457 reputation
27
bio website
location
age
visits member for 3 years
seen Sep 29 at 15:18

Sep
16
comment Do people hedge with leveraged ETFs intraday? How?
Thanks, I am aware of the issues around the close / overnight rebalance. But I think it's not that easy even if you want to hedge, say, till 2pm.
Aug
4
comment Are there providers of delayed market depth data (DOM, Level II, Order-by-Order, etc)?
check with IEX :)
Jul
29
comment Why are we obsessed over normalizing financial data?
cause otherwise you can't compare it?
Jul
25
comment Ito integral approximation by Euler?
seems legit: en.wikipedia.org/wiki/Euler%E2%80%93Maruyama_method
Jul
9
comment Why does [dz(t)]^2 converge to dt over infinitesimally short time periods?
For the first, in the original formula $T$ is not inside the summation. The sum has $N$ terms, so instead of subtracting T from the sum, you can subtract T/N from each term. For the second part - $E(\Delta z_i^2) = T/N$, so your second term is $-2 T^2/N^2,$ so the whole thing under the summation sign adds nicely.
Jul
2
comment Why does [dz(t)]^2 converge to dt over infinitesimally short time periods?
hmm, where do I miss $T/N$?
Jul
2
comment Why does [dz(t)]^2 converge to dt over infinitesimally short time periods?
Judging by the addendum you've made to your answer after reading mine, it's pretty clear to you, how it relates to the question. I'll expand if I have time and there is interest.
Jul
2
comment Why does [dz(t)]^2 converge to dt over infinitesimally short time periods?
no, the last asymptotic is from the law of iterated logarithm. Will try to find a good reference.
Jul
2
comment Why does [dz(t)]^2 converge to dt over infinitesimally short time periods?
@vonjd: no, "i.e. ..." part is wrong. You don't square a martingale, you square the sum before the limit, and the whole point is that 1) it converges 2) only if it's the second power. And both parts are manifestations of CLT.
Jul
2
comment Why does [dz(t)]^2 converge to dt over infinitesimally short time periods?
This explain neither convergence, nor why square is the right power. It's really CLT for a random walk.
Jan
10
comment Book on market microstructure
Out of curiosity, what's so HFT oriented there?
Dec
30
comment How to interpret beta meaningfully?
Just wanted to finish your logic to the complete answer: using the formula for beta above, one can see, that beta(A, relative to B) * beta(B, relative to A) = correlation between A and B. So one shouldn't expect them to be inverse of each over. In your example it just means, that the correlation between the returns is sqrt(0.48 * 0.74) ~ 0.59.
Sep
24
comment Trader's identity in a limit book
@ CharlesM (cntd): You can look at the time gaps between consecutive trades on the feed. What you will see in the statistics, is that there will be a number of events with very small/no gaps, and the rest - with significantly larger gaps. This way you figure out for the reasonable threshold. @chrisaycock: there is no such thing, as "the same time", and I think it's extremely unlikely, that NASDAQ will receive several separate orders within 1 nanosecond.
Sep
24
comment Trader's identity in a limit book
@ CharlesM - I can't say, 0 nanoseconds or 10, but yessure about nanoseconds, but yes, that's the idea
Sep
11
comment Transaction Data with Participant ID
I don't have access to this data, and can't tell 100% sure, but yes, I expect you can (if they still do it) get it from NASDAQ only.
Sep
11
comment Transaction Data with Participant ID
The papers are not hard to find (e.g. faculty.haas.berkeley.edu/hender/HFT-PD.pdf). Notice, that 1) It's NASDAQ, that distributes this info confidentially, so if anything you / your school has to request from them 2) They don't give MPID's, but rather label "HFT/non-HFT" 3) Besides what other people mentioned about identifying HFT by MPID, notice that the labels they give are based on the analysis of the trading activity, not the exact knowledge, about the firm behind MPID.
Dec
21
comment Most natural generalization of covariance/correlation to model dependence of extreme events
Agree with vanguard2k as well. They may not give you what you want, but they are well defined as long as 2nd moment is finite.
Sep
14
comment Evaluating forecasting algorithm
What's the problem? - if your forecast tells the price will go up, assume you buy, if it tell in will go down, short
Aug
16
comment Meta-view of different time-series similarity measures?
I mean that if we go nitpicking, Graph Theory is not a method of analysis of time series. More seriously, I think a question about the relation between two concrete methods of analysis, similar to your example fits the board much better, than a demand for survey of all methods and their interrelations.
Aug
15
comment Meta-view of different time-series similarity measures?
Would you like to make the question more specific? E.g. what is the relation between PCA and the following method in information theory? - otherwise, as you rightly notice, it looks like a collection of buzz words and the quest for unified theory of everything.