System Presentation

Quote from TSGannGalt:

Let's see...

1. Tradestation... That won't work, no one will take a Tradestation code or model.

2. You do the work, meaning you will have to code and create the execution engine yourself. None of the firms will take your system and implement it for you. Also, larger the firm, the more custom their execution platform is. You should be expecting to code you automation engine via FIX or at least know how to send messages via sockets. Do you expect SAC or Citadel trading with IB or some retail broker????

3. Are you willing to provide them with the source code? If you are not, why not? They are more interested in your ability as a developer. Any newbie can make one system with luck. One good system is worth nothing. One good developer is what they want.

Every once in a while, they'll call you in to develop a model with specific profiles of the market. (You don't have to, or be able to because you can only make money off of what the market provides you, but you will need to clearly tell them why you can't develop a model within their need)

4. The fund you deal with... You have to consider that the funds already have their own trading style. How is it going to appreciate their performance? Getting their past fund record and running a test showing how your system will improve their performance is equally important.

5. Most of the reputable firms know more about trading systems than you. (Considering you are using TS, not a custom tester). One simple question they will ask is what you will do when the system fails. How are you going manage it?

6. No grey box. But expect to be monitoring them full-time. If you tell them that you manually enter orders... you'll be out the door with a cute smeer. We all know that a perfect systems does not exist... so we emphasize on how you monitor the system as much as working hard to develop a model. The question you will get will be something like, "How are you going to manage the system?"

7. Obvious question you'll get is "Why aren't you trading it on your own, if you think it's great?" For me, I was trading most of the systems from home directly with a clearing firm. Though, I needed better environments to trade a few models I developed, like low latency and computer power (basically a remote server close the exchange... it was high frequency arb model). So I approached a few firms to allow me to use their execution platform.

There's more... But I think it's good for now... nothing wrong with starting a freelance developer career early.
More...

if creating the execution code is an obligation for trading our system with a larger fund - then that is not for me...I would rather start up my own hedge fund instead...and manually open the trades..:)

monitoring system full-time is a good idea..even if discretion is take out of the trading the pc and internet can fail..

I am developing my system in a general purpose model building environment - not tradestation - which has proven to be rather weak in handling complex trading ideas....not saying they cannot be coded is TS, though...

this system is the culmination of a few years of research - not something which fell on my last week in a daydream...why should I open this to them?

Semantic system development is 99,99% of all system development...coding is relatively much more easy (idea development can be compared to writing a novel while coding to grammar and syntax correction)....they can ask any programmer to code their ideas.....
 
Quote from dima777:

if creating the execution code is an obligation for trading our system with a larger fund - then that is not for me...I would rather start up my own hedge fund instead...and manually open the trades..:)

monitoring system full-time is a good idea..even if discretion is take out of the trading the pc and internet can fail..

I am developing my system in a general purpose model building environment - not tradestation - which has proven to be rather weak in handling complex trading ideas....not saying they cannot be coded is TS, though...

this system is the culmination of a few years of research - not something which fell on my last week in a daydream...why should I open this to them?

Semantic system development is 99,99% of all system development...coding is relatively much more easy (idea development can be compared to writing a novel while coding to grammar and syntax correction)....they can ask any programmer to code their ideas.....
More...

What I mention is not about good or bad ideas. It's what you will be dealing with.

Another thing is, you shouldn't expect the firms to use their resources. They've already done you a favor if they decide on going forward, and asking them to use their resources to implement a system will not give them a good impression.

Implementing a system is not a simple process for us. Coding is the easy part. There's system evaluation (analyzing what type of equipments are required and the cost), multiple performance tests(running it on UAT, then gradually increasing size) and other process (like comparison analysis between hypothetical and real money, defining a contingency plan for a shock event... cut the system, reduce size... etc. etc.) before we take to go fully live for a fund.

This is coming from a guy who's done what you're looking to do for over 6 years, in Chicago. Only thing you should be asking from them is the capital to implement it. If you start expecting anything other than capital... you'll start giving them an impression as a needy newbie.

It's like:

"I need money to trade the model.
I need a programmer to implement this.
I need a fast computer to run the system.
I need...
I need...
I need...

You guys do all the middle work and I won't give out my system. But I won't take any risks, and use any money.
I've already made the system.
You do the rest of the work..."

Doesn't sound good. Unfortunately, there are guys like me who's acquired the skills to develop, implement and manage models. There's quite alot of guys in large funds who can do it too. They're the guys you have to compare against, not the average retail developer...

Finally, you're not monitoring systems for operational risks like black outs or down service. You are expected to monitor the market for your system.
 
Quote from TSGannGalt:

What I mention is not about good or bad ideas. It's what you will be dealing with.

Another thing is, you shouldn't expect the firms to use their resources. They've already done you a favor if they decide on going forward, and asking them to use their resources to implement a system will not give them a good impression.

This is coming from a guy who's done what you're looking to do for over 6 years, in Chicago. Only thing you should be asking from them is the capital to implement it. If you start expecting anything other than capital... you'll start giving them an impression as a needy newbie.

It's like:

"I need money to trade the model.
I need a programmer to implement this.
I need a fast computer to run the system.
I need...
I need...
I need...

You guys do all the middle work and I won't give out my system. But I won't take any risks, and use any money.
I've already made the system.
You do the rest of the work..."

Doesn't sound good. Unfortunately, there are guys like me who's acquired the skills to develop, implement and manage models. There's quite alot of guys in large funds who can do it too. They're the guys you have to compare against, not the average retail developer...

Finally, you're not monitoring systems for operational risks like black outs or down service. You are expected to monitor the market for your system.
More...

i see..thanks for the detailed answer....Monitoring markets for structural changes is ok.....the system performance will indirectly signal any significant changes of the assumed underlying market dynamics...I am feeling that starting a hedge fund is not so less attractive as joining a fund if all the "dirty" work will still have to be made by me...))
 
Quote from dima777:

i see..thanks for the detailed answer....Monitoring markets for structural changes is ok.....the system performance will indirectly signal any significant changes of the assumed underlying market dynamics...I am feeling that starting a hedge fund is not so less attractive as joining a fund if all the "dirty" work will still have to be made by me...))
More...

It's not dirty work.
It's about understanding and being able to control aspects involved in trading.

Without control and understanding, there is no management.

It's nothing to be discouraged about. I started off with a prop. firm sponsoring my model and I was only able to use Tradestation. I was able to learn programming and all the other stuff with as I grew my reputation as a developer.
 
Quote from TSGannGalt:

They'll ask for your test results and come very close to reverse engineering them, to expose the risk quantitatively.

[/B]
More...

I wonder how close they can reverse engineer a system based, say, on 1000 past trades.....BTW...I can provide any kind of performance metric, starting from Sharpe ratio,,,why reverse engineer the strategy?
 
Quote from TSGannGalt:

It's not dirty work.
It's about understanding and being able to control aspects involved in trading.

Without control and understanding, there is no management.

It's nothing to be discouraged about. I started off with a prop. firm sponsoring my model and I was only able to use Tradestation. I was able to learn programming and all the other stuff with as I grew my reputation as a developer.
More...

I am coding my strategy on my own now in excel and find it ok in terms of difficulty....I do not mind learning c++ or any other language as long as my top priority stays developing and implementing a system with a positive expectation...
 
Quote from dima777:

I wonder how close they can reverse engineer a system based, say, on 1000 past trades.....BTW...I can provide any kind of performance metric, starting from Sharpe ratio,,,why reverse engineer the strategy?
More...

It's a pickle. Unfortunately there are many reputable firms, even GS, which ask you to present your model in front of their Quants in order to generate new ideas.

I don't mean to discourage you but remember that if you join big banks they make you sign a paper that makes them the sole owner of any piece of code that you develop in their including any modification on your own model. Hedge funds may go easier and let you review the confidentiality agreement with a lawyer before signing up with them.

About your system presentation, remember to layout how systematically you back tested it. To avoid curve fitting impression, you should show both in-sample calibration procedure (if any) and out-of-sample trading performance.
 
Quote from vita:

It's a pickle. Unfortunately there are many reputable firms, even GS, which ask you to present your model in front of their Quants in order to generate new ideas.

I don't mean to discourage you but remember that if you join big banks they make you sign a paper that makes them the sole owner of any piece of code that you develop in their including any modification on your own model. Hedge funds may go easier and let you review the confidentiality agreement with a lawyer before signing up with them.

About your system presentation, remember to layout how systematically you back tested it. To avoid curve fitting impression, you should show both in-sample calibration procedure (if any) and out-of-sample trading performance.
More...

thanks for your kind advice....:) I was thinking about including out-sample tests into the presentation just yesterday! :)
 
Quote from dima777:

thanks for your kind advice....:) I was thinking about including out-sample tests into the presentation just yesterday! :)
More...

Sure, here are other topics they would be interested to see in your presentation:

1) Return (PnL) Histogram
2) Equity curve
3) Drawdown Statistics
4) Sharpe Ratio (correctly calculated)
5) Average No. of trades perday
6) Win/Loss Win/Total ratios
7) Margin to Equity Ratio (Leverage) and how you calculated it
8) Hit ratio which is the same as the number of successful signals per total number of signals within a period of trading (day, weeks, etc)
9) How your stops and limits are calculated or selcted. Or whether you use rolling stops.
10) Risk limit as a percentage of equity capital (not margin) per trade, day, or week (e.g., 1.5%, 5%, 15%)
11) how you taken into account slippage, commissions and fees.
12) Backtesting procedure including in-sample calibration, out-of-sample trading, testing against time shift in data e.g. if you calibration started a month (or a day) earlier or later this should not have significant impact on performance if your system is robust.
13) Other specific details...
 
Quote from vita:

Sure, here are other topics they would be interested to see in your presentation:

1) Return (PnL) Histogram
2) Equity curve
3) Drawdown Statistics
4) Sharpe Ratio (correctly calculated)
5) Average No. of trades perday
6) Win/Loss Win/Total ratios
7) Margin to Equity Ratio (Leverage) and how you calculated it
8) Hit ratio which is the same as the number of successful signals per total number of signals within a period of trading (day, weeks, etc)
9) How your stops and limits are calculated or selcted. Or whether you use rolling stops.
10) Risk limit as a percentage of equity capital (not margin) per trade, day, or week (e.g., 1.5%, 5%, 15%)
11) how you taken into account slippage, commissions and fees.
12) Backtesting procedure including in-sample calibration, out-of-sample trading, testing against time shift in data e.g. if you calibration started a month (or a day) earlier or later this should not have significant impact on performance if your system is robust.
13) Other specific details...
More...

thank you...the list looks pretty comprehensive...
 
You must log in or register to reply here.
Back
Top