The In & Out Strategy - Continued from Quantopian

I think we should be careful here. I think the reason to why TLT isn't “working” is not because of high inflation per se, but the Fed's willingness to hike rates in light of high inflation. If the Fed took BOJ's approach, TLT would have gone the other way.

I've been reading the old Quantopian and now QuantConnect thread on/off for several weeks and finally decided to create an account.

With the recent talk about how to improve the Out strategy, I'm surprised there hasn't been a mention of the yield curve and using Treasury spreads for events like the past year where both stocks and bonds go down due to interest rates going up. Has anyone looked into that as a potential indicator for when to get out or to consider using for a cash option when it's not optimal to be in bonds?

Has anyone attempted backtesting this strategy in time periods before 2008? For example, the tech meltdown of 2001.

Any ideas on how to go about testing it? Almost all of these ETFs didn't exist then with the exception of QQQ. 

I've been experimenting with modifications to the strategy in times of rapidly rising rates. If anyone is interested in collaborating, send me a message on Telegram: @janthonym

Hi Jam, Vladimir had a limited run of this strategy dating back to 2003 in one of these threads, unfortunatley I don't have it handy at the moment, but you can find it on the forums. It performed well from 2003 to 2008 in the backtest Vladimir shared. 

Backtests from 2000 through to now for the various In-Out versions run fine if you're not using some of the more recent bond ETFs as alternative assets. From memory TLT and TIP have data back before 2007 but a lot of the other “safe” assets don't, which is why most people are restricting back test runs to 2008 onwards. 

Has anyone been able to implement this algorithm in C#?

I have been playing around with this model diving more into optimizing hedging tools and inflation plays. Using Strongs recommendation, I found an effective way to better enter/exit positions in 2022/23. Backtesting produced statistically sound results using bootstrapping test and T-Test with p < 0.01 and < 0.03 respectively. Is this the best way to test this?

Heres the performance on my local system (currently transferring to QC)

More importantly, I really enjoyed the idea of inflation “anomaly” detection. I ended up with a rather unique strat and would love thoughts because I am not sure why it works. Essentially, the allocation to TLT and UUP is decided by the covariance between the assets. I calculate this for the time range and then overlay a 300-500 day mean covariance (essentially the total mean). If the covariance is < the mean then allocate to TLT else allocate to UUP.  

The results are a lower correlation to market ETFs and lower correlation to TLT and UUP. Given the returns results in the new hedge allocation “Selected Returns”. It has less volatility and greater returns.