A Relic Trading On Feeling

[digitalnomad]

A very good comment. the part though about giving away gold is inaccurate. The open source software development community is that.

I beg to differ. I see it more like the blind leading the blind. Having read through many posts over at the Quantopian community, I couldn't help but chuckle at how clueless they were about market mechanics, microstructure, and execution.

 

[Sprout]

I beg to differ. I see it more like the blind leading the blind. Having read through many posts over at the Quantopian community, I couldn't help but chuckle at how clueless they were about market mechanics, microstructure, and execution.

I agree, as it applies to trading.

When it comes to developing software and working through different language semantics, there is nothing like receiving an workable answer to a troubleshooting question. I'll receive that gold every day.

 

[digitalnomad]

I can agree with that. Definitely some smart guys there.

 

[ironchef]

Are you saying FT and London Guardian chess columnist Leonard Barton is wrong. Kasperov just published a book explaining the psychological reasons he lost to Deep Blue in 1996. Title is "Deep thinking Where Machine Intelligence Ends and Human Creativity Begins." Basically, Kasperov was not playing against the Deep Blue computer. He was playing against a team of talented chess players (not world champion) and talented programmers who expressed their knowledge by writing a computer program. Kasperov says he was effected by the psychological tactics of the team.

Could it be possible Kasperov was a sore loser making excuses?

 

[Slartibartfast]

Could it be possible Kasperov was a sore loser making excuses?

At the time perhaps he had a point, deep learning had limitations e.g. it largely regurgitated past chess games however https://www.wired.com/2016/01/in-a-...gles-ai-beats-a-top-player-at-the-game-of-go/ last year there was a big breakthrough.

Deepmind trained against itself and hit new levels of proficiency. Apply this back to chess and Kasperov would not have a chance.

 

[ironchef]

I am not sure if this thread is the right place to post this. My reason to say deep learning and neural networks are hype is that that those algorithms run into combinatorial explosion that human thinking does not. AI was first falsified in 1972 by Cambridge professor James Lighthill (search for "The Lighthill Report"). I have written a paper that shows Lighthill's computing limited by combinatorial explosion applies to current AI ("A Popperian Falsification of AI - Lighthill's Argument Defended" ArXiv:1704:08111).

It takes some web searching to see why deep learning does not work for automated trading. In order to make neural network searching efficient enough to run on physical computers (avoid the combinatorial explosion), deep learning searching uses something called regimes. AI researchers are somewhat disingenuous in discussing regimes. I think they are just ways of pre assuming the probability distribution of what is learned. Try searching for meta learning and regimes. Deep learning works by assuming no black swans. It easy to design a sure fire option selling algorithm if say stock price distribution can only be Gaussian.

The problems with deep learning and neural networks were understood in the 1950s by John von Neumann who invented neural network automata. Neumann wrote (quoted in "John von Neumann and the Origins of Modern Computing" by W. Aspray, p. 321):

"The insight that a formal neuron network can do anything which you can describe in words a very important insight and simplifies matters enormously at low complication levels. It is by no means certain that it is a simplification on high complication levels. It is perfectly possible that on high complication levels the value of the theorem is in the reverse direction, namely, that you can express logics in terms of these efforts and the converse may not be true."

Your post and comments are too technical and complex for me to understand. Are you saying AI cannot match the human mind in trading or are you saying the opposite?

Thanks.

 

[ironchef]

At the time perhaps he had a point, deep learning had limitations e.g. it largely regurgitated past chess games however https://www.wired.com/2016/01/in-a-...gles-ai-beats-a-top-player-at-the-game-of-go/ last year there was a big breakthrough.

Deepmind trained against itself and hit new levels of proficiency. Apply this back to chess and Kasperov would not have a chance.

What about in trading?

 

[Slartibartfast]

What about in trading?

In a rush out the door so I know this is unpolished.

I am confident ML/AI can outperform humans in trading soon (the elusive issue of big-data derived "context" is cracked in 2017/18 probably) however trading is not a binary win/lose result like two player chess or go.

Under these circumstances and accounting for the casino market rules (if they win every game, people stop playing etc.) I believe the upper echelon of human (mostly) retail will be fine, just a little squeezed but they will adapt.

The CME and others will just change the rules to regulate the "Syndrome" effect.

"When everyone is super, no one will be" etc.

They need a healthy market, the perception of that anyway, "financial gas-lighting" to get retail and others to play will just get better. Some token humans will pied piper in the "liquidity providers" for years. Ok maybe a bit cynical there.

Slightly off my point but it applies to the <95% level retail traders also (who think they are being clever investing their own money), an interesting read. https://rpseawright.wordpress.com/2016/05/04/financial-gaslighting/

 

[Sprout]

Hello.

Kind of tired reading of all the quant/AI/ML/coding hype everywhere and that if you're not doing it, you'll be out competed. I'm an old fashioned brain and eyes (along with bits of guts and groins) trader. So just putting it out there and let's see in a couple of years time if I get eaten up by the machines and codes.

View attachment 175636
View attachment 175637

I'll post this year's results after it's updated for the close of this week.

Quick summary because most TLDR
- predominantly in futures & FOPs (ES & FX 6's with bits of GC, CL)
- day trading preferred
- favorite character gregor the mountain
- mostly chart reading

What have you (and other successful traders) discovered in your travels that augment "feeling" as it applies to being on the right side of the market?

 

[vanv0029]

Your post and comments are too technical and complex for me to understand. Are you saying AI cannot match the human mind in trading or are you saying the opposite?

Thanks.

I am saying that people need to be an integral part of trading systems using judgement and superior pattern recognition of the human brain. I think Linda Raschke has been saying this for quite a long time. Obviously back testing (data mining) makes sense. My criticism is aganst people who claim that new AI methods will improve to the point of eliminating people from trading. This is more technical but the AI hype algorithms such as neural networks (superceding human brain neurons say) are no better and less efficient than Monte Carlo simulations.

I also think probability (not trying to find statistical significance that is good) is no better than computer brute force enumeration searching. Argument is that the P=NP problem is just an artifact of the wrong Turing Machine computation model. For the von Neumann MRAM (random access memory abstract machine with unit multiply) P is equal to NP or put another way there is no difference between determinstic enumeration programs and non deterministic probabilistic guessing programs, i.e. programs should use smart solution space searching using enumeration and MRAM RAM table look up something normal Turing Machines can not do. I have written a paper on this that I am trying to win the Clay Prize with "Philosophical Solution to P=?NP: P is equal to NP" (arXiv:1603:06018).