Tuesday, July 1, 2008

When The CBI Spikes But The VXO Doesn't

The CBI hit 9 today and is reaching elite territory. The VXO is not near an extreme. The VXO (and/or VIX) are getting a lot of discussion lately as traders fret that a market bounce won’t happen until after a VIX spike does. It remains my contention that extreme sentiment measures like the VIX and Put/Call Ratios are “nice to haves”, not “need to haves”.

Tonight I decided to look at CBI readings of 8 or higher broken out by those times the VIX was stretched vs. those times it wasn’t. Rules for entry for the first test are as follows: 1) CBI closes greater than or equal to 8. 2) CBI closes higher than yesterday and 3) The VXO is greater than 10% above its 10-day moving average. If all three are met $100,000 worth of SPX is bought at the close. It is sold when the CBI returns to 3 or less. Results below:

Keeping the first two criteria the same and changing #3 to “VXO is LESS than 10% above its 10-day moving average” would provide the following results:

See the big difference? Neither do I. The results are almost identical.

But today the VXO wasn’t near 10% above its 10ma. In fact it was less than 5% above its 10-day MA. Below are the results when changing requirement #3 to “VXO is less than 5% above its 10-day moving average”.

It appears to me the odds favor a bouce soon with or without a higher VIX.


Anonymous said...

There is a very noticeable difference. When VXO is high, the losing trades lose more (and the winning trades win more). The ratio of avg win:avg loss goes from 1.64 to 9.85 to 12.92 while the percentage of wins remains almost constant. In other words, a lower VXO with a high CBI indicates that the market will either sit still or bounce, while a high VXO and high CBI indicates that the market is going to make a big move, probably up. This indicates that VXO is more useful for forecasting volatility than direction (as should be expected).


Rob Hanna said...

Thanks for the observation JKW.

With only 2-3 losers in each scenario and the max loss 1.7%, I wouldn't read too much into those specific trades.

Remember, the results shown are only showing the end result of the trades. They are not showing runup/drawdown for each, which would better help to measure volatility.

I beleive your idea that a higher VXO suggests higher volatility is sound. I don't think that could safely be extrapolated that from these tests, though.

Perhaps I'll show another way to demonstrate the validity of your hypothesis...


Anonymous said...

how do you adjust for data mining in all your studies?
correct me if i'm wrong but i think your trades based on breadth (sp?) were approaching the 50% randomness win (as expected if you overfit data). did you post any updates on how that is going?

Anonymous said...

i see we hit 10 on the cbi. Can you do a refresher for us. That seems to be the magic number right. There have been no losers with a 10 handle on the cbi ? Correct ?

Ps this is the best blog on the street !

Anonymous said...

I really appreciate sites like this with such thorough work done. Another sign of lack of capitulation not noted anywhere I've seen or heard is in a lack of spike in the NYSE trin (arms index), i.e. its 5 or 10 day m.a.'s. If you've addressed this and shown it not to be reliable, forgive me.