Thursday, April 17, 2008

Performance After A Gap And Go Day

One test I ran tonight looked at performance following a day when the market gapped up at least 1% and then tacked on another 1% from open to close. The first table below shows the results for the QQQQ:

Results here are quite negative.

Next I looked at the SPY:

Here results were more mixed with the max loss accounting for all of the losses in most time periods.

I also looked to see how often the gap up got filled when the market was trading below its 200 day moving average. For the Nasdaq 55% of the time the gap was filled within 5 days and 74% of the time within 10 days. For the S&P the results again were not nearly as negative. Only 41% were filled within 5 days and 55% within 10 days.

A partial retracement of today’s move at some point seems likely. It may or may not be a few days before that begins, though.


Alex said...

Do you consider Wednesday's gap in the S&P 500 to be a large gap or mid size gap? Had I waited 2 minutes after the open, I would have been inclined to believe it was a large enough move up to slam shorts. Any thoughts?

Rob Hanna said...

In the tests I've shown, large gaps were defined as 0.75% or greater. Mid-sized were 0.25% to 0.75%. Wednesday the SPY gapped up 0.98% which easily met "large" by my previous definition. This was confirmed by the large gap in QQQQ.

These number aren't magic, but if you get up in that area, you should begin thinking about a short-covering rally rather than a fade.


Anonymous said...

Thanks for all the great analysis. I was curious as to why the number of trades change as you go further into number of 'days in' trade. Shouldn't that stay the same or is there a trigger to take off the trade?

Rob Hanna said...

So that instances aren't double-counted. If the exit is 15 days out and the trigger occurs again 5 days after buying, you are already in the trade. There is no further buying. The shorter the time to exit the more instances you may get.

Sajal said...

Have you realized the impact of outliers on your results? For instance, if I ignored the worst case loss in your conclusion, I'd reach a completely different result for the average trade profit..

Rob Hanna said...

Exactly right Sajal. That's why I 1)Showed it and 2) Mentioned it.

In cases like this where the max gain or loss is significant I try to always show it as a column.


Sajal said...

That sounds good Rob. But my point remains. Say, if you take out the Max loss and reduce the number of trades by 1, you'll get the average trade to be mostly greens. This would be a completely different conclusion! If one cannot reach a consistent conclusion, then where's the edge here! :) What if you reported the median instead? What if you took out the 3-4 outliers and reported the average?

I'm not trying to nit pick, just pointing out that data mining is a hazardous occupation!

Thanks again for your interesting results.