[aleyan]

Congratulations to the quantopian team.

Trading on real data is very different from simulated trading on historic data. There are a lot of leaky abstractions and assumptions that happen during simulations that cause problems once you put real money to work. In the past, here on HN, I have criticized quantopian several times because no one has actually used quantopian to trade real money. This is no longer true.

It will take a lot of time and a lot more capital to figure out all the holes, but they are finally on the right track and I am glad they have 20 people already trading. That is 20 sets of eyes balls to look for lack of fidelity in the simulations.

Personally I am interested to see how good their transaction cost model is when trading hundreds to thousands of shares at a clip.

[noname123]

Yes, 1000x! I'm waiting to see Quantopian's forward testing with results.

Personally, I couldn't wait for their roll-out. So I wrote my own IB client and traded my Quantopian backtested strategies myself on my IB test account. Race condition, deadlocks much at 3:59PM ET at 2000+/msg burst from quote ticks (esp. when you are doing some kind of stat arb, pairs trading where one order submission is dependent on another order's fill).

Anyways interesting technical problems abound. Gl to the Quantopian team!

[novaleaf]

very interesting!!! but I'd like to try writing my own algorithm offline, in my own prefered language. (javascript)

does anyone know where I can get price history for the last few years for something like an index fund, like DIA?

[wkneepkens]

Yahoo Finance is a good option, also have an API.

[novaleaf]

thanks, i'll give that a try