Algorithmic Trading for Only $1 Per Trade

[xandman]

Btw, adding to the preposterousness of creating your own homed-baked HFT strategy:

The best algorithmic designers are being hired by vendors to create algos for EFFICIENT EXECUTION, not profits. If there were easy, reliable profits to be made ITG would have gone prop years ago.

Along with Tom Cruise, Jack Schwager has yet to release an HFT Wizards book.

 

[Gambit]

you can guess that this is an impossibility. Of course they receive payment for order flow. No way you can direct your order to specific exchanges at such cost.

I'm guessing the same but "beggars can't be choosers".

 

[Gambit]

Actually, the order flow is divided between Knight and Citadel. Talk about shark tank.

Can't this be mitigated by trading a higher time frame? You definitely don't want to play order book games with these guys. There is still a lot of room for other strategies that may involve carry, momentum and reversion at higher time frames. Additionally, this implementation allows for Quandle and other data sources. I have yet to see somebody incorporate Social Sentiment on the desktop based NinjaTrader.

I think there is an issue with techies and college kids who try to apply their domain knowledge in an arena that they are ill equipped to handle. The actual expertise required to play within the 1 min bar will not be found in the Amazon best seller list.

I guess we are all eager to see Tom Cruise in a new HFT movie.

Agree with this as well. Just pay the toll to Knight/Citadel and execute your strategy.

 

[xandman]

The domain knowledge required to build a viable HFT "business" would take years to build. Decades for me.

Hey, I could lock it all up for a holy grail, asset rotation model and go to the beach. Unfortunately, I am aware that there are other sources of fleeting alpha that could prove incredibly rewarding.

 

[xandman]

A very brief review of the docs indicate they have only Minute, Second, and Tick resolution of the data stream. If you want tick bars, you must construct them yourself.....the same if you want 5 minute bars.

Did you peruse the LEAN algorithmic engine? They have it on Github. Hard to believe some young Kiwi is the architect for the whole thing. Can't help but think he is just a spokesman for the Tradier/Citadel/Knight shark tank. Nice chap, would have to be a bonafide genius. But, geniuses working alone usually don't build sites with such polish.

 

[Sergio77]

HFT at $1/trade is the sure way to hell.

 

[globalarbtrader]

I guess there are two questions here. Does quantconnect make sense? Does tradier @ $1 make sense because you can also connect quantconnect to IB, but at the normal fee level. It looks like other people have answered the latter question - tradier probably not the way to go.

Quantconnect is basically a nice looking trade/research API, and a retail platform for a low latency connection.

On the API side the open source part is nice. Personally I would have liked to see it a a non proprietary language. This suggests the user base they're aiming at is relatively inexperienced. Me, I'm too old to learn another language. It's relatively easy to create libraries so people can hook in trading algo's writing java or python so it looks almost like pseudo code and is easy for non specialists to write.

If you're trading relatively slowly then you're essentially getting just a robust server to protect you from internet downtime or power failure at home. There are easier ways of doing just that (eg ec2, which I haven't used myself but I know people who are happy with it).

For serious low latency traders there are obviously plenty of other things out there, eg http://www.fixnetix.com/services/managed-services/ (same disclaimer applies).

So if I describe the user base that this product fills a precise niche for: relatively inexperienced people who want a nice API, don't want to have to learn a real programming language, and need a low latency connection.... it scares me a little.

GAT
Read my trolling policy, here.

 

[syswizard]

HFT at $1/trade is the sure way to hell.

I don't quite understand...
why ?

 

[VPhantom]

Hard to believe some young Kiwi is the architect for the whole thing.

FYI the GitHub contributors page shows 13 different sources of commits, though the bulk does seem to come from just 2. Punch card shows that commits usually come during business hours plus plenty of evenings. Unfortunately because the initial commit in January was already some 26,000 lines, we can't really tell if it could've been realistically done by just that one person. If he's been at it for 6-12 months before the initial commit, full-time, it looks plausible.

 

[VPhantom]

I would have liked to see it a a non proprietary language. This suggests the user base they're aiming at is relatively inexperienced.

So that'd be a +1 for Quantopian then, in Python. (Although they still have yet to disclose their actual fees — they're still in beta I guess. And if their library's proprietary (not sure) it's not really that much better than LEAN anyway.) However isn't LEAN built around C#/.NET? Not that I'm a fan but it's not quite as proprietary if that's the case.

As someone who looked into coding algos in Java 12 years ago, and ran out of patience for the platform, I'd be curious to see how things improved since then with higher-level languages like Python and decent open-source libraries. I like the hosted services (Quantopian, Quantconnect, etc.) for the turnkey aspect (backtest data included, hosted executions) but maybe homebrew on a VPS is a viable option too these days...