3 Data selection and results

We have first produced scan-to-scan light curves for all RASS sources with either a count rate larger than 0.05 cts/s or a detection likelihood larger than 10, resulting in a total of 25176 light curves. (Note that these criteria correspond to a lower sensitivity threshold as compared to the RASS Bright Source Catalog which invoked at least 15 counts and a detection likelihood $\mathrel{\copy\gaxbox}\!15$). Each of these light curves consists of about 20- 450 bins spaced at 96 min, with each bin corresponding to 10-30 s exposure time.

After ignoring 363 light curves with negative mean count rates (caused by incorrect background-subtraction in our automatic procedure) we apply two conditions to these light curves: (1) the maximum bin should have a signal-to-noise ratio S/N>3 above the mean count rate around the maximum (S/N is defined as difference between the maximum and mean count rate divided by the square root of the quadratic sum of the error of the maximum and mean count rates, respectively), and (2) the mean count rate until one bin before the maximum as well as the mean count rate for times later than 5 bins after the maximum should be consistent with zero. This yields 32 candidate sources. We then excluded visually (i) sources with double and multipeak structures (4), and (ii) sources with a rise over 2-3 bins and zero flux immediately after the maximum, e.g. like an inverted decay (3). In addition, we investigated the available pointed observations for 3 of these candidates, and excluded all 3 sources because we find persistent X-ray emission at a level below the RASS threshold. Finally, we made a correlation with various optical, infrared and radio catalogs and excluded sources with known counterparts as well as sources with likely counterparts (3). After this procedure we are left with 19 GRB afterglow candidates.