Topic: What I've learned after 3 years building an algo trading bot - page 2. (Read 287 times)
I like you saying that bots make money but also losses but the amount of loss is what is determined by risk management. Yes many traders that are using bots throw caution away and they trade with high risk with believing that bots don't fail but the reality is they fail just like they profit, so what is the point? Bots are not perfect and are not easy to create. I believe in dyor and trading limited options so that you are in control of your trade decision.
Quote from: dothebeats
Your insight on neural networks and historical data is on-point. I've watched a video wherein they tried to train an AI to trade and hopefully profit off of the bot. Turns out, using historical data on highly volatile markets as a reference for the bot will only end up in shambles, as it will always try to find and match certain patterns that it has recognized on the historical data and try to correlate it with the present, which for the most part does not really makes sense. I'm amazed with how trading bot creators, especially those which are mostly profitable, create their bots in such a way that it will ignore past trends and try to make sense of what's in front of it. I'm a coder myself but I'm on the lighter side of coding and I only do it for fun. If I were given a chance to study again, perhaps it will be AIs, machine learning, neural networks, and the deeper side of comsci just to try and 'understand' how you guys do it.
IMO you shouldn't be amazed at all, most of the retail bots are pure garbage. They are just a bunch of ifs and else statements that utilize the same ol' useless Indicators and shipped as AI. The issue is these "bots" are merely "tuned" to the "most profitable" settings they have found through brute forcing through historical data. If MACD does this and RSI does this and the 50 MA crosses this and the next candle is bullish trigger buys. The issue is everyone looks at the market through a certain lens which is through indicators if you are a neophyte and through volume profiles/Delta and VWAPs on the "Professional" side of things. The biggest issue is that both of these lenses are broken in themselves.
So no matter how good of a programmer you are, you are going to have to write or train your model with what this lens tells you. The only real advantage a bot will have over a human, in this case, would be of ability to run round the clock, it will be able to execute more trades than a human but won't change the outcome in the long run, if the lens is busted. What this industry needs is a new lens / a new perspective which will be able to counter what everyone else is doing. Wyckoff referred to it as the composite man/Operator. If you ask me personally I believe it is the case, the Composite man is the FED (Central Banks). People are made to believe that they can push the price around of their respective currencies by buying and selling while the commodity they are trading isn't owned by them. The currencies are the commodities of the Central Banks, they can set the price to whatever level they want. Just like Apple sets the price of iPhones, the Central bank can set the price of their product to whatever level they like. IMO whoever can track their footprints can beat the market.
Your insight on neural networks and historical data is on-point. I've watched a video wherein they tried to train an AI to trade and hopefully profit off of the bot. Turns out, using historical data on highly volatile markets as a reference for the bot will only end up in shambles, as it will always try to find and match certain patterns that it has recognized on the historical data and try to correlate it with the present, which for the most part does not really makes sense. I'm amazed with how trading bot creators, especially those which are mostly profitable, create their bots in such a way that it will ignore past trends and try to make sense of what's in front of it. I'm a coder myself but I'm on the lighter side of coding and I only do it for fun. If I were given a chance to study again, perhaps it will be AIs, machine learning, neural networks, and the deeper side of comsci just to try and 'understand' how you guys do it.
It must have been a great ride because coding something like that would be quite hard. And as you mentioned, "state of the art code" would be a fantastic feat because you can sell it, and if there's something special to it, it can bring you even more money. With all the things you've learned, I think most people would have different views, and there would even be more diversity in beliefs, so DYOR is one of the essential things you have covered, IMO.
Writing a state of the art code is the easiest part.
After Looking at the waves of the stock market at the first time with Zero knowledge in trading, stocks or finance, said to myself "I can write a code that will buy here, sell there and make a profit on the way. I'm not a trader and no way I'm sitting in front of graphs all day clicking buy and sell buttons."
That was easy to say being a software engineer who built numerous high performing platforms filling roles from programmer to CTO most of my life.
DYOR
The vast majority of the influencer's, podcaster's, youtuber's, tweeter's siglangroups and other shamans claiming to show you how to make money, don't know shit.
To distinguish between them you will have to listen to lot of them in order to understand what topic you should learn and understand the bits and bites of. it's an essential part of learning and there are no shortcuts.
The real people who influence this market, don't talk about it on the web but you can read their action in the exchange during drastic market changes.
Bitcoin,the most volatile market.
Once you ride this horse, others are easy to handle.
Focusing on one market to understand everything from A to Z is important. Once one has the full knowledge, it's easy to implement it on different markets.
TA's are nice but it's analyzing history. They can predict nothing. All they can tell you is what happens till this moment.
Neural Networks (AI) is subjected to what you teach the machine. if you teach historical data the output is useless.
There are tons of data out there, the ability to analyse and cross reference it in real time and decide on an action is the key part of this whole automated system, an advantage only a machine can do and it's all about data progressing.
What is the right data to make decisions upon? DYOR.
Manage losses is the only thing matters.
Bots make money all the time, but they also loose. How much they loose is up to your risk management strategy.
The key is not loosing more then earning over pre defined period of time.
To set up a profitable good risk management strategy you have to have a good understanding of the entire eco system.
You can make 90% win rate and then with one open position during a drastic market change , you can loose more then you made.
There is no such thing as not loosing a position once in a while. no matter how good your analysis is.
It's all about balancing the portfolio and not about a single position.
It never stops
-The Good.
Bot is acts like 'The Terminator'. Doesn't feel pity, remorse, fear and never stops till it's pre programmed plan complete.
Qualities like these are hard for traders to master.
-The Bad
Even the best designed autonomous car crashes sometimes on some unpredictable scenarios.
Good risk management will keep you running after recovery.
Development and learning never ends.
programming, security, databases, cloud, VPS, statistics, signal analysis, algorithms, real time processing, finance, investment, funds management, regulations, accounting, risk management, legal are just part of the topics one has to master.
ROI
Invested three years of living and gave up three years of income as CTO.
Was a he'll of a ride, one of the best, definitely do it again.
After Looking at the waves of the stock market at the first time with Zero knowledge in trading, stocks or finance, said to myself "I can write a code that will buy here, sell there and make a profit on the way. I'm not a trader and no way I'm sitting in front of graphs all day clicking buy and sell buttons."
That was easy to say being a software engineer who built numerous high performing platforms filling roles from programmer to CTO most of my life.
DYOR
The vast majority of the influencer's, podcaster's, youtuber's, tweeter's siglangroups and other shamans claiming to show you how to make money, don't know shit.
To distinguish between them you will have to listen to lot of them in order to understand what topic you should learn and understand the bits and bites of. it's an essential part of learning and there are no shortcuts.
The real people who influence this market, don't talk about it on the web but you can read their action in the exchange during drastic market changes.
Bitcoin,the most volatile market.
Once you ride this horse, others are easy to handle.
Focusing on one market to understand everything from A to Z is important. Once one has the full knowledge, it's easy to implement it on different markets.
TA's are nice but it's analyzing history. They can predict nothing. All they can tell you is what happens till this moment.
Neural Networks (AI) is subjected to what you teach the machine. if you teach historical data the output is useless.
There are tons of data out there, the ability to analyse and cross reference it in real time and decide on an action is the key part of this whole automated system, an advantage only a machine can do and it's all about data progressing.
What is the right data to make decisions upon? DYOR.
Manage losses is the only thing matters.
Bots make money all the time, but they also loose. How much they loose is up to your risk management strategy.
The key is not loosing more then earning over pre defined period of time.
To set up a profitable good risk management strategy you have to have a good understanding of the entire eco system.
You can make 90% win rate and then with one open position during a drastic market change , you can loose more then you made.
There is no such thing as not loosing a position once in a while. no matter how good your analysis is.
It's all about balancing the portfolio and not about a single position.
It never stops
-The Good.
Bot is acts like 'The Terminator'. Doesn't feel pity, remorse, fear and never stops till it's pre programmed plan complete.
Qualities like these are hard for traders to master.
-The Bad
Even the best designed autonomous car crashes sometimes on some unpredictable scenarios.
Good risk management will keep you running after recovery.
Development and learning never ends.
programming, security, databases, cloud, VPS, statistics, signal analysis, algorithms, real time processing, finance, investment, funds management, regulations, accounting, risk management, legal are just part of the topics one has to master.
ROI
Invested three years of living and gave up three years of income as CTO.
Was a he'll of a ride, one of the best, definitely do it again.
Jump to: