Comment by svedlin | Hacker News Reader

Comment by svedlin | Hacker News Reader

Folks interested in HFT might like this - HFT firm Virtu has a nearly flawless daily trading record over the past 3.4 years:

http://www.businessinsider.com/virtu-hft-only-one-losing-day...

Prospectus:

https://www.sec.gov/Archives/edgar/data/1592386/000104746914...

Looking at the S&P 500, it was positive for 55.8% of the trading days between 2010 and 2013:

http://www.crestmontresearch.com/docs/Stock-Yo-Yo.pdf

BofA had 10 trading day losses in 2013:

http://blogs.marketwatch.com/thetell/2014/02/28/goldmans-tra...

US equity HFT revenues are around $1.3 billion, making it a pretty small niche compared to wider market activity: "TABB Group estimates that US equity HFT revenues have declined from approximately $7.2 billion in 2009 to about $1.3 billion in 2014."

Law of large numbers. This is not surprising.

EDIT: Law of large numbers. Not central limit theorem. I blame alcohol.

Um, what? If anything, applying the theorem in this case would say that profits should be normally distributed, whereas it's fairly common for market makers to have perfect/near-perfect trading records.

Sorry, that was vague. My point was that market participants who make many trades will rapidly converge on their mean return, with little fluctuation. If you sample a normal distribution 10M times with a mean of 0.01%, you will very rarely see a negative sum of the samples.

The theorem says nothing of the sort. By your argument, if I flip a coin 10M times, where heads is 1 and tails is -1, and get a mean of 0, then I will very rarely see tails. You're confusing mean and variance.

Sorry. I said some pretty silly things.

See the new version of the posts. Hopefully it's a bit more coherent now.

Ok, I think your main idea was that since they've achieved consistently positive returns, then they're likely to keep it up. Even if that were true, the 800lb gorilla question is how they've managed to achieve such consistency in the first place.

They have slightly positive return per trade, and they make millions of trades per day. Their PnL is the sum of their trades -- the expected values grows like O(n), while the volatility grows like O(sqrt(n)), so for n = several million, the probability their PnL will be negative is extraordinarily low.

Reply on news.ycombinator.com