Pattern Recognition -- Any Good?

I am a visual trader and use weekly charts for this failed reversal setup. 3 EMA to determine trend. Wait for weekly close outside prior week's extreme as a reversal signal. Set order outside of that candle for continuation of trend (Failed reversal of indicated trend) Works in UP or DOWN trend.

Here's a chart of CAH. 8 trades, 6 winners. I do it with ES, CAH and a few other names in orde to have adequate number of opportunities.

upload_2020-2-1_16-14-47.png
upload_2020-2-1_16-14-47.png
 
upload_2020-2-4_19-58-10.png


ES, same setup. Last 3 entries, stay small enough to have WIDE stop and let it ride until you get a close on the wrong side of the black dotted MA. This works for me, I hope it does for others because I'm not a master trader, actually have a real job-LOL

Trade 1; 169 points.
Trade 2; 56 points
Trade 3; we'll see.
 
DTB2 said:
Trade 3 currently up 90+. If closed, that's 215 ES points since May. How many more do you need.
More...

@DTB2 Bravo! No for 3 trades, or for that matter -- regardless of how many -- Your trade system is a Gem! Thanks for sharing.
 
DTB2 said:
I am a visual trader and use weekly charts for this failed reversal setup. 3 EMA to determine trend. Wait for weekly close outside prior week's extreme as a reversal signal. Set order outside of that candle for continuation of trend (Failed reversal of indicated trend) Works in UP or DOWN trend.
More...

Simply stated. Understandable. (and It Works). Multiple Tickers will provide you with enough opportunities. So thanks again, @DTB2
 
i am a multidimensional pattern recognition expert. put simply what i do is i feed in a tensor of data (10k by 10k by 10k by 100k) at each so called timestep. these are not timeseries of old data points they are new possibly independent readings. they are the sensors she uses to make decisions. she then tells me the price of gold, always correctly. at no point in training her did i tell her the actual price of gold (not in the tensors) or the historical price. why? because its very like the actual price of gold. if i feed her the tensors in the wrong order will she get the right price? yes. why? because she does not rely primarily on a convolution of fourrier transforms and has no sense of continuity in time. this is for long termism and has no place in the making of very fast profits.
 
From my observings in the last 3 years, patterns do work. But context is the most important. If the market is bullish, if it looks like it's going to break out and some of the stocks are already popping up and front running, and you see a nice inverse head and shoulders looking to breakout... It's likely it. It also works best when market is trending. When the trend is strong, just draw the simple patterns and the breakouts and their retests work like nothing else.

I always draw the patterns, i rarely ever trade them, but i often trade them if they make a fake breakout in the counter trend direction. That's some good stuff.

But keep in mind, market conditions triumph everything.
 
i hear they can have external memory in addition to what is encoded in the weights. however my focus is very short term dislocations, i cannot design a neural net to work on this time scale if i feed it timeseries and and pass along the layer. have tried training it using aggresive stochastic gradient descent to try to batter it. i see like knocking a kiddies visual cortex out early doors he can learn to use other cells, well not this kiddie it seems to basically assume the price is the last price. i would add i tried alternative activation functions that are less smooth. pointless. best is a knockin so nothing then a binary jump then linear. calls and puts should both be used though.
 
Last edited:
TommyR said:
i am a multidimensional pattern recognition expert. put simply what i do is i feed in a tensor of data (10k by 10k by 10k by 100k) at each so called timestep. these are not timeseries of old data points they are new possibly independent readings. they are the sensors she uses to make decisions. she then tells me the price of gold, always correctly. at no point in training her did i tell her the actual price of gold (not in the tensors) or the historical price. why? because its very like the actual price of gold. if i feed her the tensors in the wrong order will she get the right price? yes. why? because she does not rely primarily on a convolution of fourrier transforms and has no sense of continuity in time. this is for long termism and has no place in the making of very fast profits.
More...
Any books I can buy/read to learn about multidimensional pattern recognition, tensor of data, convolution of Fourier transform and long termism?

Thanks in advance.
 
i will answer any questions on the above. i would not recommend a book . a fourier transform is a long termist solution however if you are a beginner you can for your first project use it on the vix, it can do this very well, and there are good reasons why it should be quite good. you can use the simple one no need to decay these amplitudes because the function we are appoximating goes on forever and are extremely stable. word of advice: there are some very long term periods so use millions of terms. there are loads of programs you can buy to perform it. The idea is very simple: cos and sin are a so called orthogonal basis. as such we can write a function as a power series wrt to this basis and the coefficients for a given frequency are unique. Skill required:integrate a dummy on the unit disk, find the zeroes of the sin function, realise that integrating between -pi and pi is an inner product for this hoax.
 
You must log in or register to reply here.
Back
Top