Cost of Programmer to automate

[tradelink]

I deliver automated strategies at flat-fee pricing. No hourly rates. Start as low as $300.

I generally build strategies in c#, but other languages are available.

Generally I develop for Assent traders, but so long as your broker isn't from the stone-age it won't be an issue.

trade@franta.com
786-302-4115

 

[Jerry030]

Quote from tortoise:

Where does one find a "developer"?

First you need a specification for what you want of some kind, in general terms. Anyone who will quote you a rate without knowing what the project involves is ...well, someone you don't want to hire. The reason is since they are getting paid by the hour they have no need to worry if they can do it within your budget (assuming you have one) so they won’t bother asking about the details before they try to underbid each other.

On the other hand if you can't define what you want in general terms, most competent programmers and developers won't want to work with you as it's usually a mess.....

I'd put a listing on dice.com, where you will have people offer competitive bids on your project. Remember in life you tend get what you pay for.

Jerry030

 

[fatrat]

I find this thread horribly depressing.

I used to do small projects and stuff for free to keep my skills sharp, but now I am regulated and no longer do anything outside of work -- aside from pet projects. They pay me nicely to compensate for my inability to trade anything except on paper or in long-holding periods/intervals.

The only people truly qualified to assess the quality of a programmer are other programmers. Even then, programmers have disputes about languages, methodologies, etc. Just ask anyone who has ever interviewed for a job, especially regarding what they thought was important vs. what the interviewer thought was important.

That having been said, the OP should just sack a few k on simple, idiotic systems on a few different programmers from foreign lands. Figure out which guy delivers reliability on some partially-profitable or non-profitable model and put the ATS on a paper account. Then go back to the programmer who gave you the best work and ask for more work.

For example, there's no statistical edge in using technical analysis indicators. Make up some dumb system buying oversold and selling overbought on RSI or an equivalently dumb project. So the best way to find a programmer among the set of programmers is to hire 2-3 low wage foreigners, then pit their work against each other and see where the features match up, and where there is consistency and accuracy. When you finally settle on someone's work, then move on to push a more advanced project.

There's some magic in this advice, because if you get 3 products from 3rd worlders vs 1 product from an American programmer, you have an idea of what can go right and what can go wrong from the various attempts at development. Eventually, you will find a groove and characteristics of programmers that you like -- just based on your experience and your interactions with the guy you work with.

American programmers will cost you a lot, because the economics of living for us is completely different than that of someone living in India. You are not going to get top Wall St. talent, however -- the people who can code high quality financial stuff do code high quality financial stuff and they whore themselves out to Wall St.

 

[Jerry030]

Quote from fatrat:

You are not going to get top Wall St. talent, however -- the people who can code high quality financial stuff do code high quality financial stuff and they whore themselves out to Wall St.

Yes, those very skilled people may be working for Wall Street, but the really brilliant ones are working and trading for themselves. These are the ones who in addition to implementing great ideas and concepts from someone else, can also generate their own insights and discoveries AND implement them. These are the people with unlimited market potential.

The Wall Street crowd has certain limitations. If you have ever work in a large corporation, you know that with about 4 managers, 7 VPs and several departments involved in major undertakings little that is new, or radically better will ever be implemented. They mostly reinvent the wheel that they heard about at the last development confab.

In fact my own theory as to why the performance of the major funds and private capital managers is so anemic is not lack of hardware, ideas or budget but the fact that software doesn’t obey the laws of Newtonian physics...it's more of a Quantum kind of thing. To explain: if I'm a corporation and I hire an engineer to invent a great new mouse trap which will make a ton of money, I can control the outcome. The mouse trap can be patented. If the engineer decides that he wants to take the idea I paid him to develop and go into business for himself, I can sue him for patent infringement and employment contract violation. Any use of the mouse trap is visible in the marketplace.

Trading software implementation has none of these characteristics. In theory nobody knows if what a programmer is trading is something he created or is something the hedge fund he used to work for paid him $100,000 to program and he retained a copy.

Beyond that I can see where it world be impossible for Wall Street to get a bunch of programmers to create a hugely profitable applications. Imagine this...boss says "OK guys if this system trades like we hope it will, the CEO says all you guys will get a $10,000 bonus and an all expenses paid vacation in Orlando, so get to work so we can make a lot of money".

The programmers code away and when they start testing they realize that in fact the system will make outstanding profits.

Now do they either individually or collectively do which of the following?

1) Call the boss and tell him it works, it's done and ready for trading
and get sun tan lotion of that company paid vacation

2) Figure they should do a little “out of office” testing with their own account while telling everyone that the system is still being debugged.

3) After finding that it does make a ton of money, introduce some really hard to find bugs in the official copy, transfer the "good" system into their private property, open an offshore account, resign in a couple of weeks and vanish into the haze to trade from the beach in Antigua.

I suspect the really great trading ideas are discovered and implemented by individual trader/programmer/developers and never get anywhere near a big Wall Street firm.

Jerry030

 

[RedRat]

3) After finding that it does make a ton of money, introduce some really hard to find bugs in the official copy, transfer the "good" system into their private property, open an offshore account, resign in a couple of weeks and vanish into the haze to trade from the beach in Antigua.

I suspect the really great trading ideas are discovered and implemented by individual trader/programmer/developers and never get anywhere near a big Wall Street firm.

Programmers do not introduce/discover ideas. They implement code according to Business Analytic specifications. And they can not "hide profitability".

Also starting such a profitable system may require huge costs individual can not provide. Say market arbitrage system, you need high performance servers, direct lines to the brokerage (estimate $5k/month), low commissions (be the member of the exchange) so on.


Let me compare individual with large Investment Bank. He can work hard and discover great idea. He can implement it and find out it is profitable. He is quickly at changing software and accomodating it to current market skipping so many testing stages. He can trade huge leverage, estimating drawdown to be 70% while IB can only suffer 5% drawdown - so he is expecting huge annual %, say 300%.

But individual has his own disadvantages. This lack of testing can
lead to bugs in real version. I myself suffered many times because of bugs in my programs or in TWS API. As individual has small account he can not perform a good diversification and his drawdown may be larger then expected => blowout the whole account.

Individual has his own qualities like lazyness, lack of discipline, lack of testing...

So in long time IB may be better ;-).
RR

 

[maxpi]

I could use some expertise. I can't imagine having somebody else develope a system and code it though. I'm coding things in OpenQuant and basically I have to babysit the thing, watch it all day, debug it in realtime, debug it in backtesting, etc... how I could even direct somebody else in all this I don't know. I'm struggling with C# a lot though, is there a website or a good source to hook me up with somebody that could answer my coding problems a little piece at a time? I used to use Experts-Exchange but it seems to have degenerated. I can't get specific enough answers out of the OpenQuant forum, they direct everything at C# programmers with a high level of competency, I'll never be that, I can program but I don't have the C# "genes". I have code up and running but there are things I would like to do that so far I can't, not at my level of coding capability..... and I don't want to share the strat, I just want to ask some specific questions and get some specific answers, you would think some programmers would like that business model to earn a little spare cash....

 

[RedRat]

I could use some expertise. I can't imagine having somebody else develope a system and code it though. I'm coding things in OpenQuant and basically I have to babysit the thing, watch it all day, debug it in realtime, debug it in backtesting, etc... how I could even direct somebody else in all this I don't know. I'm struggling with C# a lot though, is there a website or a good source to hook me up with somebody that could answer my coding problems a little piece at a time? I used to use Experts-Exchange but it seems to have degenerated. I can't get specific enough answers out of the OpenQuant forum, they direct everything at C# programmers with a high level of competency, I'll never be that, I can program but I don't have the C# "genes". I have code up and running but there are things I would like to do that so far I can't, not at my level of coding capability..... and I don't want to share the strat, I just want to ask some specific questions and get some specific answers, you would think some programmers would like that business model to earn a little spare cash....

There are some programming boards like
www.codeproject.com
www.codeguru.com

you can ask your questions there. But your questions may be OpenQuant specific so you will get no answer.


as for

and I don't want to share the strat, I just want to ask some specific questions and get some specific answers

One of my projects was to implement the following system. Each day I read a .csv file with list of limit orders, that file is generated by trader every day using his private algorithms. When market is near the desired limit - I place it to the market. Plus additional functionality. There is NO sharing of the strategy.

On the other side these high-level programs have restrictions. For example old version of OpenQuant supported only one strategy, don't know how good is 2.0 version. I also do not know how good is it in trading several different markets.

I have experience in developing strategy for myself, trading ER2. And it was really hard, I spent about 2 years. And it is profitable . But I CAN NOT develop such a strategy using ANY of high-level programs. It is completely custom.

 

[johnpinochet]

Quote from Jerry030:

Depends on the language you want it coded in. Something in general use like VB is less expensive. A good place to look for programmers is dice.com

If your trading platform has the ability to execute encapsulated code and the system you want to implement can be defined by modeling then there is a much faster and cheaper alternative. Many of the general purpose data mining applications have the ability to take historical examples along with your desired goal and turn them into a predictive model. The results can be exported to VB, C++, Java, XML, etc to be used by another application (your trading platform).

Since the software does the heavy lifting, the human effort is limited to defining the inputs and the goal you want to predict, so the human hours are a fraction of the cost of having a programmer code a complex rule based system. This is the approach I went to after spending dozens or hundreds of hours programming ideas only to find they didn't work as well as I thought or they stopped working after the market evolved its patterns over time. With a model you can remodel the market as it evolves or even weekly as opposed to having a static programmed system.

The actual run time to generate a model for most commercial grade predictive applications is less than 24 hours. This may compare with weeks and weeks of programming and testing time for a complex trading system.

Jerry,

Do you have some links to your favorite commercial predictive applications?

I've probably spent multiples of 10,000 USD over the years on all types of software, but the absolute best one I use (aside from the stuff I made myself) is WEKA and it is free.

http://www.cs.waikato.ac.nz/ml/weka/

It is a piece of cake to use if you like Java. Java and Python are my languages of choice. I slice and dice VLF's (minimum of 10 to 50 Gig files) with Python (my code) and follow that up with various Java wrappers (my code) to WEKA.

 

[tradelink]

weka is very cool. only downside is it's java (which I personally am not a huge fan of).


R is less graphical than weka, but just as free and has bigger community and more plugin support. it's very easy to use, will be familiar to mathematica/matlab users.

and then there's always excel, bc nothing beats a pivottable.

 

[Jerry030]

Quote from RedRat:

Programmers do not introduce/discover ideas. They implement code according to Business Analytic specifications. And they can not "hide profitability".

Also starting such a profitable system may require huge costs individual can not provide. Say market arbitrage system, you need high performance servers, direct lines to the brokerage (estimate $5k/month), low commissions (be the member of the exchange) so on.


Let me compare individual with large Investment Bank. He can work hard and discover great idea. He can implement it and find out it is profitable. He is quickly at changing software and accomodating it to current market skipping so many testing stages. He can trade huge leverage, estimating drawdown to be 70% while IB can only suffer 5% drawdown - so he is expecting huge annual %, say 300%.

But individual has his own disadvantages. This lack of testing can
lead to bugs in real version. I myself suffered many times because of bugs in my programs or in TWS API. As individual has small account he can not perform a good diversification and his drawdown may be larger then expected => blowout the whole account.

Individual has his own qualities like lazyness, lack of discipline, lack of testing...

So in long time IB may be better ;-).
RR

It sounds like you don't know any good programmers.

I used to manage about 35 of them. Mind you I'm not talking about somebody that bought themselves a couple of computer books on Amazon and calls themselves a programmer, but real professionals. They tend to be extremely clever, creative, and individualistic and are quite capable of understanding the potential of the software they develop.

But it's the same reason that say a large well financed comapny like IBM didn't invent Google or Yahoo. An individual working for for a tiny fraction of the value of their effort has a very different impact than a person wroking for 100% of the value of what they do, as in the case of the creators of Google and similar "new ideas".