Search results

Yes I agree that the spread is not due to the low interest from institutions. I was thinking that in the absolute spread sense, the higher priced stocks will get you more spread because of the fixed percentage you mentioned (i.e., 0.03% of 200 is wider than 0.03% of 20). Also, I keep thinking...

Got it that makes sense too. If you are assuming a 1 tick spread yes that's true. If you look for very illiquid instruments however, the spread often is proportional to the price (e.g., $300 instrument might have 10 cent spread, whereas $10 instrument might be down to one or two pennies) so in...

I see, that makes sense -- but if your commission is per share (my case at IB), then its the other way around right?

Yes I think you are right on with this as well. I am in fact looking into this and so far it appears opportunities are few and far between, but I think they do exist and are very likely exploitable... One question: you mentioned the larger firms are interested in lower priced instruments...

very nice descriptions, thanks for these -- are you trying to implement variations of these yourself?

Thanks for this, this is really good info / motivation. I have been studying the order book lately and it does seem like it will lead to a viable strategy.

Basically you are saying "hey I got lucky for a while, and so can you (1 in 10 chance)!"

Good luck. I think you may be on to something as more than a handful of academic quant finance papers have pointed out the statistical significance of 3-5 day negative correlation of returns.

found this published in 2010 http://mitsloan.mit.edu/neudc/papers/paper_383.pdf talks about how large institutions exploit inefficiencies which always disappear as the years go by. Looks at an interesting case of the Indian commodities market which opened in 2003 and started out with a lot...

There is a well known statistical fact: the poorer your statistics (e.g., small sample size, existence of correlations), the more likely you will get extreme values.

April fools!!

I think it doesn't make sense to do what you did. As you pointed out yourself, a regression assumes a linear relationship like: PL = a0+ a1*param1 + a2*param2 + ... However as we all know changing the parameters of your system can make a profitable system to be completely unprofitable, and...

Actually my experience is IB TWS does not hog the CPU. I run it every day during market hours and my computer is cool as a cucumber. However, what I find is Firefox is extremely expensive. It eats not only CPU time, but has a huge memory footprint. Moreover there are definite bugs in Firefox --...

This is my biggest worry about swing trading. In swing trading, your trading frequency prevents you from really knowing if you have a statistically significant positive expectancy. Those of you who think day-trading is impossible and swing trading easier are likely only riding on the illusion of...

I found the microsoft search indexer a huge nuisance that slows down my computer AND keeps the fan on high all the time. I never use this slow and clunky search feature in microsoft anyway so I turned it off and now my computer is fast, cool, and quiet. This indexer will not show up in task...

So what advice would you give a trader who is just starting at a major bank, hedge fund, or top prop firm. They should just give up? Or stick to HFT, that's it?

In other words, to further clarify, out of the 12, there were 10 with okay sample size (still small but may be okay). Out of the 10, 5 worked, 4 failed, and 1 was borderline. If the borderline is conservatively assigned to the failed side, then 5 worked, and 5 failed. The 5 ones that worked will...

Okay thanks. Just looking at the window you attached, it looks like out of the 12 visible signals in the window, may be 5 of them look somewhat real (around 10 or greater trades, with around 65-70% success rate or better), 4 of them had success rate <= 50% (i.e., didn't work), 2 of them with...

What is the "CL" column? Thanks.

Yes, I can see in this case, it has a good chance of being significant -- i.e., if you optimized only once in sample, then run a much larger OOS and it is still very profitable. The null hypothesis version would be: run 1000 random sims with different random number seeds. Pick the best...