Methodology: how are strategies tested and deployed

[Peticolas]

When I look at a strategy like the NASDAQ 100, I understand the high returns but the risk metrics look incredible. Can you describe your process for developing these strategies? Are they developed at one point in time, and released, or are they reformulated each year to best fit the data? (e.g. when was the last time the parameters of the NASDAQ 100 were reformulated?)

There may be information on this in your software, but I don’t have access to that, as I’ve noted previously.

[Trevor]

I can’t speak to the details of the NASDAQ 100 strategy, but just in general, LI strategies do change occasionally, but not often. Definitely not once a year. The only recent change I can think of was the addition of TIPs to the hedge sub-strategy. Technically that affected most (all?) of the strategies, but was very targeted. The only other changes I can remember were the addition of the hedge sub-strategy to the bond strategy, which was getting clobbered, and I think a change to the constituents of one or all of the core portfolios. The sense I get is not one of “let’s update this strategy so we can make the backtested results look really good”, but more of “something fundamental changed in the market that partially invalidated the assumptions of the strategy, and we need to fix that”.

[Peticolas]

Thanks Trevor. If I had been able to get the software to work, I probably would have been on-board, but the installation file throws errors that no one seems interested in helping me with.

[Trevor]

Oh, too bad. I’ve actually never used Quant Trader (not a Windows guy) or I’d try to help you. FWIW I do just fine without it, but I can understand if you like to tinker with strategies that Quant Trader might be a big selling point.

[Vangelis]

Thank you both for your feedback.

Robert, apologies for not helping more with QT, usually installing QuantTrader just works. It’s true that the best way to understand how our strategies are constructed is to run and play around with parameters in QuantTrader.

I created a simple video of downloading and running QT. Maybe it will help pinpoint the problem.
https://youtu.be/EJMB_66N6ww

[Peticolas]

Thanks Vangelis. Your video shows what I expected, but when I enter my username and password, the application threw two errors (screenshots of which I have reported in emails), and then exits the application. It does not appear to be an issue with credentials. We did solve that problem. Tried the normal thing, rebooted, but it still throws the two errors.

[Peticolas]

There’s an article in which Vangelis describes LI testing as walk-forward analysis. My question is whether LI uses anchored walk-forward analysis (always starting at the same data point) or rolling walk forward analysis (discarding old data).

[Frank1 Grossmann]

We can only do a rolling walk forward analysis, which means that we run our optimizer automatically for a fixed period in the past. This will give us a new optimized look-back period which we apply to the next month to calculate the allocation. One month later we repeat the same thing. So the three-year period will always move by one month. We have seen however, that the optimum look-back period is more or less always between 45 and 100 days and this type of adaptive look-back period does not really improve the strategy results.

[Peticolas]

Thank you, Frank.

[Author]

Posts