is supply and demand. For me the only tweaking needed is to better identify the areas of s and d. Anyway good luck in your venture

thanks bud.... isn't understanding S&D same as saying you need to understand "Einstein's theory of relativity"...the more you read the deeper it gets. It easy saying go and learn S&D but when you starting poking that's where all the methodologies and trading styles come into play?

Interesting approach. Will be following your journey. Would love to see how your machine learning develop in the future. Waiting to see the future posts with more details as you mentioned.

Good luck.

Predictions for 23rd July.

target is 50% point from friday's range.

For example an entry for USD-CHF would look like this...

all four predictions above are going in right directions. Example of EUR-AUD Above

Close to hitting CHF-JPY TP.... see above.

Predictions for 24th July.

Hi, buddy,

I am interested in the setup work of your machine learning model on live trading.

If you do get a good machine learning model, could you please share us how you put it in live trading?

Thank you.