DivPS is a per share function , therefore only looks at the common shares and their dividends. In the case of multiple common shares we use the one that Compustat flags a “primary”. DivPS does not look at financial statements. Please see the docs “This data comes from the corporate actions stream.”
PayRatio5YAvg uses the annual financial statements total dividends line item. There’s no fallback and must have 5 valid values. We’ll make the documentation clearer.
I discussed this some more internally, and I think we will need to change PayoutRatio5YAvg once more so that it handles missing data during preliminary reporting. Please see image below. It’s for AMGN and you can see from “CompleteStmt” the preliminary data lasted about 1 month. For a whole month PayRatio5YAvg went to N/A (depicted as a straight line in our fundamental charts). A better value would have been the previous one, around 21.5, until the final data came in to make it 29.9.
We’ll announce this as an upcoming change.
NOTE: PayRatio is particularly affected by prelim data since it’s using a cash flow item. Other 5Yavg ratios might be ok. We’re investigating them now.
Thanks Marco, really appreciate you guys looking into this further.
All:
It should be clear from this discussion (and countless others) that there is no way any of us can write a system and actually trade that system. The best we can do is trade whatever approximation Marco allows. Shame.
Bill
Bill, please explain. Thanks
Bill,
I feel the need to defend the product somewhat. As a technical software developer in a previous life, I perhaps have a unique perspective on the difficulties being handled in the background. Although I have gotten quite frustrated over the last couple weeks about a few things, what Marco and team have put together is really an amazing product. At this price point there is nothing I can find that competes. Yes it is true that there are systems through which you could have more direct control, but they would require dealing with directly with the data quality headaches that Marco and others handle for us in the background and the majority of the users don’t have the time or the horsepower for that. Additionally, the cost of actually getting that data is grossly prohibitive, an annual subscription for Compustat and Capital IQ runs in the $30k range. Moreover you would have to deal with data management problems, user interface development, simulation design minutiae, etc. etc. etc. The thought of it makes me want to vomit.
So in short, I expect problems like this to come up from time to time. It is important though, that Marco and team are not dismissive of users concerns, and that they do not always resort to the fall back position of “overfitting” whenever they make a system change that changes users results. While overfitting is a very real issue in systems like this, it is very difficult to separate it from what might actually be errors in the system causing a degradation of trading signal quality. This can only be done with deep painstaking analysis.
One unsolicited suggestion for Marco and team based on my technical software experience is to make the upgrade/change process more deliberate. There is a reason that traditional software uses versions and beta testing for new versions. Cloud based software implementations remove some of the logistical requirements for using such a system, but in a technical software application it is critically important that you have your power users test new changes (especially system-wide changes) before you implement them. I believe this is especially true if they want to have a robust Ready2Go User community. Ready2Go users cannot deep test the suggestions they are getting from Ready2Go models, and as such if a change is made to the underlying system (like the way NAs are handled) it can really damage the reliability of those models and users will have no recourse or way to know what the changes might have done. One off changes to a certain formula or introducing new formulas maybe you can get away with no testing, but for changes that truly have a system wide effect you really need to have some type of beta testing first with key users and especially Ready2Go developers.
Okay that’s my five dollars worth of shouting into the void. Hope you guys have a good day.
Cheers,
Daniel
Well said Daniel!