Search results

You should think about your opportunity cost (what you be doing with your capital when you are out of the trade) and how many trades you get to take with each system. Maybe you could look at a day as your unit instead of a trade. For the time you are not in your trade assume you are in cash...

Thanks for taking questions on this stuff! Have you followed the academic research on HFT, and do you have any opinions on it? One particular question I have, there is a paper by Baron Brogaard and Kirilenko that use S&P 500 futures data to estimate how much HFTs make trading the ES. They...

Yes, those were the ones I had in mind.

Three academic papers you may want to look at - Amihud (2002), Pastor and Stambaugh (I forget the year), and Chordia Roll and Subrahmanyam (2005). The Amihud and P&S methods are crude to avoid using intraday data and trade signing, but get close to estimating how much volume it takes to move a...

That is not what I see in the data. The 1-minute price impact I mentioned is the same if I look 5 or 30 minutes out, so it looks like it is permanent. This is consistent with the theory discussed in Harris, where on average the trader that sends the marketable order is better informed. But...

I ran a test that suggests (1) and/or (2) is likely to be true. On a large sample of NASDAQ trades I calculated the average effective spread as 2.5 bps and the average 1-minute price impact as 3.6 bps (both relative to the pre-trade midpoint). The price impact measures the change in midpoint...

if you mean the equity risk premium, close but not exactly: http://www.indexuniverse.com/sections/news/8869-factorshares-launches-5-spread-etfs.html

Look at this paper on trading mutual funds with stale prices for some ideas: http://pages.stern.nyu.edu/~rwhitela/papers/fundtiming%20faj02.pdf I think I have seen 3 or 4 others like this, but don't remember anything with a dramatically different approach. Just brainstorming, if you can...

CFA won't open doors. Not having it may close some though.

Sorry, most of us without Asperger's have ADD.

I gotta back up Mike here. I had a boss who was getting a phd at night - he had worked his way into a quant position without a phd and felt like he needed one but had a good job he didn't want to leave. Serious university? Not top 10, but definitely not a joke. How do you define serious...

Unfortunately this is tough. I used to work at two fixed income places that didn't really care what people did as long as you didn't trade in the stocks of companies we did business with, but I think that is rare. One was kind of cool - a bunch of underemployed analysts with bloomberg...

I know the conversation has strayed a little, but back to the original spy vs es part ... there is an interesting angle nobody mentioned. I think the answer depends on what you think you can achieve with your cash while you are rolling es, that could swamp the other cost considerations. With...

I still think a logit regression is your best bet. The problem with a normal regression is you will probably get predicted values < 0 or > 1. A logit (or probit, they are similar) doesn't have this issue - predicted values are all between 0 and 1. And the output can be interpretes as the odds...

You probably want to try a logit or probit model.

Can you compare VWAP over short term bars over the two data sources? That should be more robust to individual trade prices than OHLC. If that is close but you still need OHLC for your strategy, maybe you can try your system on bars built out of short term VWAPS, like a 15 minute bar made out...

I have created equal weighted indexes/baskets with equal dollar weights in each. Not for any deep reason, it just seems like the simplest approach and is good enough for me. I usually work in returns, so the index return for and index of A, B, and C would be 1/3 the return on A + 1/3 return B...

I have been thinking about this stuff for a while but haven't had time to get past collecting data. I have worked with databases of limit order books but not for anything predictive. My approach will probably be to come up with a few measurements that capture the status of the book and do...

I disagree, it is not that simple of a question. In general there are noise terms in autoregressive models so they are not deterministic. Autocorrelation of 1: P2 = P1 + noise ---> random walk. Autocorrelation of -1: P2 = - P1 + noise ---> don't know what to call this and can't exist...

My take on this is that many of the posters in this thread are groping different parts of the elephant. Most of what I have read is right, but right under a set of assumptions that are not quite specified and may not apply. The gambler's fallacy is relevant when you have independent...