Hi Chris — thanks very much for reaching out. We have been using Algo Catalog as a source of fully specified mechanical rules, then independently testing some of the more interesting ones on our own point-in-time Nasdaq-100 and S&P 500 data.

A few things would be especially useful to know if you have inclination and time to share:

For Nasdaq-100 and S&P 500 backtests, do you use the actual historical point-in-time constituents at each date, or current constituents carried backward? If point-in-time, what is the membership-data source?

What is the exact execution convention in the backtests — when is the signal calculated and when is the trade assumed executed? E.g. I have found that trading on the 2nd-to-last day of the month based on the closing data the 3rd-to-last is better than trading the first of the monh or the last day of the month.

Would you consider making the strategy definitions downloadable in a machine-readable table — universe, filters/ranks in sequence, Top-N, rebalance frequency, timing rule, costs, etc.? That would be extremely useful for people independently reproducing and comparing the algorithms.

Last I looked some strategies said they were public but when you open them they indicated they would be someday. Is there a likely date when those will be?

Similarly, would you consider providing downloadable monthly return histories? We have found that this is much more useful than CAGR/MDD alone when testing whether several good algorithms actually diversify one another.

Do you test neighboring parameter values to see whether a result is robust — e.g. Top20 vs Top25 vs Top30 — or primarily the published exact specification?

Do you keep records of strategies/factors you tested and rejected? Even broad information on approaches that repeatedly failed would save independent researchers quite a lot of duplicated work.

Have you noticed that criteria that work well within the Nasdaq-100 often fail when transferred unchanged to the S&P 500? If so, are there particular factors or two-stage combinations you have found transfer especially well between those universes?

One feature request would be the ability to filter/search the catalog simultaneously by universe, rebalance frequency, number of holdings, CAGR, MDD, Sharpe, and factor family. That would make it much easier to discover algorithms with a particular return/risk profile.

Thanks again — the site has been very useful to us precisely because you publish enough of the actual mechanical rules for them to be independently tested. I realize you have proprietary data to protect, but you asked for questions, so here they are!