Automated Trading - FAQs

[nonlinear5]

Programming is a craft that takes time to master. The universal rule of "10,000 hours of practice" probably applies. It's true that you can pick a "Learn language ABC in 21 days" and be able to write some "Hello, world" programs after a month. But an automated trading system (if written from scratch) is typically a complex software conglomerate which requires certain amount of skill to design, build, and maintain.

If you are looking to automate some fairly simple strategies, your best option is the trading platform with a scripting language. When you use such a platform, most of the automation tasks are delegated to the functionality that has already been built into that platform. Becoming proficient enough with the scripting language can be probably bound to reasonable amount of time.

If you are looking to build a trading system from scratch, and you don't have any programming knowledge and experience, it would probably take you 10 years of dedicated study and work. If you hire professional developers to do it for you, it would cost you a lot of money.

I'd say, go for the "low hanging fruit", which is a narrow-scope, highly specialized scripting language designed specifically for building trading strategies, and supported by a trading platform (NinjaTrader, TradeStation, etc.)

 

[Alpha Trader]

@[USER]eusdaiki[/USER] this is exactly what my buddy has done (as mentioned previously).

I completely forgot to elaborate on the point I made above. The cut his programmer/partner gets (and the partnership is not 50:50) is a good 6 figure/month.

My buddy is a pretty giving and generous person to whom money does not hold much value. He see's it as an occurrence of him being able to impact his partners life (and everyone connected to him) in a positive way. However, you have to admit that this is a unique situation.

 

[IAS_LLC]

I would say that trading can be learned in much the same way coding can be learned. The difference is lack of discipline in trading results in financial ruin, where as lack of discipline in coding results in segmentation faults .... and financial ruin if you are using your code to trade.

 

[Alpha Trader]

Nicely put. I agree with the last part, but not the first. In regards to trading being learnt, yes it can. The problem is most can't implement what they learn and a lot of it is to do with the psychology of it rather than intelligence. Hence, the success rate being so low.

In coding, you don't put hard money on the line and the consequences of your actions do not show up immediately. These are only a few things that make trading "difficult".

 

[Spearhead]

Programming is a craft that takes time to master. The universal rule of "10,000 hours of practice" probably applies. It's true that you can pick a "Learn language ABC in 21 days" and be able to write some "Hello, world" programs after a month. But an automated trading system (if written from scratch) is typically a complex software conglomerate which requires certain amount of skill to design, build, and maintain.
If you are looking to build a trading system from scratch, and you don't have any programming knowledge and experience, it would probably take you 10 years of dedicated study and work. If you hire professional developers to do it for you, it would cost you a lot of money.

I think there's a big difference between writing commercial quality code at a real software development shop as opposed to writing something for your own personal use which no one else will ever need to look at it. Assuming you're a discretionary trader with a profitable strategy, I'd think you can code up your strategy for yourself with much less than the 10000 hours of practice/10 years of study - which is probably the amount of practice time it would take to command a $200K salary at Google. I may be overgeneralizing as this is probably very strategy-dependent, just emphasizing that a large amount of effort in a formal software dev setting is spend programming to standards, writing tests, writing legible code, etc.

 

[IAS_LLC]

In coding, you don't put hard money on the line and the consequences of your actions do not show up immediately. These are only a few things that make trading "difficult".

That assumes you are coding for the sake of coding, which unless you are an extremely bored individual, you are not. You are coding to accomplish a task, and it is the task that is inherently difficult. Clicking the mouse is to chart trading, as coding is to automated trading.

To me the biggest advantage of automated trading (other than speed), is removing judgement from the equation. My code will never violate my risk managment rules.... but my hand on the mouse might... So, the coding takes some of the psychology out of the equation.

 

[Spearhead]

Nicely put. I agree with the last part, but not the first. In regards to trading being learnt, yes it can. The problem is most can't implement what they learn and a lot of it is to do with the psychology of it rather than intelligence. Hence, the success rate being so low.

In coding, you don't put hard money on the line and the consequences of your actions do not show up immediately. These are only a few things that make trading "difficult".

My gut is telling me that saying "trading can be learnt" is trivializing it a bit especially when you're equating it with learning how to code. I don't quite think it's learn-able in the same way as learning to code, which probably just comes down to time/effort/practice. In trading, there's no motivation for the profitable to share knowledge, and every motivation to keep things secret, so getting an education is pretty tough.

I would think that most successful algo traders were profitable discretionary traders, or learned from their previous employers, before they got into the algo game. I would "think" that there aren't many successful algo traders that started out purely as a software guy who decided to get into trading.

So, with a profitable strategy in hand, there's a really high bar before I'd be willing to consider outsourcing or taking on a partner who I'd have to share strategies with.

 

[Alpha Trader]

You are coding to accomplish a task, and it is the task that is inherently difficult. Clicking the mouse is to chart trading, as coding is to automated trading.

Beautifully put!

To me the biggest advantage of automated trading (other than speed), is removing judgement from the equation.

I am a heavily price action based trader and I thrive on dissecting discrete anomalies. These anomalies are numerous in number and coding for each would not make sense. Also, it would be extremely difficult to write a code for them as most times these irregularities are taken advantage off based on context. Basically, the judgement factor is important to my strategy.

The coding I would want is for simple, straight forward strategies that have a high probability. So whilst I am trading my manual strategies, it would be nice to have an algo make additional executions. My advantage for coding would be diversification and the quantitative factor.

 

[IAS_LLC]

If they are simple, it wouldn't be worth it to hire a programmer anyway, just us NinjaTrader or TradeStation. If you don't mind me asking, how do you know these "simple" strategies are high probability? What frequency are you looking at?

Beautifully put!



I am a heavily price action based trader and I thrive on dissecting discrete anomalies. These anomalies are numerous in number and coding for each would not make sense. Also, it would be extremely difficult to write a code for them as most times these irregularities are taken advantage off based on context. Basically, the judgement factor is important to my strategy.

The coding I would want is for simple, straight forward strategies that have a high probability. So whilst I am trading my manual strategies, it would be nice to have an algo make additional executions. My advantage for coding would be diversification and the quantitative factor.

 

[rwk]

. . . Do you think a 10:90 split would attract someone, if we ASSUME that the strategies can bring in $500k/ year? On this account, the programmer would get a split of $50K/year. . .

On first pass, 10% sounds reasonable, as money managers often get something like 2/20. But if we're talking about a startup without a solid track record and smallish capital base, there would be a lot of risk for the programmer.

Some years ago a buddy proposed that I automate his trading. I don't recall the split, but I think it was something like 20 or 25%. I agreed, but then there was a delay in getting his funding. During the delay, I reconsidered and backed out. I decided I didn't like the risk/reward profile, and moreso I didn't like committing to a project without seeing what I was getting into. I don't cope well with tedium anymore. Now that I program for myself, I like it better, and I have no interest in programming for pay.