At how many standard deviations away from the mean would you guys normally give up on a spread trade? When do you say "something's wrong, I'm bailing out"?
with it being earnings season, i'm having trouble finding any pairs without news. how long do you guys wait before entering a pair that has had news recently? specifically earnings.
The further away the better IMO, with the stdev you can double. On a losing trade usually the thing start to drift way slowly and therefore the stdev does not shoot up that much. So the mean will catch up. If something really go wrong then it's more judgmental than math. If the distance from mean really goes to 5 you look at the story that the stocks have to tell and you make a judgment call.
In mean reversion strategies you have some justification to double down on a losing position, but if you keep your "bailout" point too far away (in terms of standard deviations) then you risk making a serious loss in the event of a merger or the pair relationship falling apart. I'm backtesting so I can't "make a judgement call". I need a quantitative answer, but thanks for your input.
True, but in my experience this is not how it works. Total Keops is right in what he says about stddev. When the trades goes wrong, the ratio tends to hug the 2x stddev line as it goes against you. The problem is that, when/if it comes back in to say, the mean, you could still be losing money. One solution is to set initial stop loss/take gain targets based on the stdev parameters when you open the position. However, the problem with this is that, this may stop you out just when the trade is a about to turn in your favour. I would suggest incorporating a quantitative target of % from the mean in order to deal with the problems of solely using stddev. I'd be interested in what you come up with as a solution to this problem.
I've seen pair spreads that pushed past 2 standard deviations and made it to 3. No disrespect intended, but I think you might have just been lucky if your experiences have shown pairs not to spread past 2 stddev. I don't see the benefit of using a % from the mean rather than standard deviation for the "bailout" point. Standard deviation tells you the probability of the spread reaching a certain point, whereas a % is rather arbitrary.
No diss taken. You can easilly find 4 stdev away. I'm sure there are 5-6 every now and then. But I dont think that the biggest losers are on big stdev moves. Again, the stdev may not be a good trig to get out. And BTW mergers are announced overnight so if your in there is nothing you can do. I suggest you try many different set-up and see how it reacts; time stop, % stop, $ amount stop, close at the mean ... Try to spread the big losers among many trades so they dont count as much overall? In any case, as long as you get far from the mean in absolute value or in stdev I dont get nervous usually, what I dont like is when in the red and the mean catch up but the ratio dose not go back to the mean. Look at the chart of the losers and see how they go. And dont forget that the events of august-november 2008 where extremes and may not come back in the next 10 years (I know they will some day). You may want to discard that data because there is a probability that you would not have played the game, your call. Good to see people do some real anlysis. There is a lot of books and papers on pairs trading. They can be very usefull. Take my comments for what they are worth, I dont trade pairs that succesfully yet. But still, I understand the numbers.