Gaps Trading - Too Good To Be True?

[Laissez Faire]

Quote from Pipflow:

I tried this gap trading earlier, but it was true that these gaps actually did not close in certain currency pairs and thus i had to incur some huge losses. Not a good strategy to rely upon.

Just to expand on what I said in my previous post.

First of all, I don't consider gaps in currencies the same way I would an index. I have limited experience with currencies and while they have sessions for each time zone, it is a market that trades continuously, so I'm inclined to believe that the open price for the NY session compared to 24 hours ago matters less than in equities. The gap from the weekend close may be of some significance though, but that's only a maximum of one per week.

Index futures also trade overnight, but it's not a 'true' 24 hour market since it's based on the cash market that trades 6 1/2 hour each day.

Second, it seems like your strategy was only based on gap closes? Many times, a gap will continue in the direction of the gap. Blindly betting on a gap to close is a losing strategy. Often the gap is a sign that you should trade WITH the gap, not against it.

Most gaps close sooner or later on a long enough time line, but obviously this is about day trading and what is relevant then if the gap closes today or not. A larger gap may often be the start of a larger move and then that gap will be left behind for a long, long time.

 

[gmst]

Quote from Kevin Schmit:

Good link. Incomplete though. Arxiv has some recent papers that haven't made it to the CFM site yet:

http://arxiv.org/find/all/1/au:+Bouchaud/0/1/0/all/0/1

I found 1309.3102, the nested factor model paper from September 2013, particularly interesting. Also one of his random matrix papers was very useful to me. Often, as in the above referenced factor decomposition paper, Bouchaud takes the method only so far, apparently leaving the working out of the next logical step to the reader.

Not all of his papers are useful. He occasionally publishes papers that seem to be complete misdirection, and also co-authors with a number of cranks-in-all-but-name.

Thank You Kevin!

Never knew that Cornell library is freely available online!!

Going through the paper seems like a lot of work and then implementing it also seems like a lot of work You found it very useful in your trading - if yes, in what way? Did it improve your sharpe, lead to more consistent returns, reduced risk, led to better stock portfolios etc.?

 

[Scataphagos]

Quote from lightrader:

Hi , I recently read in the web a few stories of people that claim to make a lot of money in gap day trading strategies (about 1-2% per day on average, net of commissions and losing trades).

Stock index futures gaps can be traded also.... "gap and go", "gap and crap", others. The key is to figure out which one it is today... which is more likely, if you can... and be flexible when you see "It's not the one I played for".

 

[Joman]

Quote from Kevin Schmit:

Good link. Incomplete though. Arxiv has some recent papers that haven't made it to the CFM site yet:

http://arxiv.org/find/all/1/au:+Bouchaud/0/1/0/all/0/1

I found 1309.3102, the nested factor model paper from September 2013, particularly interesting. Also one of his random matrix papers was very useful to me. Often, as in the above referenced factor decomposition paper, Bouchaud takes the method only so far, apparently leaving the working out of the next logical step to the reader.

Not all of his papers are useful. He occasionally publishes papers that seem to be complete misdirection, and also co-authors with a number of cranks-in-all-but-name.

Thanks for the additional information Kevin, I'll have a look at his research again.

 

[gmst]

Quote from gmst:

Thank You Kevin!

Never knew that Cornell library is freely available online!!

Going through the paper seems like a lot of work and then implementing it also seems like a lot of work You found it very useful in your trading - if yes, in what way? Did it improve your sharpe, lead to more consistent returns, reduced risk, led to better stock portfolios etc.?

So, I went thru the paper (that Kevin quoted) in a bit more detail. From an applications point of view, I have come to following conclusion:

This paper will prove useful to risk manage or construct superior equity-equity options-index-index options portfolios (basically portfolios with convex payoffs) which have been hitherto modeled using copulas. Techniques in this paper are more in line with real market experience and are better than bivariate copulas.

However, for an individual trader or even a small group of 4-5 traders, information in this paper is useless (I am willing to change my opinion if Kevin or someone else can convince me otherwise.) Since no-one on ET other than if they work for a 100+million fund, are going to apply such advanced analytics for their risk management.

 

[Kevin Schmit]

Quote from gmst:

However, for an individual trader or even a small group of 4-5 traders, information in this paper is useless ...

I believe you are incorrect.The obtained common vol factors can be used to re-estimate the original 10 or so factors, which were originally built off of log price changes scaled to unit variance. Then iterate until convergence or near convergence. This leads to a significantly superior factor decomposition of market price changes. Superior factor loadings lead to superior trading for individuals or small groups.

If I have time later, I will reply to your previous post and lay out how this, and some of Bouchaud's other work (on the level of endogeneity of overnight returns and vols) can be applied to index gap trading.

Of all the authors I have found most helpful (Tsay, Stambaugh, Jarrow, Box & Tiao, etc.), Bouchaud is the only one who runs a successful multi-billion dollar hedge fund. Perhaps that is why I find his methods more directly applicable.

 

[gmst]

Quote from Kevin Schmit:


If I have time later, I will reply to your previous post and lay out how this, and some of Bouchaud's other work (on the level of endogeneity of overnight returns and vols) can be applied to index gap trading.

Kevin, I would be very interested in that reply. Thanks a lot!

 

[gmst]

Quote from Kevin Schmit:

Of all the authors I have found most helpful (Tsay, Stambaugh, Jarrow, Box & Tiao, etc.), Bouchaud is the only one who runs a successful multi-billion dollar hedge fund. Perhaps that is why I find his methods more directly applicable.

I do not remember reading anything from Tsay and Stambaugh - just shows there is so much there that I never got a chance to explore. Curious - You don't think Andrew Lo's work is practically helpful?

 

[Ghost of Cutten]

Tip - most gaps have no predictive value. Try to focus on what type of gaps, and what subsequent market behaviour, are highly predictive. Greater magnitude of gap, stronger emotions, more important news, and higher volatility off the open are all signs that there may be something interesting going on.

Most really good trades are pretty easy to identify, but rare. The trick is in passing up all those more marginal trade setups that are hard to identify, and thus hard to trade, but which happen very frequently. Good setups almost trade themselves, average setups are tricky and easy to make mistakes with.

 

[xelite777]

Quote from Ghost of Cutten:

Tip - most gaps have no predictive value

Hi Ghost of Cutten.

Is this statement based on extensive research or simply your personal opinion?