much easier on a correlated pair...its been years but the site we used to find individual equities was impactopia dot com. I wonder if that is still available. It might make for a good illustration to find a more typical iteration. Fannie/Freddie was one I traded. The banking industry, the home builders and the oils service providers where the meat back in the day....lol bet GS/LEH doesn't trade anymore.
I have most of the code we would need to plot a candelstick chart of the ratio written in other indicators. Give me a day or two and I can put something together. The question would be what ratio to plot?
The following 2 users say Thank You to Xav1029 for this post:
the delta between two symbols would be a good place to start. I'm thinking about if the delta needs to be ratioed at this point..or ever. The idea is to keep the long and the short dollar neutral. Is dollar neutral important?
One thing we might look at, or defer to, is the respective typical daily ranges of the products. In a highly correlated situation you do theoretically have greatly reduced risk. To me that means I'd be more comfortable in the longer time frames. As crazy as it sounds hourly...or longer...lol
So using this delta concept, narrow, buy one sell the other, wide, sell one buy the other. I guess looking at the individual charts you could easy enough do the same thing....so is this consolidation a step toward simplifying or making the simple complicated.
For me, I think I'd like the view alongside what I already do.
On another platform you could do the calc in the symbol field.... this minus the quantity this times this. I don't know if it is that easy or even necessary.
I think the real trick would be to create a basket risk on/risk off trade that you could fire with a low freq algo. I mean if you where using 60 min or 240 min charts you might trade a couple three times a week..AND if you have that theoretical hedge are you in a no worries situation?
Hey lol sorry I was thinking of what to chip in but had to make a few phone calls all of a sudden and forgot to come back here. @Rikers11 may be a better person to go to for FX from what I see. My group does a lot of stat arb but not of this variety. Before he chips in though, I am guessing that both of us share some skepticism about this strategy, because the spreads in spot FX and futures have been flattened (they had computers to trade this as early as the 1980s). See under "Explain some abandoned / public trading algorithm and how it works": "I was a quant hedge fund trader for 11 years. Ask me anything."
1. You are right to look at longer time scales for your signal. Just running a quick check, the standard deviation of the daily returns on the EURUSD is about $0.0350, so that approximates to $0.00350/sqrt(60/3*24) = $0.0016 on the 3 minute time scale that you were looking at. (By the way, it's more accurate to do this calculation by taking the standard deviation of minute returns and scaling it.) If you aim for 1 standard deviation, achieve a win rate of 60%, then the expected unleveraged return is $0.00096 per lot. What are your commissions, leverage and the quality of fills from your ECN? After transaction costs, the strategy may just incur consistent losses. Pick a time scale where the spreads are wide enough.
2. The hedge is of course not a safeguard, and you always have to be on a lookout for policy and settlement reasons that a spread will suddenly not mean-revert. One example I've heard of this year is AAPL-SBUX, which would have made awesome hedges for one another for the whole year up till August, but blew up spectacularly. The spread makes no sense, but probably many mutual and hedge funds were switching between technology and consumer non-cyclicals using Apple and Starbucks as proxies until the end of summer. Another classic example is the CHF, where many stat arb models in FX died when they pegged the CHF.
The following 3 users say Thank You to artemiso for this post:
Please register on futures.io to view futures trading content such as post attachment(s), image(s), and screenshot(s).
Also, what you're talking about seems to be more of pairs trade than statistical arbitrage, as you are essentially just building a synthetic eur/cad pair. You might look to find arbitrage opportunities between a synthetic product less the cost to carry, and the spot product. But that doesn't mean these pairs of pairs don't offer interesting directional trading opportunities, you may find limiting your exposure to the dollar may increase the "trendiness", making it slightly easier to predict.
"If I agreed with you, we'd both be wrong."
Last edited by addchild; November 14th, 2012 at 12:14 AM.
Sry, I'm not sure what's the goal. If you're doing pairs trading the best place to look isn't FX. As mentioned you'll be just trading a synthetic, if you want to get of the dollar just trade the non dollar cross. Triangular arb is not possible, market makers upload their quotes automatically for all crosses and the leftover is arb instantly (someone leaving an order sitting without watching the majors for example). Also FX is highly news driven without the scheduled news (yesterday was a good example) so if you are doing any kind of short term arb you get blown of if you can't shut down properly (as to say the "tails are fat"). You might think that yesterday was a great example of RTM but it's only because on the aftermath the news was denied by the Germans....
for software you can experiment with ArbMaker. (I have no affiliation and don't use the software).
Last edited by Rikers11; November 14th, 2012 at 02:09 AM.
The following 2 users say Thank You to Rikers11 for this post: