deathflute

How did you guys arrive at 15% for 150K?

What do you guys think about the terms of the deal? Isn't 15% a lot to give out for just 150k?

I have been part of a team which used kdb heavily and this is not really true. The best part of kdb is that it requires almost no administration and at the end of the day becomes just another programming environment/tool for you to use. Hiring a team of people who just do kdb is only done by banks or other unimaginative large teams.

Well written article. I think most people who do not have a background in data are unaware of the various options out there and fall for the marketing behind hadoop like tools.

I would urge people doing analytics to take a look at kdb+ from kx. Unless you have ridiculously large amounts of data(> 200 TB), I can bet that you would be better off with kdb. The only downside is that it costs a lot of money which is a pity.

Thanks for an excellent post. I would like to see any one indicator explained in detail as well. I am sure it would be of interest to a lot of people.

I am also confused about how latency arbs work. If you are co-located at one exchange, wouldn't you be further away from other exchanges?

"Then you play the same game on the opposite side to unload your position."

Don't you have to play the game at both sides of the book simultaneously? Otherwise, you would risk adverse selection.

It is interesting that you believe that E(compensation in a start up) > E(compensation running a HFT strategy). The popular media is awash with reports of billions of dollars being made by HFT traders. Are they completely off?

What about HFTs that are running market taking strategies?

To add to (b) Midpoint orders have to be hidden though because lit markets cannot allow displayable sub-penny orders.

Thanks, this is a very clear article on the most important ingredients of HFT. I am curious to hear the apology but I am not sure there is anything to apologize for. The market marker is taking on substantial risk of adverse selection and is being compensated for it. As long as legal means are used to achieve that end, this is not really different from any other business strategy. Causal implied benefits to society from any business are subjective and the most important consequences are usually latent.

Anyways, I am curious to know if elaborate prediction models are used here or it is mostly a game of adjusting prices based on correlated instruments? Also, how does one build a simulator for such strategies?

Hi,

I might have some interesting projects for you. Can you email me at hn At machine Dot imap Dot cc? I couldn't find an email address in your profile.

As they say there is more than one way to mouth a cliche.

I have always been skeptical about such services. It is like collecting pennies in front of a steamroller. Sooner or later, a tail event like this is bound to happen. A distribution of hosting experiences has to be fat-tailed. It is okay to deal with that risk when even the worst case scenario is acceptable. But, for your own home how can it be acceptable?

Most of what you wrote made some sense until I read the bit about linked-in being indispensable ;)

Arthur's languages are awesome, no doubt about it. But the fact remains that they are proprietary and prohibitively so. As much as you romanticize k/q, I bet you can't even afford to run it on your home computer, let alone do any real work on it.

Yep. Money is not lost, it just belongs to someone else ;)

Sorry, that doesn't make sense. HFT is actually not at all capital intensive compared to some other strategies that hedge funds run. The whole point of HFT is to produce high sharpe returns on a small capital base. These strategies do not scale.

I agree. But if they are making money off it, then why are they talking about it? The trading community hasn't exactly been known for its altruism.

It is quite puzzling why all of a sudden there is a rush to write about hft. Historically, systematic trading has been very secretive and quite understandably so. And now all of a sudden you can read about building hft systems on wordpress blogs. Either returns have virtually diminished from such strategies or in the view of recent popular backlash, there is a concerted drive to talk about it.

Thanks for recommending that. I'll probably pick up his latest algorithms book.

Sorry, I should have been more specific.CLR is a great book, but it is too referential IMO. I am looking for an easier reading.

I agree. 2000$ is a _very_ cheap 1 BDR in manhattan, more likely an "alcove" studio. You really have to go 2500$+ to have anything decent.

I should have been more clear - My goal is to become a competent lisp programmer first and foremost. I will probably end up choosing clojure for my projects because I already have lots of java code and a jvm lisp dialect is more suitable.

I would scale down your medical operation bit by bit while spending more time building the skills necessary to launch your start up. Only once you are confident enough of pulling it off, would I work full time on it. Basically, not only you have the risk of your start up being a complete failure but also that you might not be able to pull it off technically. The latter risk can be completely hedged away by my first piece of advice.

Definitely Big Sur. Eat at the Big Sur bakery.

How can I make short-term investments in early stage startups?

I like your point of data oriented design. Do you have any suggestions/links on how I can learn more about it? Thanks.

I wanted to write a high-level API on top on multiprocessing, but it seems that this project has already done this. However, it is unclear if this has been used/tested much with python 2.6.

If you have used this package, please share your comments.

Thanks for your detailed response!

I was looking at those. Do you have any experience with them?