What correlation?

Trading Discussion
1

Greetings. I have been trying to improve my use of correlation values to diversify risk when trading multiple pairs concurrently. Typically, given a signal for a prospect trade that includes values for the correlation of the prospect pair with currently open or pending orders. If x = correlation value, then my target would be to keep all pairs in the range -0.35 < x < 0.35.

I have decided that I might be doing this all wrong for the following reasons:

  • By not considering likely future correlation between pairs for the duration of the trade(s)
  • By not looking at the relevant property for effective diversification

The first point, predicting if the pairs will continue to correlate in the target range for the time horizon of the trade(s).

Secondly, correlation values are usually calculated using the prices of each instrument in sequence in a time interval rather than the change in prices of each instrument in suquential time. So which one is better?

I have written a quick indicator to compare the two and have the following results from the graphs:

  • Price based correlations tend to swing more widely between -1 and 1. If you compare USDCHF & EURUSD which are highly negatively correlated, most of the time the price based correlation value will sit below -0.65 whilst at other times it could be as high as +.65.

  • Conversely relative price change based correlation is very steady, the USDCHF vs EURUSD example is almost always below -0.7.

I tried a simplified comparision in excel looking at 2 sequences of numbers , one sequence starting at 1 and the other starting at a 100. In both cases I increased the prices by 1 over the next 49 cells to give the sequence:

  • (1) 1,2,3,4,5,6,7…48 and
  • (2) 100,101,102…,.147

In case 1 (price based) the correlation value was 1 and in case 2 (change of price based) the correlation value was 0.45.

In sequence one the price increased 48 fold in 49 price changes from 1 to 48. In sequence two the price increased by 47% in the same time.

Case two appears to be better on the face of it, but if you are entering trades in two pairs with a fixed TP & SL of the same pip values for both pairs then the absolute price move is more important (so case 1 wins).

If you are comparing pairs like USDJPY (109.6…) to EURUSD (1.171…) then case 2 might be the better choice. Not sure.

2

I think I might be leaning on relative change in price because it compares the change in price as the central property and not as a derivative property of the absolute prices. Also it is a weighted measure so inherently tells you something about the comparative volatility in the price changes: 1 to 48 vs 100 to 147.

Also, you often read about the reliability of indices as a tool for the comparison of markets based on whether they are a market cap weighted index like the FTSE100 or an unweighted index like the DOW 30, so a similar principle may apply.

3

The two values you are are describing is correlation and beta. I track correlation of my positions and their beta against each other. I also track my portfolio correlation and beta weighted deltas. If my correlation or beta weighted deltas get too high i reduce. Just because you are trading different pairs or products, doesnt mean you are diversified, which is where correlation and beta come in. Compared to other traders, i have a very low draw down, which is largely due to having true diversification.

When you start trading larger sums, say a retirement account, this becomes that much more important for capital preservation purposes.

1 Like
4

Makes sense. Still a risky enterprise though. I will be launching a thread soon about why you should give all your money to ropunzel instead. :grinning:

I think you are right it is a derivative of the beta for comparing security returns, the only difference is that the denominator comprises of the standard deviation of both pairs multiplied together rather than just the benchmark security variance.

How do you find them both (correlation vs Beta) as a diversification tool i.e. which is better from your experience.

1 Like
5

Sure, anything that offers a decent return will have risk involved, but diversification of correlation and strategy allows you to achieve similar returns while taking on less risk.

I track both correlation and beta and as you mentioned are quite different calculations. I use an online portfolio tool where you can feed it all of the tickers in your portfolio and it will produce a correlation matrix. I then put a heat map on it to see if I am too highly correlation in anything. I also take the matrix and calculate my overall portfolio correlation which I like to keep below 0.35. My brokers platform calculates my beta weighted delta’s (weighted against the SP500) so I can see if I have too much market risk.

1 Like
6

@krugman25

When comparing the correlations of two pairs, I have decided to split the price movement into two parts:

  • Direction &
  • Magnitude

Direction: If the price closes up for one pair then its direction is 1 and if it closes down the direction is -1. If I use USDCHF & EURUSD as an example. The direction sequence for the last 50 bars on EURUSD might look like {1,-1,1,-1,1…} and similarly USDCHF might be a sequence of {1,1,1,-1…} etc…

So I will check the correlation of those two sequences to find the relationship in the price movement between those two pairs.

Magnitude: I will use the percentage change in price to get another two sequences that show the magnitude of the price changes, which I can also check the correlations for. The magnitude is in absolute terms so ignores the direction of the change in price which should be captured instead by the previous.

here the magnitude is M = (Close(1) - Close(2) )/ Close(2) then M = sqrt(M x M)

This is what the graphs show for USDCHF & EURUSD.

EURUSDH1 (1).png1358×435 18.3 KB

The red line is the Direction correlation and the blue line is the Magnitude correlation.

This shows that for USDCHF & EURUSD the price changes are very negatively correlated so tend to move in opposite directions, but the magnitude of those opposite moves are very positively correlated. The prices move in opposite directions and to the same degree.

This is the graph for the traditional correlation based on the prices

EURUSDH1 (1).png1358×435 17.5 KB

I am going to stick with this because it gives me more information about what is happening than the traditional correlation. One thing that I have noticed is that most pairs have a positive correlation in the magnitude of price moves. I think this positive correlation reflects the volatility in the market as a whole.

I think the main thing to consider in that case would be the correlation in the direction of the price movement. For that I will use the following to qualify trade diversity:

  • Probability that directional correlation is in range -0.35 < x < 0.35 based on last 1 year data
  • Average reversion time in bars back to acceptable range after a breakout.

This will let me see the average correlation as a probability value, quantify the risk of a break out from the acceptable range and indicate how long I should expect to wait on average for a correlation breakout to correct.

One years back data should be enough. I expect the interdependence of pairs will continually change with time, or not. Whatever the case laziness says do 1 year.

2 Likes
7

Good on ya, nice work.

You seem to have recognized the pheomena where cross market correlation begins to increase with volatility.

8

I think that you trade US Equities? In that market the VIX volatility index captures the volatility implied by the option prices, I don’t think there is a similar product for forex. But due to their interdependence it would make sense for the volatility across the market to positively correlate to some degree in most cases.

9

@krugman25 Can you think of any way one can test the new approach to see if it improves the diversification of a portfolio. You mentioned that you look at overall portfolio correlation as part of your process, may be you could apply the old system with the old rules to select tradable pairs and also the new system, then compare the portfolio correlation of both at the end. Compare using old system.

10

Yes. I trade equities, but also trade pretty much everything. Volatility and expected moves are calculated from options prices. Believe it or not there are options on Forex. It is called options on futures and there are futures for all of the majors. Because of that there is an IV and IVR for the pairs, not to mention the ability to trade FX using options strategies.

11

That’s cool, I actually trade spot and spot options mostly at the moment. I prefer to write than buy and trade a simple hybrid of a covered call or covered put in the opposite direction to where I expect the market to break out. I prefer it to trading the break out psychologically; it is compartmentalised and the returns are known in advance. The strategy also includes a spot trade in the direction of the breakout which without going into detail should ensure a break even or better with probability 67% on each trade. And most importantly there is edge, in no small part due to the option premium tilting the balance.

I am familiar with implied volatility (IV) but not with an aggregated IV for the forex market as a whole like in the case of VIX for equities. Is there a product like VIX for FX out there? it would make more sense for there to be one in Options of Futures which is exchange traded rather than OTC Options on Spot.

Although I am not familiar with futures and the brokers that I spoke to to potentially trade options on the CME exchange were prohibitive with 1) leverage 2) nature of referral 3) by providing options on futures only.

12

I also only write and never buy options. The only reason to buy an option would be if it was part of a larger strategy that involved selling options like a calendar spread. All of my options trades generate negative theta and if I want some directional profit I create a certain amount of positive or negative delta. I trade as you mentioned which is contrarian plays, especially large selloffs where I would sell puts for theta and delta or sell strangles for just theta.

I actually trade using a pretty standard price action strategy that many people in these forums and in Forex in general talk about, except I trade it using options which makes the success rate much higher. For example if there is a pin bar reversal, the standard price action rules state entering on the break of the pin bar and place your stop at the other end. Too often price will break and then pull all the way back and hit the stop, and then immediately turn around and zoom off in the original direction. Most traders get chewed up by that type of market action and throw in the towel on price action trading. If I trade with options I can trade the same way except I don’t have hard stops, and if price moves up and then turns back around I am just sitting on my option collecting premium and waiting for it all to work out. In addition if price moves against me in a major way I can convert my naked put/call into a straddle and begin collecting double the theta.

I have never had problems making money trading straight price action in Forex, but my success rate has gone even higher incorporating options and there are a lot more options( no pun intended) when markets start moving against the trade, and I never really have to take a loss. I can keep rolling my options contracts forward into perpetuity until my trade becomes profitable.

I am not sure what problems you were having. Options on futures leverage should be even smaller than trading the futures themselves since an options on futures contract represents a single contract (compared to options on equities which represents 100 shares), naturally it will then require equal or less leverage than trading the future itself. And futures already require very low capital. Most futures are 20 - 100 to 1 in their leverage. I recently switched to TastyWorks and they will be rolling out options on futures in a few short weeks. They also allow naked options trading on retirement accounts which I think will make them on of the first to allow it.

13

That’s very interesting. I see you actually trade option value movements then since you look at the theta or decay in value with time. I tend to trade them with a view to rig some edge into my spot trades by taking the premium, so theta is not that important for me because I hold all options to expiry for cash settlement.

They are however hedged with a spot pending order in case the market moves against me. Rolling expiry? is that american style options? I’m not sure I can do that, I trade European style only. Since I have been trading options I can say I pretty much never lose and there is piece of mind.

14

The difference between European and American is when the options can be exercised, but that doesnt have any effect on the things I am talking about. When i talk about rolling, I am talking about rolling your positions to the next month cycle. I personally sell options with 45-60 DTE and buy them bqcm with 10-14 DTE. Thats because options become the most risking during the last 1-2 weeks before expiry. In those last 2 weeks gamma risk exponentially increases. On the other side you also dont want to go too far out in DTE because theta is too low. So the sweet spot is holding an option in the 2 to 6 weeks range because theta is good and gamma risk is lower. Since I dont hold to expiry, my theta numbers are telling me how much premium I am collecting each day assuming price doesnt change, because I will eventually have to buy those contracts back when I roll my positions forward.

I will also close my positions before that 2 week period if I have collected at least 50% of the total premium. If you think about it in terms of risk/reward, as your options lose value the trade becomes more risky. Think about it this way, if your option has lost 90% of its value you then have 9 times risk for 1 times profit. The most you can make is that last 10% and could give back 90% of your profits, or more to get that last 10%. Thats not just my opinion, there has been a lot of quant studies showing closing at 50% profit or 2 weeks to DTE produces the best profit. It is also about capital allocation. Why keep your trading capital tied up in an option that only has 10% left of profits when you can close and move onto a new trade where you start back at 100% profit potential. Many times I squeeze 50% of profit out of a 6 week contract in just a week or two. Rather than waiting another 5 weeks to get that last 50% I can move onto a new trade and start over. In this way I am boosting my daily P&L as well.

1 Like
15

That’s very interesting, so you trade options properly then and without a spot component i.e. looking for opportunities where the premiums are potentially overpriced and then riding the decay until the risk reward utility starts to decline. Do you use exchange traded products.

I trade options on a weekly cycle. I don’t like long term trading generally on any product so far. I think I might move more long term when my AUM reaches target figures, but then I will probably go on to indices.

Why did you decide to settle on options, or how active are you on the other markets and products that you trade.

1 Like
16

No, at heart I am a price action candlestick and contrarian trader. I still trade price action on the spot market using PA. Its unfortunate that most traders fail at PA because I have been very successful at it. For example i just sold a put when gold dropped to 1292, becuase i expect price to pop back up. When price popped up my short put dropped im value by 60%. I Basically squeezed 60% of an 60 DTE option in 1 day. But the theta was beneficial as well because if price didnt pop up and just drifted sideways I would have still collected the daily theta decay. So I use options for both directional from the spot and theta decay of the option itsrlf which improves my cost basis and give my trade more profit edge.

I have traded options longer than Forex, but I went all in on options because I realized they are by far the most profitable and highest probability way to trade. My P&L swings are much smoother as well which is important since I trade my wife and I’s nest egg, not just a fun money account.

I trade pretty much all markets. I rarely trade stocks directly because they are not capital efficient, equities and futures give traders access to just about every market and both have options trading vehicles.

17

I use SaxoTrader for options trading who do you use. I haven’t really looked at trading the decay in option prices because their spreads seem very high (Saxo). So I am looking to see if the option will expire out-of-money.

I make the directional bet based on the anticipation of a break out in the spot. That is problematic because breakouts are usually from low volatility period into higher ones (caused by the break out) when you should be selling options in the other way i.e. from high vol to low vol when the premiums are high to low.

18

I use SaxoTrader for options trading who do you use. I haven’t really looked at trading the decay in option prices because their spreads seem very high (Saxo). So I am looking to see if the option will expire out-of-money.

I use TastyWorks and would highly suggest if you want to improve your options trading strategy start follow the crew at TastyTrade. They have 10 data scientists on staff and are constantly running quant analysis and various strategies to see what produces the best results. I only trade highly liquid options markets. I am looking for something where spread is about 1 penny per 50$ of options profit. So if I have a 250$ option (as in 250$ of extrinsic value) I will tolerate about a nickle in spread. That’s my maximum tolerance, I still try to get better than that. Also dont ever pay at market. I have had cases where the market was 10 cents wide and I sold the options 2 cents from bid. That adds up over time.

I make the directional bet based on the anticipation of a break out in the spot. That is problematic because breakouts are usually from low volatility period into higher ones (caused by the break out) when you should be selling options in the other way i.e. from high vol to low vol when the premiums are high to low.

That’s why you should only use options for directional trades when going long on down moves, that way you are selling puts into high volatility so you can profit from both the volatility crush and directional move up. If you do the opposite and sell a call for a down move, even if price falls and you make money on the direction you could get some big IV expansion, which hurts the trade. In which case it may be better to just short the underlying for a downward bet and to use put options for a long bet after a large selloff.

19

not sure about this but looking at the site.

20

Hey. Back to the correlation stuff. I have decided that stripping the pairs down to direction and magnitude is unnecessary. If I wanted to test for which correlating property is most appropriate between absolute price and the others considered then I would look at the predictions for both against what actually happened. The what actually happened bit would be the correlation of the price changes after the prediction. So that is probably the answer.

I will look at the percentage change in price (weighted) as the property to check the correlations for. Also, if I want correlations to be within a defined range, there is no need to check the average time to revert back to that range after a breakout, at least for my purposes. The probability of being within that range over the last x number of bars is enough to get an idea of the movement in the correlation.

I am placing trades for 1 week on the H1 candle as the primary time frame. 1 week = 120 H1 Bars. But I m looking back 3 months (1440 bars) to find the probability within range. Therefore I just need to make sure that a probability within range in last 3 months (1440 bars) is a good predictor of future correlation for 120 bars. Everything else is probably an over complication.