Any advice on taking strategies to hedge fund?

[Punta]

Hi all,

I have been working on automated strategies lately. Based on a technical analysis of historical EOD data, I have found 3-4 trading patterns that, based on backtesting, should be extremely profitable. I am trying to set up meetings with hedge funds, and so far have had one bite - meeting is next week.

The way I see it, assuming that the algos are for real, there are a few options:

1) I go and work for a prop. trading firm, take my algos with me.

2) I sell them the IP, and take some cut of profits.

3) Try and raise capital to do something solo.

Regarding #1: Is anyone aware of typical deals when taking IP like this into a job - obviously in some sense, the IP ownership is different than if the algos were developed while working for the company.

Regarding #2: Is this even plausible? Is it wise to trust a company to volunteer how profitable your algo is turning out for them??

Regarding #3: Is this a crazy idea if I have zero experience running something like this?

 

[Gringotts Bank]

You've put the cart well before the horse.

You must trade it LIVE. Make actual trades, for at leats a few months. Only if your results match the backtesting do you have something of value. What's the hurry??? Go to your meeting with some real broker reports.

I don't for one second believe 900% is possible with a purely mechanical system. I don't even believe 50% is possible, year in year out. 100-200% in one year is absolutely possible, because I have seen managed funds achieve this on odd occasions. But that's very different to what you're claiming.

This is starting to sound a bit crazy, punta. A bit like sir ollivers posts - buying small countries etc. Be real, ok?

 

[Punta]

Yeah I realise that it sounds totally crazy. That's why I'm trying to talk with people who know what they're doing, about taking it forward.

If it helps, I do have a PhD in geophysics, and my profession is based around researching and analysing large real world data sets, so I'm reasonably confident that my technical analysis will stand up. I've also had a friend check my code, so don't think this will boil down to an error there.

Perhaps, as others have suggested, the strategies will turn out to be totally unimplementable due to volume issues etc. Either way, it will be an interesting road...

 

[Gringotts Bank]

Here's some stats I've taken from Jack Schwager's interviews of top US traders:

--Michael Steinhardt (30% annualized return over a 21-year period ending Spring 1988)

--Richard Dennis (who turned a mere $400 into over $200 million over a 17-year period)

--Dana Galante (who compiled a record 15% annualized return by managing a short-only fund during the period from 1994 to 1999)

--Steve Cohen (with an annualized return of 90% over a 7-year period – and who is now in the Forbes Top 400 List)

Now there's punta who can return 900% pa on Aussie stocks!! That would blow even Paul Rotter out of the water.

But you won't even trade your system for one week in order to find out whether it works a little bit?? I conclude you're wasting everyone's time with your threads. Better to stop posting I think.

 

[prawn_86]

I conclude you're wasting everyone's time with your threads. Better to stop posting I think.

Probably could have been worded better, but i think most here will agree.

Trade it live for a week, taking manual entries and see how you end up. If its an EOD system then it should be fairly easy to do. Or even just paper trade it for a month or so showing what entries/exits/volumes you would have taken.

I would love to hear what the hedge fund says, im surprised they would meet with someone who has only backtested data

 

[Punta]

Alright guys, did't mean to instill hostility here.

Maybe I should have posted a chart for a strategy that returns more like 80% and people would have been more open to discussions.

One thing to bear in mind is that it will not be possible to push major dollars through a system like mine. Maybe the returns of systematic funds like Medallion/Winton (who take a similar purely technical approach as I would like to), are made up of some strategies returning very large percentages, and some strategies suitable for moving serious coin that return significantly less.

Also, with EOD data for backtesting, I currently assume that I can buy/sell at exactly open/close, which on the one hand is all you can really do when looking for patterns in EOD data, but might be fatally simplistic.

Either way, I'm not trying to dupe anyone.

As you were....

 

[Gringotts Bank]

Punta, sorry if the tone is a bit harsh but you keep AVOIDING THE ANSWER that is right in front of you.

If you trade it for ONE WEEK, you will know whether it is likely to make -15%pa or 80%pa.

Why won't you do it? Why don't you want the answer? Is it because it might not work out? You have to roll the dice sooner or later. I'd suggest doing it before any meetings, so you have a chance to cancel.

 

[Punta]

Sure, that's obviously great advice. I would love to do exactly that. For a guy like me though, trading a strategy for a week involves a lot more effort on the learning curve than finding and backtesting patterns in data.

To go from never having traded a share to buying and shorting multiple per day, and at the crack of open/close, is actually a steep learning curve to climb.

On the other hand, emailing a cover letter and CV to prop. funds is relatively easy!

Still though, totally agreed, it would be far better to do something 'live' even on a tiny scale, before taking it to prop. funds. Good advice.

 

[tothemax6]

Why don't you run it on the market, live, but pretend? I.e. 'paper trading'?

Secondly, as far as I know, IP does not exist in the algo trading industry. The 'IP' is 'I keep it secret'. It is not the same business as say, physical products - where you can detect if a competitor is using your patents. Hedge funds do not have to divulge the algorithms which their computers use to trade, hence patent enforcement is impossible, and the patent is actually damaging since it divulges the secrets.

Thirdly, does the algorithm have to trade so many times per day? Could you tell it to randomly trade only 1 in 4 of each signaled trade? Or only the 'best' signals?
Really, if you are that sure your system will work, you would find a way to execute it on the market.

 

[boofis]

....For a guy like me though, trading a strategy for a week involves a lot more effort on the learning curve than finding and backtesting patterns in data.....

Alright, I'll do you a deal....You send me your strategies, I'll trade it with 5k initial investment for 4 weeks and I'll give you back half of what I earn/lose. Heads I win, tails you lose. How could you pass up such an offer?! :