Algorithms and Trading Systems based on Chess game strategies - page 5

 
laplacianlab:

I am not clear about your point number

3) Don't forget that the main result here must be chess tactics and strategy ideas that we can code 

Laplacianlab, I totally agree with you that we can't lose the focus of talk about Chess tactic and strategy abstract concepts in general. So go ahead about that.

Why impossible? We can test and reverse engineer the chess tactics and strategies to create such algorithms  (as rule 3 states), if we have a complete system, or just create one from an abstract concept (the working stuff by now, that you are asking for we don't lose).

In my point of view, the system idea is a complementary way to build something more realistic and automatic, that will join all this.

And the phrase "playing against the market" is just a metaphor, we all know that this is a fictitious game and that we are far away from some day have ROI on this way.

But the dream is free ;-)

 
figurelli:
Laplacianlab, I totally agree with you that we can't lose the focus of talk about Chess tactic and strategy abstract concepts in general. So go ahead about that.

Why impossible? We can test and reverse engineer the chess tactics and strategies to create such algorithms  (as rule 3 states), if we have a complete system, or just create one from an abstract concept (the working stuff by now, that you are asking for we don't lose).

In my point of view, the system idea is a complementary way to build something more realistic and automatic, that will join all this.

And the phrase "playing against the market" is just a metaphor, we all know that this is a fictitious game and that we are far away from some day have ROI on this way.

But the dream is free ;-)

Ok,

I have no idea how real chess algorithms are programmed, so, for now, I would start modelling the following concepts (UML classes, or whatever) for this fundamental event-based EA:

Pieces

Long term economic events (these are markets' long-term breaths)

  1. Economic collapse
  2. Public debt crisis in Western countries
  3. China's growth in the coming years
  4. Social movements in Western countries
  5. The next mini ice age
  6. The influence in people of upcoming Austrian-School ideas
  7. ...

Short term economic events (these are the news)

  1. EEUU
  2. Europe
  3. China
  4. Australia
  5. ...

All the above are supposed to move currencies, commodities, etc.

EA's Cognitive Habilities

Strategy

  1. Patience
  2. Initiative
  3. OpportunityCost
  4. ...

The above determines how the EA behaves to what's happening in the market.

 
Laplacianlab, I like this approach, however can you explain more your idea?

For example, how this model will became an algorithm for real trade?
 

Chess is a game of complete information, both players can see all the other players moves and pieces unlike in a market. Chess has a limited amount of possibilities where a market is infinite. I would be more inclined to liken trading a market to a game of Poker. Although Poker is also quite different from trading, there is an added sense of randomness that both trading and poker have as well as them both being games of incomplete information. 

One thing I learned in school was (back then) that the computer could not solve the game of chess as there are too many possibilities, instead the computer must look ahead to all the possibilities that it can for as many turns as it can and then score the outcome of all those possibilities and choose the move with the highest score of worst case scenario of that particular move in that particular scenario. I do derive one of my MQL programming techniques from what I learned in school about programming a chess computer, namely, the technique of scoring possible entry or exit (or other action) and only entering or exitting once the threshold is met in the score.

Even in the scenerio of programming an ea, one would not/could not score the worst case scenario of a given action instead one would give higher scores to things that increase the probability of positive results.

"Thinking ahead" in chess can be likened to back testing although of course they are quite different over all. 

 
bendex77: Chess is a game of complete information, both players can see all the other players moves and pieces unlike in a market. Chess has a limited amount of possibilities where a market is infinite. I would be more inclined to liken trading a market to a game of Poker.

I totally agree with this guy. Interesting thread tho. I didn't say anything because I didn't want to be mood_killer but those were my taught exactly.

I've liked playing chess growing up mainly because my father always played with his friends. They would sometimes talk about how deep someone's move perception was. Meaning how many moves into the future this person could predict good move. Modern computers can make a pretty lengthy depth_perception compared to humans; but Super-Computer vs another Super-Computer will always end in a draw. They'll always play the most efficient moves which leads to a draw.

The closest I came to relating Forex to chess was in my early_days when I learned about candle_sticks. I taught why not equate different candlestick pattern to different ranking of pieces on a chess board, but that was about as far as I went. Later it came as no surprise that I equate trading to a game of poker despite my efforts to try equating trading to Blackjack for the same above reasons. Blackjack has a limited amount of possible outcomes like only 52 cards are gonna come out of that deck & if only 4_cards are remaining and no A's has been played then all remaining cards must be A's. The Forex-market has no such thing and I like the way Alain described it earlier.

angevoyageur: On each turn there dozens of possibilities of movement in Chess. At each time they are only 2 possibilities for the market to move up or down.

True the market can also go sideways. Or a price may not necessarily change on next bar. But the simplicity and complexity of the market may come even closer to another game .... flip of coin. <- And this, allot of people do-not want to accept ... not even me :) 

 
angevoyageur: On each turn there dozens of possibilities of movement in Chess. At each time they are only 2 possibilities for the market to move up or down.
Not so simple. Even if you consider a simple tick to be a "turn," the market has other dimensions to consider such as: time (when will it tick up or down?),  amount or price (how far will it move up or down?). Even just these two dimensions alone give infinite possibilities....
 
figurelli:
Laplacianlab, I like this approach, however can you explain more your idea?

For example, how this model will became an algorithm for real trade?

Thanks for your interest in this idea. I think I'm not going to be able to code anything about it, for now, but I can develop it a bit more cause dreaming is free and this is a brainstorming, isn't it?

Pieces

Let's assume that we've been able to identify the long term news which move the market (China's growth, gurus' predictions, debt crisis, etc.) and we are able to make questions such as the following:

  1. What do people think about Gold in 2014?
  2. When will China stop growing?
  3. ...

As we said, we finally were able to code a RDF Ontology named Collective Intuition, so right now our MQL5 EAs can perform queries like the above thanks to SPARQL. The knowledge needed for building this Ontology has been extracted from a number of different sources.

As a result, now we can combine that knowledge with a news calendar in order to place orders in markets. This is a fundamental robot based on the why of things.


EA's Cognitive Habilities

Carlsen, Polgar and Karpov are simply different people. I'm sure they won't play the same chess game against Deep Blue.

Maybe they can be modeled this way (I don't know):

Carlsen

  1. Patience = 35%
  2. Initiative = 80%
  3. OpportunityCost = 55%
  4. ...

Karpov

  1. Patience = 65%
  2. Initiative = 70%
  3. OpportunityCost = 85%
  4. ...

So why don't we try to quantify our EA's cognitive skills like this in order for it to act differently in a number of different situations? That can be done very easily with a set of parameters!

Then later we can start recording the EA's operations. If we see that results are not very good, then we can change the initial strategy.

 
Then later we can start recording the EA's operations. If we see that results are not very good, then we can change the initial strategy.
However, how could an Expert Advisor like that auto-regulate its strategic vision? Maybe this is a paradox, or not, but don't worry... there's the term self-reference to explain this. Click here to know a bit more about self-reference.
 
Ubzen:

I totally agree with this guy. Interesting thread tho. I didn't say anything because I didn't want to be mood_killer but those were my taught exactly.

Hi Ubzen, thanks for sharing, I think you have too much to contribute here and just started to do that.

Actually, Victor Allis estimated game-tree complexity of chess "to be at least 10123, based on an average branching factor of 35 and an average game length of 80. As a comparison, the number of atoms in the observable universe, to which it is often compared, is estimated to be between 4×1079 and 1081". Someone can state that the number of atoms in the observable universe is finite. Indeed, but probably we all agree that it is an impressive number.

So, the easy task here is associate tactics from chess to trading, since we can code it as concepts. And the hard one (that I call a dream) is create a model and system to do this 100% automatically.

I liked too much what you told about you chess/candle insights, as this may be also a way to address the dream, as Jordi (laplacianlab) ideas too. By the way, I could see a week ago a picture connecting all the points to make this dream a reality.

But I believe that the solution and architecture I started to show is just one way, and the main idea here is explore several concepts. In this sense, I decided not contaminate other ideas and critics, passing my insights step by step, that is what I'm doing right now.

 
laplacianlab:

So why don't we try to quantify our EA's cognitive skills like this in order for it to act differently in a number of different situations? That can be done very easily with a set of parameters! 

Then later we can start recording the EA's operations. If we see that results are not very good, then we can change the initial strategy.

Thanks, now I see it better, but I still can't connect the points.

Maybe to help this, think in the the following steps:

  • Dream Step 1: Imagine a match YOU x Market (just EUR/USD, for instance, any timeframe)
  • Dream Step 2: How do you decide what piece move (and where) using EUR/USD graph/news/etc.?
  • Dream Step 3: How EUR/USD graph/news/etc. will indicate the virtual market move (what piece and where)?

If you can write a code for this 3 steps, and explain the algorithms to address that, you have an Eureka too, as in my opinion, any dream model must address these 3 steps, if we really want to emulate this game and not just use conceptual tactics models.

By the way, I wrote this dream steps as a rule in the first post, so we can improve it better.