My Experience In A Semi-Professional Trading Group
Introduction
My experience in the professional trading field has been brief and unsuccessful however I'd like to share it with you so you can know how a semi-professional trading group look like. I don't like to talk about it in general because it wasn't that great and was a waste of time however this can be of interest for people's looking to enter into this field.
So I'll talk about the structure of the group and what tools I used for trading.
How It Started
Sevilla is said to be the capital of Andalusia, one of most visited regions in Spain, I live near this city and I was surprised to see that a new trading group was forming and recruiting trader's. At that time I was 18 year old lacking in skills (I'am still lacking), the only thing I could do was build technical indicators. I met one of the guy who was in control of the group and he was kind enough to let me pass a testing period where I would be required to prove consistency and profitability.
The Group
The group would be funded from a private fund from USA, a first funding was allocated to us and another group, we where therefore in direct competition, of course the one with the best results would get a more important funding while the other group would be discarded from the project.
The Structure
The group although small still consisted in several departments described as follows :
Department of Risk's :
this department consisted in various quantitative analyst that would test the trading models of the traders, in short they would test the stability of the model and see if they where profitable using mathematical (stochastic) models. Because such step was time consuming traders in their testing period could skip it however this step is mandatory for any trader who passed their test.
Department of Trading :
Self explanatory, this is where traders trades, the group was specialized in high frequency trading, financial instruments used where mainly currencies.
There was also another department but I'll talk about it later.
Traders executed trades from metatrader 4 with the possibility of using other platforms, in this last case this would require testing and configuration and approval of the supervisor .
All the data generated by traders where periodically reviewed, we could access descriptive analysis of our trading periods, I can tell you I wasn't doing great.
High frequency trading expose the trader to various factors such as latency (network/devices), fiber optics where therefore used for the network. Monitors where required for fast response, for example you wouldn't use a tv screen connected to your PC as display since it would allow for more latency, it might sound ridiculous but latency can really be an important factor in trading.
Proprietary System Department
The funder of the group had a proprietary algorithmic trading system, actually I could learn a lot from this.
When we talk about auto trading we can think of us lazing around while having something trading for us, well that wasn't really the case here, the proprietary system entered trades on its own however it still required human interaction. Basically after some time a trader from the trading department would go operate with the proprietary system, his mission was basically turn off the system during high volatility periods, most of the time economic event's, I hadn't the possibility to do that.
It was possible to use the system with our own capital, this would require the user to sign a contract and provide a certain capital, it was appealing because the system showed performance of 60% accuracy (it's a crazy impressive result), profits made would be splitted.
My Experience
First of all let's describe the strategy I used, the strategy would consist in two filter crosses with :
A main filter who aimed to provide optimal noise removal while keeping high reactivity and a signal filter who aimed to provide optimal crosses with the main filter.
Those filters where based on two groups of filters, the main filter F was using :
An exponential moving average, a fractal adaptive moving average, a least squares moving average and a recursive filter S described as follows :
S = (p*price+(1-p)*S) + k*(F - F')
Where p = running hurst exponent estimated via the rescaled range method and k = square root of p
The signal filter F' was using :
A least squares moving average, a simple volatility stop and a filter S' described as follows :
S' = (p*price+(1-p)*S') + k*(F - F')
Each filter F and F' is calculated from a weighted sum that maximized the equity function of the strategy. Because I wasn't patient I tossed aside the optimization of the filters settings and used a period of 50 for every filters in the main group and a period of 200 for every filters in the signal group.
The optimization of the weights was done using gradient descent, because I was in a hurry I just used the weights found after a low number of epochs. Everything was done wrong, from picking the filters in the groups to training and testing the model.
I lasted 4 days, seeing I was doing horribly the supervisor said that it was better to stop there, and indeed it was the right thing to do.
Conclusion
I couldn't learn much about the professional world of trading, you think it's hard but it's always harder than what you first thought. However it's really interesting to see how things are done. Now my experience is not the standard experience anyone would have, things might be different for other peoples, however in any cases no matter where, risk must always be minimized.
Classical technical analysis was used by some traders, while others used more complex models that where this time well trained and tested. I don't know what happened to the group, they might have failed to be better than the other one but i wasn't there enough time to care about that.
So let's summarize : I was lazy and irrational, my model could have been interesting but the choice of the filters where random, the model was badly known, a friend helped me set it up, the training and testing was horrendous, results ? A failed test and lot of shame, but it's alright.
This is actually really important, when you deploy a strategy, being for automated trading or not you must understand it at 100%, every bit of it, do you plan to use a sma crossover strategy ? So you must know how a sma is calculated, what are it's characteristics, what is the relationship between it's period and it's frequency/phase response and many more.
So in short : know your model like if your life depended on it, get your model tested and validated by a professional or know how to backtest it "the right way", never buy other people indicators/strategies without the guarantee of a lifetime support, creating the strategies by yourself or alongside a professional who can provide support is always better, then deploy it the right way, in mt4 it was easier.
Well that's it, i'am still happy to have learned something from this experience and I hope you'll learn something from it to. Thanks for reading.
My experience in the professional trading field has been brief and unsuccessful however I'd like to share it with you so you can know how a semi-professional trading group look like. I don't like to talk about it in general because it wasn't that great and was a waste of time however this can be of interest for people's looking to enter into this field.
So I'll talk about the structure of the group and what tools I used for trading.
How It Started
Sevilla is said to be the capital of Andalusia, one of most visited regions in Spain, I live near this city and I was surprised to see that a new trading group was forming and recruiting trader's. At that time I was 18 year old lacking in skills (I'am still lacking), the only thing I could do was build technical indicators. I met one of the guy who was in control of the group and he was kind enough to let me pass a testing period where I would be required to prove consistency and profitability.
The Group
The group would be funded from a private fund from USA, a first funding was allocated to us and another group, we where therefore in direct competition, of course the one with the best results would get a more important funding while the other group would be discarded from the project.
The Structure
The group although small still consisted in several departments described as follows :
Department of Risk's :
this department consisted in various quantitative analyst that would test the trading models of the traders, in short they would test the stability of the model and see if they where profitable using mathematical (stochastic) models. Because such step was time consuming traders in their testing period could skip it however this step is mandatory for any trader who passed their test.
Department of Trading :
Self explanatory, this is where traders trades, the group was specialized in high frequency trading, financial instruments used where mainly currencies.
There was also another department but I'll talk about it later.
Traders executed trades from metatrader 4 with the possibility of using other platforms, in this last case this would require testing and configuration and approval of the supervisor .
All the data generated by traders where periodically reviewed, we could access descriptive analysis of our trading periods, I can tell you I wasn't doing great.
High frequency trading expose the trader to various factors such as latency (network/devices), fiber optics where therefore used for the network. Monitors where required for fast response, for example you wouldn't use a tv screen connected to your PC as display since it would allow for more latency, it might sound ridiculous but latency can really be an important factor in trading.
Proprietary System Department
The funder of the group had a proprietary algorithmic trading system, actually I could learn a lot from this.
When we talk about auto trading we can think of us lazing around while having something trading for us, well that wasn't really the case here, the proprietary system entered trades on its own however it still required human interaction. Basically after some time a trader from the trading department would go operate with the proprietary system, his mission was basically turn off the system during high volatility periods, most of the time economic event's, I hadn't the possibility to do that.
It was possible to use the system with our own capital, this would require the user to sign a contract and provide a certain capital, it was appealing because the system showed performance of 60% accuracy (it's a crazy impressive result), profits made would be splitted.
My Experience
First of all let's describe the strategy I used, the strategy would consist in two filter crosses with :
A main filter who aimed to provide optimal noise removal while keeping high reactivity and a signal filter who aimed to provide optimal crosses with the main filter.
Those filters where based on two groups of filters, the main filter F was using :
An exponential moving average, a fractal adaptive moving average, a least squares moving average and a recursive filter S described as follows :
S = (p*price+(1-p)*S) + k*(F - F')
Where p = running hurst exponent estimated via the rescaled range method and k = square root of p
The signal filter F' was using :
A least squares moving average, a simple volatility stop and a filter S' described as follows :
S' = (p*price+(1-p)*S') + k*(F - F')
Each filter F and F' is calculated from a weighted sum that maximized the equity function of the strategy. Because I wasn't patient I tossed aside the optimization of the filters settings and used a period of 50 for every filters in the main group and a period of 200 for every filters in the signal group.
The optimization of the weights was done using gradient descent, because I was in a hurry I just used the weights found after a low number of epochs. Everything was done wrong, from picking the filters in the groups to training and testing the model.
I lasted 4 days, seeing I was doing horribly the supervisor said that it was better to stop there, and indeed it was the right thing to do.
Conclusion
I couldn't learn much about the professional world of trading, you think it's hard but it's always harder than what you first thought. However it's really interesting to see how things are done. Now my experience is not the standard experience anyone would have, things might be different for other peoples, however in any cases no matter where, risk must always be minimized.
Classical technical analysis was used by some traders, while others used more complex models that where this time well trained and tested. I don't know what happened to the group, they might have failed to be better than the other one but i wasn't there enough time to care about that.
So let's summarize : I was lazy and irrational, my model could have been interesting but the choice of the filters where random, the model was badly known, a friend helped me set it up, the training and testing was horrendous, results ? A failed test and lot of shame, but it's alright.
This is actually really important, when you deploy a strategy, being for automated trading or not you must understand it at 100%, every bit of it, do you plan to use a sma crossover strategy ? So you must know how a sma is calculated, what are it's characteristics, what is the relationship between it's period and it's frequency/phase response and many more.
So in short : know your model like if your life depended on it, get your model tested and validated by a professional or know how to backtest it "the right way", never buy other people indicators/strategies without the guarantee of a lifetime support, creating the strategies by yourself or alongside a professional who can provide support is always better, then deploy it the right way, in mt4 it was easier.
Well that's it, i'am still happy to have learned something from this experience and I hope you'll learn something from it to. Thanks for reading.
Check out the indicators we are making at luxalgo: www.tradingview.com/u/LuxAlgo/