2023-01-06 Risk Model update & correction:
1/9 Model posted positive results beat SP500 and BM by 97 & 12bps.
Bullish EM, EAFE, & SP500 helped. While shorting US High Yield & R2K hurt.
NatGas was the worst performer & EM the best.
Next week is long all except BTC & R2K
Correction to update last week.
2/9 On Thursday the model indicated to go short several Equity markets on Friday. However, after posting this in my last tweet (now deleted) this was a model error due to a miscalculation of implied volatility.
3/9 The risk model calculates implied volatility (IV), across time, of options on ETFs to gauge risk. Generally, when there is a liquid options market, I can calculate the IV at any point by interpolating the nearest options chains to that time.
4/9 Generally it is efficient to use the options chain immediately before and after the desired IV date.
For example, 30 days from today is 2/6/2023. So to find the 30 day IV for SPY, you would make a calculation based on the 2/3/2023 and 2/10/2023 options chains.
5/9 Generally it is efficient to use the options chain immediately before and after the desired IV date.
6/9 For example, 30 days from today is 2/6/2023. So to find the 30 day IV for SPY, you would make a calculation based on the 2/3/2023 and 2/10/2023 options chains.
7/9 Since this method is less efficient, I will use a volume weighted average of several timeframes to infer an appropriate IV measure; so instead of the usual (+1/-1), average(+2/-1, +2/-2, +1/-2) etc.
And that's where the model error gave a false signal.
8/9 Instead of isolating the low volume ETFs at issue, I ran all of the ETFs using the (+1/-2) options chain, without averaging.
9/9 This led to false signals, indicating false shorts in several markets. I have corrected this mistake and there were no changes to the risk model on Friday.
Long way of saying that mistakes happen, this is just a paper model, and this is also not investment advice.
Cheers ADThis post is based on this twitter thread.