Price of an algorithm for automated trading.

[Ash1972]

Anyone who has developed a worthwhile, robust trading system would in my opinion try to manage both HIS OWN and other people's money with it. He'd put some effort into starting and marketing his own fund rather than just try to flog the code.

What would I need to know to decide whether the system was robust? Well..

1) BACKTEST the system on AT LEAST 25 years' worth of historical data across ALL liquid markets. I'm not interested if the system worked brilliantly on Brent Crude for the last 6 months. The backtesting should assume a realistic transaction cost for each trade. I'd expect to see the backtesting done on software that can test open position management algorithms on a portfolio level - and they're not too many packages like that.

2) Based on these results:

i) What is the worst drawdown? How long, how deep, and could you live with such an experience?

ii) What's the annual variance of returns like? High variance means a very volatile (i.e. painful to live with) system.

iii) Finally - does the average annual return look decent?

And after all that I'd never BUY the algorithm. I'd let the developer manage some cash and pay him 2+20.

 

[TraderZones]

Quote from Unit001:

At this point my opinion is that your system is worthless
equity curve looks ugly, but not only that, I would need a whole lot more data in order to see if your system is worthy anything
8 times more, and I am not kidding

you are asking for 8 million based on trades so far, sorry old boy, I would not be a buyer (at this point)

Based on your rather naive original question, I think your opinion may be about as "worthless" as the system you are judging...

Since we know zero about you, try being a little less judgmental and a little more like the anonymous thread starter you are - humble.

 

[TSGannGalt]

Good replies...

Only if I came into this earlier...

I would have called the OP a retarded monkey...

 

[Unit001]

Quote from TSGannGalt:

Good replies...

Only if I came into this earlier...

I would have called the OP a retarded monkey...

person on this thread claims that he sold an algorithm for 5 million

I suggest we both focus on that if true, rather than on who here is stupid, you or me, or both or none

 

[TSGannGalt]

Quote from Unit001:

person on this thread claims that he sold an algorithm for 5 million

I suggest we both focus on that if true, rather than on who here is stupid, you or me, or both or none

I believe it to be true because:

1. It's Mark Brown. You're talking about a living legend, here. He's well connected in Chicago especially with the top funds and private trading firms. You can't buy reputation and a track record. Personally, I used to follow his posts in Chuck's site for when I first started. When I got my first contract to develop models for a firm, I E-mail the dude and helped me out a lot. He's my "self-proclaimed" mentor of mine, I don't worship him and I'm always critical about what he writes... but I have nothing but respect for him.

2. You have a bunch of small independent quant. groups developing risk models for firms (both financial and non-financial) with a $50+ Million contract. Quantifying risk and having the ability to monitor them is (or at least used to be) big business.

3. There's different kinds of contracts for system developers. The trading industry is small and still a face-to-face business. 1/2 of the $5 Million is most likely from the reputation, track record and name of Mark Brown. The other 1/2 is content of the system, like what everyone mentions scalability, nature of the system like Market Making, Straight Arb., Sophisticated Risk Monitoring, Grey Box / Prop. Scalable, Options, Equity, Futures, Required On-going support/service, providing source... etc. etc.

Finally. Mark Brown does business with his ability as a developer (from what I've heard), not the models or system he developed.

 

[MarkBrown]

Quote from Unit001:

if this is true, how did you solve the problem of proving it works and not giving it away until money transfers to your account?

I would ask for at least 40 million.

easy i funded an account with 200k and after one year net 1.4mil - took my 200 back and they paid additional 3.8mil.

pretty good marketing scheme huh?

off topic check out www.tradevec.com its up and coming!

 

[bwolinsky]

Quote from Unit001:

sorry to scold you some more, but as a system designer myself, I would never trust collective 2 with it
if collective2 involves any kind of code sharing with that site
you would never see me do business with them

I call it common sense

No code sharing, Just results.

 

[knocks420]

700% in 1 year!!!

What kind of leverage, and DD?!

Quote from MarkBrown:

easy i funded an account with 200k and after one year net 1.4mil - took my 200 back and they paid additional 3.8mil.

pretty good marketing scheme huh?

off topic check out www.tradevec.com its up and coming!

 

[TSGannGalt]

Quote from knocks420:

700% in 1 year!!!

What kind of leverage, and DD?!

I wrote about getting help from Mark Brown and based on his advice, the attached recap is what I developed. Unfortunately, I can't provide the actual record of trades that the client did. It was first written in Tradestation and I had the code left in my CD archive so I pulled out the code and ran a test during the period it was live.

The test setup:

E-mini S&P
$4 R/T Commish
$12.50 Slippage after R/T
1 contract

In real life, the size traded was 5-20 contracts so it's not that scalable. It was also running on NQ, ZN, EuroStoxx, DAX, Bobl and a few other smaller index futures.

When I developed this, I approached a prop. firm in Chicago and they took a chance and invested $10,000 on me to trade the model. My contract was 0 salary and 20% of the profits. I couldn't program in any language so I gave the source code to the IT guy in the firm who helped me automate and implement the model.

The deal is not great but it got me started as a systematic trader and a freelance developer.

 

[TSGannGalt]

Adding to the previous post...

- While it lasted, the performance is better than the xls I posted because of the slippage I put in it. (Exiting all trades 1 tick negative of the open of the next bar)

- It doesn't work now. Luckily, system downfall was a gradual process so I didn't get a large DD from the particular system.

- I was very lucky considering how I developed models back then. I sucked but this system became a curse and a standard for the type of systems I "needed" to develop. Also, it was shameful that I couldn't program so I started studying how to, so that I can automate models on my own (which would let me get a larger % of the performance)

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

So... I now have a question:

It took me about 4 months to find a firm that would provide the extra capital and IT resources to implement this live. How much would you have paid for this, if I approached you around Dec. 2002 with Tradestation's hypothetical result? OP, bwolinsky and other system developers... I'd love to hear from you guys...

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

Also, hopefully this serves as a guide towards the calibre of the type of performance and quality for a professional system developer. But there's plenty of one hit wonders, the bigger challenge and the core requirement of this profession is the continuity of the quality.

I find it funny when I see people on Collective or ET posting mediocre systems that they want to sell. More mediocre the developer is, they obsesses about keeping their secret and source. Systems fail and they never last. The live product doesn't matter how long the system backtests well (kind of...), it's about having a system that is profitable "now", not the past or tomorrow (though... the past is a "measure").

Some people have the wrong idea about trading models/systems, even more about the industry and business....