Support Desk | ETFOptimize.com
Knowledgebase
ETFOptimize Support > Help Desk > Knowledgebase

Search help:


Do You Make Changes to Your Model's Algorithms Once Launched?

Solution

  

Q:  Do You Make Changes to Your Model's Algorithms Once Launched? 

A: Our Policy on Revisions:

We rarely revise the Buy/Sell Rules, proprietary Ranking Systems, or anything of critical consequence to a model's basic construction that can change its signals. A significant change to the strategy's construction could make the strategy's initial equity curve and risk statistics meaningless. Significant changes would mean that the model's history is no longer accurately representing its potential future.

Of course, the market can change significantly from year to year, and few quantitative strategies can work well in every condition.  For example, since 2009, the Federal Reserve has been the most significant market mover, and we have found it necessary to develop monitoring and signals around its activity.

However, we might occasionally make minor modifications to our Premium Strategies. For example, we have updated the average commission charged on ETF trades as brokers have steadily reduced them from competitive industry pressure. Commissions fell from $250 per trade in the 1960s to $0 in October 2019. That's when virtually every established broker dropped commissions for all stocks and ETFs. Broker's revenue now comes from selling the metadata from trades. While commissions were a small percentage of each transaction, with compounding, they added up over decades. Eliminated today, commissions are no longer a part of investors' considerations, and we have removed accounting for commissions in our Premium Strategy tracking.

Another example is in each strategy's ETF Universe. ETF issuers regularly introduce new funds while dropping others if there is poor demand. In the early years (1993-2010), we had to modify the custom ETF Universes for our strategies nearly every month as new ETFs were added and failures removed by their issuers.

However, as the industry has stabilized and matured, the ETFs that we use in our models have required fewer changes. That's because we make sure the positions recommended are well-established (more than two years since their introduction), have substantial also have ratings of 4 or 5 (out of 5) from FactSet for the ease of investing $1 million in a single trade. Screening for those characteristics ensures a stable Universe of ETFs for each model, but it is a living model – meaning that ETFs can be included or fall out of the Universe on the margins if their characteristics change substantially.

Also, our strategy designers keep abreast of innovations in systematic investment design and we regularly test new ideas for investment factors in our analysis platform.

When we develop a unique selection approach or innovative risk-identification technique, we either incorporate it into an existing strategy if it is more of a refinement – or we might launch a new product around that approach if it is a significant change. However, I should add that we aren't adding/or abandoning models constantly. A great deal of consideration goes into the introduction of a new model, and we might run it privately for several years before introducing it to the public. We currently have many prototype strategies running out of sample (OOS) behind the scenes to ensure their robustness, and we always have a batch of unique ideas to exploit market anomalies under development. 

After we have a core of about eight-ten excellent strategies on the site, we will likely make a few changes unless there is a unique new opportunity and a supply/demand disequilibrium we have not previously foreseen. If you have a recommendation for a model that will have broad appeal, please let us know with a support ticket at https://etfoptimize.com/support. 

 

 
Was this article helpful? yes / no
Related articles Are Commissions Included in Transactions?
How Long Has Your Firm Been in Business? What's Your Company History?
Article details
Article ID: 5
Category: Knowledgebase
Rating (Votes): Article rated 3.2/5.0 (16)

 
« Go back