Frequently asked questions

Uncle Stock is a fundamental stock screener for investors. It allows you to search for the best stock picks using financial statement information. We offer our users a huge amount of financial ratios, advanced scoring mechanisms, a broad coverage of financial markets and the possibility to test your investing strategy historically, using a backtest.

Every new user gets a free trial of 2 weeks to learn about the product before taking the paid licence.

The best browser for Uncle Stock is Google Chrome, the second best is Opera.

We only keep your name and email adress. We certainly do not give this information to other parties.

In our glossary.

You can pay using PayPal or bank transfer.

Yes we do, every user has a personalised link which can be used to get a referral reward.

Yes there is: we have an active facebook user group.

There are over 450 metrics available. For these metrics, multiple sub-metrics are available, giving a total of 7.600 metrics, of which 1.500 can be used in searching.

A more detailed list can be found here.

Financial results from US come from Sharadar. Financial results for the rest of the world come from Eod Historical Data. Price history comes from Eod Historical Data.

Price data is delayed. Price delay depends on the market. For Nasdaq markets, price data is end of previous day when the market is open.

Uncle Stock tries to import new financial results within 2 weeks after they are available at source.

It depends on the popularity of the stock. On average indicators are recalculated every 10 days using the last price. At least this is once per month.

Uncle Stock tries to make indicators actual when viewing, by recalculating all metrics based on the latest price (and any new results) if there are new results or the price deviates too much from the one used at previous calculation.

There is a monthly background processing of all stocks, in which

1. it is checked whether it is time to try getting new results. In the months that results are expected, this is 2 weeks.
2. recalculate all metrics based on the latest price (and any new results)

When a stock is consulted by a user, the same happens (this to give popular stocks a shorter cycle)

1. it is checked whether it is time to try getting new results. In the months that results are expected, this is 2 weeks, otherwise 1 month.
2. recalculate all metrics based on the latest price (and any new results), if there are new results or the price deviates too much from the one used at previous calculation.

You can more or less define your own universe: Just do a search, and when saving it, in the save dialog, mark it as default. The criteria of that search will act as default criteria for a new search.

If a market is supported, you can simply add a new stock to the screener by giving in the ticker in the Yahoo notation once in the search. Afterwards, it will be included in the screener forever.

When entering a name of symbol, Uncle will show a list of hinted stocks, that can be selected. Important: The hinted stocks are the ones that were already loaded in Uncle before. An existing symbol for which no stock is hinted (because it was never loaded), can also be entered! Entering such existing symbol will load the stock in the Uncle Stock datastore, and make it available for everyone.

More information on looking up a stock can be found here.

A screen is the combination of a sort metric with some filters. The filters selects which stocks are included, the sort metric determines which stocks come on top of the screen.

The backtest checks the effectiveness of a set of criteria historically. It can look back 10 years in time and also compares yearly renewal with buy and hold.

The backtest always takes all the stocks meeting the filter criteria with a maximum of 24. If there are over 24 stocks meeting the criteria, the sort is used to select the best 24.

There is just a little bit of survivorship bias. Delisted stocks are kept in the selection if they were delisted, but not in the year of selection – so later, which is already good. However if they were delisted in that year, they are removed from the result set because I can’t know whether the stock is delisted because of a take-over or because of bankruptcy. So I can’t know the actual return of that stock in that year. And if a stock has missing historical data, it will not be found by the backtest (as it searches on historical data) so these stocks will not have impact.

Uncle Stock is certainly resistent against look-ahead bias. As we only use results from the previous year.

Every year at the beginning of April. For statistical purposes, this as as good as Jan to Jan, but April has the advantage that most financial results should be there. So it is a better moment to test, as the purpose of the backtest is to test for the effectivity of the criteria.

The backtest has a final part that is less or more as one backtest period. The annualization takes this correctly into account.

As of two months after the start of a quarter, the backtest can simulate buying the start of that quarter. So buying at Q1 (April 1st) of this year is available as of June 1st.

The backtest uses only year results. For rebalancing start of Q2, Q3 or Q4, the year results of previous year are taken. For rebalancing at start of Q1, the year results of 2 years ago are taken. The market value (price) component of the metrics takes the price at the start of the quarter. So Price/EPS at start of Q2 2014, uses price at start of Q2 2014 divided by EPS of 2013. Because of this, start of Q2 is the most effective moment to test a strategy.

Note that the current query uses quarter results and 12 month trailing values.

Yes, the return is calculated using the dividend adjusted price.

In a few months time, there are actually hundreds of tiny changes, which adds up to big change. Improvements on data (more stocks, more financial results available/depth, data corrections, changes in interpretation), as changes to the algorithms, refinements and corrections. I always reindex full history, to have the historical data be inline with current code and data. So the current backtest is just better.