Quant technology

I’m a technology dude (software, EE, math) partnered with a duo of other other technology guys (no WS, general finance, analyst, trading practice) and we’ve built a black box system that generates near-term trading signals (long and brief) for equities, ETFs , and indices (and most likely commodities and currencies – but we’ve not tested that yet). The accuracy levels for both win/loss rates and percent come back per signal emerges to be very high (to us, anyway). Anyone have any insights on how to get such a system in front of people who could direct funding and investment capital toward us?

While the core analysis engine is adequately accomplish for us to test/backtest the system there is substantial work that can be done (and should be done) to improve spectacle, scale up, automate and integrate, etc. We’d consider taking funding for these reasons – and ideally we’d want a playmate who could compliment our system with all the abilities and resources we lack.

I appreciate any thoughts you might have.

• Log in or register to post comments

Want to Vote on this Content?! No WSO Credits ?

Comments ( five )

• ny_trader
• ?
• | Four
• | Rank: Chimp

Well very first you want to make sure that you have a legitimate system and not something that will be throated out in actual trading . Lots of ideas that look good on paper/backtesting get absolutely killed when it comes to factoring in slippage, transacation costs, errors/biases in backtesting, and just simply overfitting the data.

Does your model work on a broad range of ETFs, equities? That would very likely avoid overfitting bias. Does it generate a sufficently large pnl/trade to make up for slippage and commissions? Then you can most likely embark thinking about the next step.

Want to Vote on this Content?! No WSO Credits ?

• wiseclam
• O
• | 7
• | Rank: Chimp

Excellent comments and questions. Let me take a shot at answering them and you (and others) can comment on next steps, if applicable.

The “signals” our system generates are buy/sell signals that are intended to identify a trading chance that would last days. A typical “buy and hold” trade might be ten days (while we are not limited to this, let’s assume a ten day buy-and-hold strategy for this discussion). Ten days of movement is, on average, a decent size budge – I think enough to overcome commissions and to reduce the influence of slippage.

We can run against any security with serial (typically daily) data. We’ve found that higher liquidity securities tend to yield greater accuracy. And we can run the system (For Your Information. it’s not a “model” – it’s not an algorithm or even a set of algorithms that are static) and see excellent results against a broad range of equities, ETFs and indices. Of course we can also generate bad results, so a key part of the system is a method to differentiate good securities/signals from bad, a priori.

So. I think the longer trades (multi-day) and the high liquidity securities helps address some of the issues you are raising (volume/liquidity, transaction costs, slippage).

On overfitting and general spectacle evaluation. we evaluate all of our results (signals) relative to chance. If a particular ETF traded up 55% of the days over a backtest period we need to see results that radically exceed a 55% “win rate”, as this is achievable with random trading . The same rules apply to the average comebacks for our signals – they need to be 2X or more above chance trading comes back rates.

We apply what I think is a very high standard to our results. But of course the ultimate standard is account value growth, and we haven’t got there yet. We’d like to further refine/enhance the core engine in our system; address automation and integration ( back office stuff) functionality and generally scale our capability (e.g., run thousands of securities, not ems of securities) before we get to a place where we would actively/strenuously trade. There are also abilities issues. we are not WS guys and we don’t know how to optimally trade, manage risks, etc.

One model for “next steps” is to playmate with a hard that can bring a) venture capital to develop/accomplish the system, b) investment capital to run a petite fund based on our signals and c) expertise and resources to run such a fund effectively. Other approaches exist, too, of course, including a unspoiled bootstrap treatment. I’m not sure we have the financial means to go that route – again, we’re tech guys not very compensated WS guys :-).

Want to Vote on this Content?! No WSO Credits ?

• nameback
• HF
• | 82
• | Rank: Senior Monkey

Wiseclam — I’m right there with you, mate! I’m a stats stud and my fucking partners are stats guys and we’ve built a quant system that backtests utterly well, and I’ll just tell you what hurdles we’ve encountered so far.

(1) Find a professional trader you know, through a friend or whatever, who can look over your backtesting and make sure you’re not leaving out any significant factors, like transaction costs and slippage. This will be one of the very first things people ask you. over and over and over again.

(Two) No one believes backtesting in finance. Why? Because most of it isn’t rigorous, because most finance people aren’t well versed enough in statistical rigor to backtest with rigorously out-of-training-set data. Overfit is rampant. Our solution to this was to raise $70k from friends and family for a proof-of-concept fund, which we’ve been running for two months and is up 15.3% so far. We’re attempting to raise money with this track record, and our extensive, rigorous backtesting, but most places want to see six months – one year of real-world PnL before investing; and that’s the ones that are friendly to start-ups.

(Trio) Legal fees and back office support. It takes 15-20k to get a hedge fund set up, at the low end of the range. Then you have to think about whether you need to be or want to be registered investment advisors (registration takes a while and costs another few thousand). Then you need back office support, like auditors/accounting and infrastructure. Fund of Funds and Seeders will look for this stuff. There are places that provide funding as well as backoffice support, but the more infrastructure you have, the lighter it will be. We have none, and it’s rough going.

(Four) Develop marketing materials. They need to look slick and include a shitload of detail. We have some decent ones but literally every time we send them out we end up revising them to include more detail because people ask for things we didn’t think to include. Not just your backtesting, not just your live testing, but also methodology, individual bios, philosophy, strategy, process, ratios (beta, Sharpe, Sorentino, etc), peak-to-trough analysis, volatility, etc.

(Five) Pound the pavement. Network incessantly and go to every finance/hedge fund event you can find — crash them if you have to. Get your model in front of high net worth individuals that you know or your friends/family know, and get your model in front of Fund of Funds and Seeder Funds. Give out a lot of business cards. Get a lot of business cards. Keep at it until you find that crucial anchor investor willing to take a risk on an unproven system and give you that very first $1mil in AUM, and grow from there.

(6) Everyone wants you to already have money, already have investors, already have backoffice support. It’s one of those “I need a car to get a job but I need a job to get a car” type of situations. This is why when commencing up it’s lighter to concentrate on high net worth individuals than Fund of Funds and Seeders, because individuals are less stringent generally, however still very hard to get.

We’ve been building our system for eight months, live testing for two, and raising money for two, and we’re basically nowhere. It’s rough out there, but if we keep at it we know that we’ll eventually break through and get some real AUM, and once our legal shit and back office shit is taken care of and we have 1-5mil under our belt, building on it will be MUCH lighter.

Edit: It’s funny, your system sounds somewhat similar to ours, actually. I’d be interested in learning more about it, feel free to PM me.

Related movie: