What have I been doing this week?
Research is still on going.
I am happy with my Recognition Primed Decision model, that I discussed last week, in helping me decide more consistently what the context is and how I trade it.
My main area of focus this week has been looking at current market conditions and looking for clues when that could be changing.
I spent a lot of time research trades days in the Bund. In the past I did a lot of research, trying to predict trend days and did not managed to find a trade-able edge in these stats.
Before I go further I must stress that this research is in the Bund, and could / does not translate necessary to other markets.
I found stats that says that if we have 3 days that are the same / similar in terms of volume profile, then there is a 75% chance that the next day will be different, but that does not mean it will be trend day.
The next step was to look to see if there are any clues at the beginning of the session, that a trend day is likely to happen.
I looked at the volume/delta/range for the first 60 mins of the session and the first 120 mins of the session and compared non-trend days to trend days. Whilst there is a slight increase in these data points, it would be hard to tie these into a practical application of a trend day.
For example, the mode of the first 60-minute range of a trend day compared to a non-trend day is 2 ticks larger over a sample of 400 trend days.
The Open drive is a rubbish predictor of trend days in the Bund, both in the full and cash open. Whilst a lot of trend days have these, so do a lot of non-trend days. And actually, open drives are a great target stats wise.
At the moment, my best indicator is a failure of a good level (and the better the failure better the indicator), that normally involves a losing trade. So, my plan was to try and gather data to find clues that could indicate that a trend day is in the works.
My next step is to go through each trend day manually and looking for patterns and clues from a visual point of view. This is labour intensive and boring and could easily result in zilch but is the next logical step.
But I have done this before. I once printed out all these charts and studied each one, to help define my basic context model.
But I will stick on some sounds and get down to it after posting this.
Basic research and coding has been done. I am now in the testing phase, make sure that they send the right orders at the right time.
I cannot remember who, and I cannot find them in the comments section, recommended this blog on automated trading alongside manual context, which is exactly the approach I am taking. Whoever that was, thank you, it has helped consolidate my thoughts about where I am taking my trading.
This approach makes it more complicated, as I do not want the automated system on all the time, only when in an area of interest that aligns with my context.
So the trigger has to have its own button. Obviously one for buy and sell.
I had a few teething problems with not sending orders and then sending too may orders.
Testing this is quite slow, first I used a dummy trigger to test the order sending, and that was quick. I am now testing this using the correct trigger to see if they do what I am expecting them to do. Lol
This is all on sim at the moment. I think I would have choked when the during the first phase it sent 5 x 2 lots before I could even finish my slurp of tea, if this had been real.
Once I am happy I will start testing with small size on my real-life account.
Of course, all this does not affect the trades, which are auto fades with blind orders.
And there are a couple of instances in the order flow which I have not managed to work out, how to represent these as code. But I could spend the next few months tweaking this. I need to get it to a stage to take it live so I can get some feedback on how practical it is, etc.
I am still doing my trade plans and daily stats every day, to keep in touch with the market.
I did not post Friday plan, as I was working on the keel boat that I race during the summer and did not get back till very late.