Volatility mastery

C2
1

Guys and girls,

I have been trading quantitatively since 2003, on a daily basis. I have developed a combination of systems to maximise returns from volatility.

I have combined these into a system which was up 22% last year, is up 13% this year so far, and has had a max drawdown of only 12%. It is not very correlated to the S&P500. The Sharpe ratio is 2.7.
(those results are real for me but “hypothetical” in that your mileage may vary - standard disclosure).

You may not have noticed it as it shows in the Grid there are only 2 trades; it has had a lot more than that, but Collective only counts full exits of positions. Most of the trades increase and decrease the overall position size (so therefore don’t show in the stats).

Market orders are used - so it is easy to trade. It is only $49 per month. Take a look! If you have any questions, please let me know.

PS: Note that the chart including Interactive Brokers’ commissions also includes IB’s $99 autotrade fee (Collective2 has to include this in the charts). If you trade manually as I do the commissions are extremely low - I pay about $1 per trade (depending on size) - and there are not many trades.

Thanks
Rod

2

would you say that the 0.588 correlation to the SnP500 (based on C2’s calculation) is incorrect? or do you mean that your strategy’s correlation is not close to 1.00? Correlation is a measure of linearity…

3

@Roujee, the 0.588 correlation to the S&P500 is correct. That is, it is not very correlated as <<1. Correlation is not a measure of linearity - it is a measure of how closely two time series move together.

If one has a system very correlated with the S&P, this will provide some diversification.

4

A correlation of .588 is a very high correlation. That is much higher than the correlation between being a Republican & being a conservative, or the correlation between years of education and income.

The correlation coefficient being computed here is almost certainly a Pearson correlation, that is, a linear measure of how closely two variables move together. It would be possible to get an even higher correlation if one of the variables were transformed to reflect a curvilinear relationship.

BTW, my PhD is in quantitative methods.

1 Like
5

There is no bright line rule for what is highly correlated. For models here, there would be some benefit in having a negative correlation between models because that would smooth returns more than having a correlated and an uncorrelated strategy. IMO, I would consider a strategy to be uncorrelated with the SP500 if the correlation coefficient was lower than .20 or .15. But that’s just my opinon.

1 Like
6

@RodneyBryant, what I mean by linearity can be illustrated in the following example:

S1 = 2, 3, 4, 5, 6, …
S2 = 4, 6, 8, 10, 12,…

Correlation between S1 and S2 = 1

S1 = 2, 3, 4, 5, 6, …
S2 = 4, 9, 16, 25, 36,…

Correlation here is less than 1 (0.989).

So even though every time S1 went up S2 went up (and will go up, given S2= S1 * S1), it does not mean correlation is 1.

Now, if someone is simply short the VIX, this strategy will have a high correlation with the SnP500… not 1.00, but maybe 0.60.

So when you say your strategy is not highly correlated with the SnP, that’s not exactly right.

@QuantitativeModels, I would rather see the correlation of log returns between the SnP500 vs. the strategy.

7

Hello Rodney, The stats look good, and congrats on the Vol strategy. Im a huge fan of Vol strategy myself. So lets get down to business eh. U said " I have developed a combination of systems to maximise returns from volatility". This is a good way to go, so in this portfolio, there are another strategy (But i see 1 trades in trade history tab, thus need to sort this issue out). If im a investor of your strategy, i want to know how many percent of your portfolio that get allocate to Vol strategy.
This strategy are trading XIV which is inverse, should have somewhat really high Beta . So with high beta i would like to know how much alpha got attracted . i thought Collective 2 have this statistic ( Please Collective 2 we need this stat in order to gauge the performance of the system).
If Im an investor im seeing some good things and not good at the current state of the system.

  • Returns look nice and realistic . i say conservative Vol strategy Draw down right now is 12 % for Vol strategy (tricky one).
  • High Beta (.57) but if it have a good alpha then that Beta level is acceptable. (again Collective 2 gives Beta with no Alpha). Well i guess we can use Sharp, Carma but if im a investor i would like to know how much Alpha is being generated by the system along with Sharpe and other fancy stats.
  • @System developer, If the system have almost 9 months record how come C2 have just only 2 trades ? Is all the return for those months came from 2 trades, i usually download the trade history excel file to analyze in python or Stata , but for this system i downloaded the EXCEL file and it showed 1 trade. Therefore, what is going on here ? all those statistics, that shown in front page, are based on 1 or 2 trades ??.

Again no trade history exist… need to sort out

8

Hi @QuantitativeModels. I agree with what you say. Over the last 10 years of backtesting the system had a 0.22 correlation to the S&P500.
Regards
Rod

9

Hi @Actuary4life
Thanks for your message. Regarding the number of trades, I mentioned that in my opening post “You may not have noticed it as it shows in the Grid there are only 2 trades; it has had a lot more than that, but Collective only counts full exits of positions. Most of the trades increase and decrease the overall position size (so therefore don’t show in the stats).”

There were actually 15 trades over the time period. Collective2 only measure FULL exits of positions as being trades. For example, if I buy 100 shares of something, and sell 30, it does not count that as a trade. This system does that a lot.

I have asked Collective2 to fix this part of the platform but it isn’t on their priority list.

Regarding a “combination of systems” - they are all combined in this one system on Collective2. ie. I run the systems daily at my end, and put the orders in Volatility Leader on Collective2 and in my own trading account.

Regarding alpha, that is just outperformance versus a relevant index. You can see that on the chart (QuantitativeModels will probably correct me but that’s the summary of it).

Kind regards
Rod

10

Hi @Roujee. You write " if someone is simply short the VIX, this strategy will have a high correlation with the SnP500… not 1.00, but maybe 0.60." I am not simply short the VIX. I am short the VIX using XIV (long) at various times (not all the time), and at certain times I am long VXX (but not during the last 9 months).

Thanks and regards
Rod

11

@RodneyBryant
I did not say you were short the VIX. I was simply illustrating that someone with a short VIX strategy could have a correlation of 0.60. I’m trying to say that a correlation of .588 is still high.

Regards…

12

Thanks. That’s helpful. So backtesting, you have a low correlation (.22) with the SP500.

1 Like
13

Thanks for answering, Im trading the VIX quantitatively myself, developing and testing many different models to exploit cantango and backwardation. Now that is out of the way.
You said " Over the last 10 years of backtesting the system had a 0.22 correlation to the S&P500.". Now question arise,
where do you get 10 years of XIV data (XIV started around 2011 i think), so did you build you data using VIX back to 2004 ( I did this and it could be done) but problem arise by doing that too. Thus it is hard for me to have a picture of Beta .22 by just trading the XIV alone (Now if you want to do something for fund. DO THIS If you run Regression on the log returns between XIV and SP500 since 2010 using YAHOO data, you will see that the beta > 2.1 and alpha is about .005) . Now that is some serious Beta, (more than 3x leverage ETFs like UPRO, SPXL. (Im using STATA sofware to run this FYI). Im just trying to UNDERSTAND your system before the real money get allocate. Therefore, please no hard feeling ( I dont have any other motives here).
Regarding Alpha, I know i can look at the chart that your system is beating the SP benchmark. But you and me know if we are trading quantitatively those chart does not mean a dang thing.
You see, i like the system, wish there more history so maybe i use CAPM model to attract the Beta and Alpha stats so i have a better picture of what im looking at.
So in the end, im watching it closely.
Lets get some more history eh.

1 Like
14

XIV is priced according to a formula based on futures prices. For the period before 2011, you can buy data going back to 2004 reporting the estimated value of XIV or VXX or whatever, computed by using the relevant formula. Then use the estimated values before 2011 and the real market data since 2011.

Vance Harwood at Six Figure Investing developed the database and sells it:

https://sixfigureinvesting.com/2013/09/backtests-for-volatility-etn-etf/

If you are developing a volatility strategy, you need to check it through the downdrafts in 2008, 2009, and 2011. Most of the current high-flyers on Collective2 have not yet been through a big challenge; some will make it through OK; some won’t.

1 Like
15

Yes, i agreed with you, that testing again 2008 will be important. But most won’t
I did not know they sell that. I used to follow this website https://quantstrattrader.wordpress.com/2015/10/21/how-well-can-you-scale-your-strategy/
He did build the data somewhere, now i can not find the article.

XIV return are good but if look closely you will see this High Beta + Bull Market = Impressive returns.
So question is how long until the big drawdown shows up ?.

16

@QuantitativeModels, great answer - that is exactly what I do. I use Harwood’s data. Thanks.

17

I am long VXX in downturns.

1 Like
18

Yeah, i think VXX will get split again soon.

19

So you are “long VXX in downturns,” but not now? Is that correct?

Being long VXX right now would be VERY risky. With a wasting asset, you must be right not only on direction but on timing. Backtesting volatility, I see some periods where my models show that it would be profitable to be long VXX/UVXY, but I am skeptical that going long VXX will work consistently enough out of backtesting to be worth the risk. I’ll probably nibble a little bit at the edges occasionally, but I doubt I’ll take any big positions going long VXX/UVXY.

1 Like
20

Hi @QuantitativeModels. I should have been more specific: I am long VXX in significant market downturns. I’m definitely not long now. The triggers are all quantitative, not discretionary.

Thanks and regards
Rod