What does it take to succeed as a ATS trader

In my opinion, to develop any sort of 'quantitative' model, the tools required increase in complexity exponentially along with the strategy.

Creating a simple screen for stocks that moved above their 200DMA but below their 50DMA? Excel will do fine.

Trying to write custom back testing software? You sure as hell better have some programming experience. Hell, to increase the speed of my back testing software, I utilize Amazon's EC2 system to distribute tasks. There isn't a formula for that in Excel.

Let me paraphrase Bookstaber in his most recent book, The Demon of Our Own Design, when he says that you would never expect someone with any sort of financial experience to apply to a university for a teaching job in the sciences based solely on their 'mental capabilities' proven by their experience. The same goes the other way around -- how does a physicist or mathematician, with absolutely no economics experience, have any leg to stand on when developing models? Sure, the complex 'math' helps, but without a fundamental understanding of the market, it is futile.

A real mix is needed. Software Engineering skills are required for building strong, reliable software. Computer Science skills are required for developing fast algorithms. Economics/Finance is required for actually understanding why the edge makes sense.

Sometimes you might get lucky, where your edge exploits a hole in the concept of the market, and not economics -- in which case, the market can be treated like any other sort of 'unknown' for mathematicians -- but those sorts of exploits are rarer than rare.

In my extremely humble opinion, to be a well rounded quant developer, you need all the skills. On the other hand, being a jack of all trades makes you a master of none. If you can afford it, having a team works best -- those who can conceptually design the models and those who can implement them best.

-C
 
Quote from Corey:

In my opinion, to develop any sort of 'quantitative' model, the tools required increase in complexity exponentially along with the strategy.

Creating a simple screen for stocks that moved above their 200DMA but below their 50DMA? Excel will do fine.

Trying to write custom back testing software? You sure as hell better have some programming experience. Hell, to increase the speed of my back testing software, I utilize Amazon's EC2 system to distribute tasks. There isn't a formula for that in Excel.

Let me paraphrase Bookstaber in his most recent book, The Demon of Our Own Design, when he says that you would never expect someone with any sort of financial experience to apply to a university for a teaching job in the sciences based solely on their 'mental capabilities' proven by their experience. The same goes the other way around -- how does a physicist or mathematician, with absolutely no economics experience, have any leg to stand on when developing models? Sure, the complex 'math' helps, but without a fundamental understanding of the market, it is futile.

A real mix is needed. Software Engineering skills are required for building strong, reliable software. Computer Science skills are required for developing fast algorithms. Economics/Finance is required for actually understanding why the edge makes sense.

Sometimes you might get lucky, where your edge exploits a hole in the concept of the market, and not economics -- in which case, the market can be treated like any other sort of 'unknown' for mathematicians -- but those sorts of exploits are rarer than rare.

In my extremely humble opinion, to be a well rounded quant developer, you need all the skills. On the other hand, being a jack of all trades makes you a master of none. If you can afford it, having a team works best -- those who can conceptually design the models and those who can implement them best.

-C
More...

System traders end up becoming "Jack of all trades" from the outside. Basically, we acquire skills that require us to trade models systematically.

The focus of our profession is very simple. It's to develop of trading model that is profitable. The objective is simple and very similar to discretionary traders. Though, the biggest challenge is the environment...

Discretionary trading is very much a rational process. In a very simple term, a discretionary traders need to do what's "probable at the moment".

Systematic trading is not rational but a logical process. Also, in a very simple term, a systematic traders need to do what's "probable in the long run".

It's hard to imagine for someone on one side of the trade (rational / logical) but the two are approached differently. Psychologically (perceptive), both are the same. But perspectively, they are different.

One good example, would be the degree of competency towards your trading platform. For a discretionary trader, the technical competency required would be to understand their trading package they trade under. (It's even mentioned in Douglass to get proficient with the trading platform and it's very much true) For a systematic trader, the technical competency goes as far as the ability to send orders directly with the exchange (or clearing firm).

From my personal experience, the level of entry for trading has been getting higher...

Back in the old days, if you wanted to trade for a living, you would work under someone with a seat membership. Time passes and people start thinking of trading multiple markets and they go electronic... More time passes and people expect to trade more strategies and market.

Time passes and industry (markets) change with it. Livermore... Market Wizard... or "Waxie"... has been trading based on their time and their strategy of the time.

Just a reminder is... you can only make money trading from what the markets offer you (and what you percieve out of it). You can't force a strategy/style and expect it to make money because you've been working so damn hard (typical approach is to curve-fit). Hard work is a necessity... the difference is trading smart.

If you're trading in a way, where the trader next to you can do.... you'll have a higher vulnerability towards the way you trade. Doing something the guy next to you cannot do, is what "we" define as an edge. If retail traders have the potential of exposing a certain edge, most likely retail guys would expose it, considering the population.

Not to be defending myself... but that's why I continue to trade in an non-retail environment. I'm not as smart as some guys out there... especially considering myself knowing about autistic ("Rainman"-like) traders out there like D.R. (former-Schonfeld) or M. (Photon) in Chicago. Especially, when they're the reason I quit discretionary trading....
 
I personally just found nothing to disagree with in your post, Mr. Galt. Though, since my experience in the field is very limited -- and only to automated investment systems -- I would love to hear about your experiences and what you are currently trading now. I figure the only thing free in this world is advice (which is rarely free anymore...), so I might as well try to get it from those who are successful whenever I can.

Personally, I found that the best thing I could do was to find subsets of stocks that all have a specific factor in common and run back testing and allow the computer to identify the common trends. I find it brings very little personal bias to the situation, and I can generally develop a fundamental thesis as to why the strategy should outperform after the computer has delivered it. Then I genetically engineer a weighting scheme, back test, forward test, and set 'er loose. Since these are simply investment vehicles, testing forward is very easy -- generally it just requires a spreadsheet and some sort of EOD datafeed. I don't think I will ever achieve the results of an automatic trading system, but I can beat buy and hold and I don't have the same draw-downs as most trading systems. As well, the 'edge' that I take advantage of generally is human emotion -- which never disappears from the market -- so besides regenerating the weights every couple of years, there isn't much risk of the system not working -- especially where each subset is designed to work in a specific market.

But, as I said, that is my very limited experience. I am still in college :D
 
1. I've written my past experiences in my previous post. I'm a bit lazy to repost them.

2. I have access to most electronic markets globally.

3. My suggestion is to write a list of things you would like to do, or a list of things that you believe may help you trade better and a simple reason why. Like:

- T1 Line. To decrease latency.

- Cleaner data. To increase test accuracy and order execution quality.

- Faster backtesting engine. To decrease test time, and increase the ability to test complicated models.

- Genetic Algo. To increase optimization process speed.

- Neural Net. To self-adjust the parameters on a real-time basis.

- FIX execution. For flexible order execution and management.

- A real-time scanner. For automating the list of stocks, that follows my criteria.

etc...
etc...
etc...

Then give a hard look at your current situation and you'll figure out what you're missing. And you just do what it takes to fill the gaps between what you would like to do and what you can do.

Seriously, what else can you expect? You can't expect people to provide you with an easy way out because there is none.

---------------------------------

All you have to do is act professional and don't compromise. The worst thing you can be doing is feel guilty/regret for not doing something you should have done, compromising, and lying to yourself making excuses.

Things may take time but it's not hard acquiring skills. If you can't do something, go back a step and learn the prerequisite stuff. If Calculus is too hard for you... start with Algebra 2, if not Alg. 1... or +-x/
 
Quote from TSGannGalt:

...What's required for a retail ATS trader to trade against them?
More...
1) Mathematical Finance. If you understand how assets are priced, both theoretically and psychologically, you are 90% of the way there.
2) Access to capital, probably at least $1M in buying power.
3) A first rate, fast enough datafeed. GIGO.
4) A first rate connection that is reliable and fast enough.
5) Some minimal testbed to at least test the ideas on a non live account, historical or otherwise. This is more complicated than it sounds. Things like TradeStation are not what I am talking about. Alternatively, a genuine insight that needs no testing. This is usually some form of arbitrage or market making, so probably out of the real of the retail trader.
6) Minimal programming experience to get started (excel is fine), but eventually you need at least these:
a) A strong position manager that understands realtime risk and margin across many assets.
b) A good order manager that works with complicated realtime systems.
c) Eventually your system will likely benefit from being programmed in a real programming language.

I have never read as many academic papers/books on theoretical finance as I am currently, and I am sure that I am 50% better at comming up with viable systems than I ever have been.

nitro
 
Quote from nitro:

1) Mathematical Finance. If you understand how assets are priced, both theoretically and psychologically, you are 90% of the way there.
2) Access to capital, probably at least $1M in buying power.
3) A first rate, fast enough datafeed. GIGO.
4) A first rate connection that is reliable and fast enough.
5) Some minimal testbed to at least test the ideas on a non live account, historical or otherwise. This is more complicated than it sounds. Things like TradeStation are not what I am talking about. Alternatively, a genuine insight that needs no testing. This is usually some form of arbitrage or market making, so probably out of the real of the retail trader.
6) Minimal programming experience to get started (excel is fine), but eventually you need at least these:
a) A strong position manager that understands realtime risk and margin across many assets.
b) A good order manager that works with complicated realtime systems.
c) Eventually your system will likely benefit from being programmed in a real programming language.

I have never read as many academic papers/books on theoretical finance as I am currently, and I am sure that I am 50% better at comming up with viable systems than I ever have been.

nitro
More...

I agree on most of what's written. Well...

1 - I sort of agree. Obviously, I'm not as competent in this field compared to you. Though... talking from a purely model development point-of-view, the MF theoretical models require "significant" amount of tweaks for it to be applicable in real market conditions. For example, I tend to see that it becomes a non-ending pursuit for the fastest and biggest computer.

Anyways, I wouldn't put my whole weight for MF in terms of model development. Of course, a whole lot more for Risk Management purposes.

2 - Definitely.

3 - Yes.

4 - Yes.

5 - Sort of. I have a good idea where you're coming from but I tend to be skeptical on "pure genuine insight" these days. At the least, the "genuine insight" needs to be proven genuine... Especially, for the sake, of developing models around the "genuine insight".

In the same context of #1, my philosophy is "Anything goes, as long as I TEST EVERYTHING". And it's tricky trying to duplicate them and I'm not able to do all of it... as of yet.

6 - Sort of. I wouldn't use Excel in anyway, for full ATS models.

- My current focus is on the IT side. I guess I'm using up all the time I used for making sushi for learning IT stuff. And of course, I'm 1000+% better at coming up with viable systems than I ever have been. (Compared to my days in Chicago :p )

What's new, nitro?
 
You must log in or register to reply here.
Back
Top