At this very moment Ray the not quite Random Robot is rapidly incorporating some AI into what passes for his neural network in the probably forlorn hope that he will be soon be able to understand both physics and the BTC chart.

Please give my best to Ray and tell him to keep the faith; there is beauty in randomness.

What if randomness is non-existent and we are too primitive to realize the algorithma behind what we call random. :)

Mehmet, I would argue that is exactly the case. Given the right algorithm and computing power, we could track any given molecule from its position starting 1,000 years ago to its current location. However, with our current state of technology, that isn't possible so some things, while not random, are effectively random.

Scott - Ray has asked me to pass his festive felicitations on to you and any "robots" you may know. He also idly enquired what you might make of this "shock news!"?

https://www.theguardian.com/technology/2017/dec/22/bitcoin-price-plunges-2000-12-hours-year-end-rally-fizzles-out

Just a not-quite-random walk around Wall Street?

You got it my friend

Ray has asked me to pass on this festive image of the BTC "year end rally":

https://www.linkedin.com/feed/update/urn:li:activity:6343967600791289856?commentUrn=urn%3Ali%3Acomment%3A%28activity%3A6343967600791289856%2C6350314171564986368%29

Great article but rather than just read about algorithms, consider experiencing their extraordinary potential trading Forex at RagingFX.com

Thanks Tom, I do trade FX. I don't mean to give away any of your group's secrets but the best "secret" I have found in trading in general but especially in FX - if you are in a trade and you are praying, you are over leveraged.

Trade small, preferably in the 2:1 area for FX (even 10:1 is too high in my opinion) and you will never blow out an account. When trading at 2:1, even the most stubborn of traders can recognize they are on the wrong side of the trade before they can do serious damage to their account.

I always thought of a algorithm as an exercise to well define one's discretionary trading abilities. If the model fails then maybe the method, concept or idea also had flaws or was not quite as well thought out as previously believed. The human mind has a convenient way of bridging right over missing information by making assumptions that algorithms demand clarity on.

Mark, you got the gist of the article spot on. We humans can fill in the gaps and extrapolate in ways almost impossible for an algorithm. The best we can do is get the main points close and little by little refine what we can.

Scott Boulette - excellent points That's why we choose to include an algo-specific Forex training course with our service.

...but that is not all.

Given that the algo does the heavy lifting for us, the course we offer can focus on aspects often ignored by traders, such as daily risk management and ratio-trading strategies based on market sentiment. ...plus we teach these principles as they apply to the algo, not just in general. My point being that its not just the algo that is in focus here but also all that comes with FX training too. ...definitely disruptive in more ways than one.

Check us out at RagingFX.com.

Tom, I agree that what many traders consider to be ancillary to their trading is actually more at the core of what makes a good trader than they may realize. Any education on these points will have a positive impact on the traders who take it to heart.

Scott, I liked the article, particularly the issue they made out of the vaguely or ill defined terms they had to modify frequently during their walk. Algos used for trading are seldom done that way, and are not generally modified on the fly. As I see it, the coders doing the algo know little about trading but lots about coding, and the traders oftentimes know little of coding and lots about trading. This vague set up is used to come up with trade algorithms more often than not. Of course management knows even less about coding, and sometimes even about trading, but want to have algos and AI operative so they can increase sales. The code done this way is not capable of doing what the traders want because the traders cannot explain what they want with enough detail to the coders and so on and so forth. A nice pickle to find yourself it.

Oscar, that is precisely the issue I intend to solve in 2018. As an individual trader who by luck happens to be a decent programmer and through work happens to know the exchange rules almost by heart, I can bridge the gap between the components you talk about. However (isn't there always a however?), trade, code, sleep - pick two.

I have come to the conclusion it takes a group to compete at the level we used to be able to achieve on our own.

Scott you are absolutely correct in this regard. A small group of coders that trade and traders that code may be able to clear up this obstacle. All the best for 2018

It comes to:

Know where you want to go and what you are looking to find.

Most of us (me too)think we know.

Until we put our thinking on paper and it looks confusing even for us not only to the Algo.

I explained my method to strangers and then I understood that I have no idea what the _____ I am talking about.

Anyway enjoyed the article thanks for bringing this to us.

Respectfully,

Eli Weiss

@Eli, you are quite welcome.

Traders can make money at scale so I always think I would rather know a good trader than a good coder, lol. Actually, this business needs both and as Oscar wrote above, if you can find the combination in a person or group, you can do great things.

Great...

Welcome to RagingFX.com

Excellent performance - what leverage do you use (i.e. 1:1, 10:1, 50:1)

The article doesn't say much other than the approach and simple set up, what it did have is the concept of what are you trying answer. That logic is consistent with any analysis. As a person who trades and a person who programs, the mystery or misuse of the term algorithm is somewhat irksome, but then I consider does it matter if they understand.

Todd Turney expresses a sentiment that I too have applied, keep the rule simple, have something that you can define, write it down, see if you can validate it and then figure out if you want to make it programmatic.

There will continue to be misuse of the terms AI and ML along with also by people who hear the words, think they understand and want to seem sophisticated. Then there will be some who do understand the terms and will focus on how to implement them when it makes sense.

@William - I have found there to be two major camps on the issue of algorithms (no doubt there are many more I am not cognizant of). I am in the camp of only programming algorithms designed to produce alpa (vs execution algorithms) where I have some basis for understanding the logic behind the algorithm. I know others who apply ML (which I consider a subset of AI) to a set of data and see what pops out.

My primary issue with the ML approach is that if you don't know why it works, you likely can't predict the conditions under which the algorithm breaks down.

I know this may have little to do with algos but it has a lot to do with understanding about how we trade. Like Eli Weiss said, you can trade successfully and be unable to explain what it is you do to be successful with your trading. I have a similar situation in a way. I am trying to better understand my trading methods. I have decided specific portions of the methods need to be put in writing explaining the more instinctive actions I take, jumps is what I call these actions. I have found that writing about jumps is really difficult, not to say useless. The real reason for the difficulty is the fact that the jumps may be largely subconscious in nature, and I just do not understand them well enough to describe them. It would be interesting if anyone, anywhere, could possibly be able to translate any such jumps (which I think all traders do one way or the other) first into writing and then into an algo that approximates what we actually do.

Oscar Cartaya most of the technical trading mothods can be compiled. I can try to help you to understand better your trading patterns

Shalom, I do not think you read my post properly, I act instinctively at crucial times and do well. I cannot explain why. Sometimes I do something and other times, in a similar situation, I do something else. This has absolutely NOTHING at all to do with technical trading methods which fix the response to indicator activity or market price activity.

@Oscar I think I understand what you are referring to because I went through a similar exercise several years ago when I decided to make the transition from hft to higher time frame strategies (still in the seconds to minutes time horizon).

I found that I was processing things at a lower level (subconsciously) than I was aware of. For example, the market might look "jittery" causing me to wait for a break and then retrace. Jittery came in many different forms and it took over a year of experimentation/coding to reduce that to a set of numbers usable by an algo. By reducing the overall sentiment (your jump) to its component parts and aligning each of those parts one by one so that they didn't contradict the whole and then reducing all that to code was a huge undertaking but well worth the effort.

Absolutely simple and ingenious term👌

Good article. What's funny is that many algorithms are referred to as 'AI' these days, simply because they're relatively good at meeting peoples' expectations. Applying that lens, my washing machine certainly qualifies as 'AI'!

@Janusz - that explains why laundry is such a mystery; I am just not that fluent in AI. For those who enjoy a good spoof, do a search on the magic laundry basket on YouTube. If you don't get the joke, there is not much hope for your current relationship, lol.

The theory of algorithms is quite developed during the XXth century. Certain specifics of the XXIst century is brought by intensive development of quantum algorithms. Algorithms narrowed for trading require Markov times, which I also name Dynkin-Neftci time in "Trading Strategies with Position Limits" https://arxiv.org/pdf/1712.07649.pdf . You may consider this and previous "The Wandering of Corn" https://arxiv.org/pdf/1704.01179.pdf as a complementary reading on trading strategies, when algorithms are applied for trading futures. Best Regards, Valerii

@Valerii - I have read the "The Wandering of Corn" and it is excellent so I am definitely looking forward to reading "Trading Strategies with Position Limits". Thank you for posting those links.

I have been considering whether a classification system could be set up to describe price action regardless of time frame and use that in conjunction with other data analysis to match the current regime (i.e. today, the last 4 hours, the last 100 milliseconds) with past regimes as a predictive measure.

Possibly this could be set up as a sort of semi open source group with anyone willing to contribute able to share in the results. In some ways this would mirror how academic articles are written (which is why your links above reminded me to mention this thought). Have you seen anything like this tried and if so, did it work?

Scott, thanks. Ability to describe price action regardless of time frame assumes self-similarity. A theoretical Brownian motion is self-similar and zooming in or out creates images, where one cannot say about the time frame until [s]he is given a hint. Futures prices are discrete. Big positions and/or leverage magnify discreteness to an extent, where self-similarity stops working while zooming in. This is experimental fact. Not only a view of a chart changes but other properties analyzed statistically become different. Relationships between discrete, often equidistant, price levels and irregular waiting times, time intervals between ticks, become important. Best Regards, Valerii

@Valerii, I agree 100% but you put it much better than I possibly could.

@Valerii - I have started reading your paper "Trading Strategies with Position Limits" and I already know I am going to like it. Why? Because of your spot on observation - "Technical analysis is anathema to the academic world".

I can now highly recommend anyone interested in the higher frequency trades and especially any that depend on a Markov process read "Trading Strategies with Position Limits". Don't let the math intimidate you; skip it if it does. In fact, if you read nothing else, section 10 alone is worth the effort.