Since the trickle \"problem\" was \"solved\" about 10 days ago, I\'ve been getting an average credit of around 9000 per day. While this is great, it isn\'t correct. However, I haven\'t seen anyone else complaining...

1. Does anyone else have a similar problem?

2. If not, what could be happening to my results?

The average credit for CPDN in BOINC has never worked well and continues on in that tradition. The recalculation that occurred a week or two ago after all the previous problems resulted in particularly goofy RACs. These may even out after a long time to something reasonable, or they may not. In CPDN, the only reliable indicator of work done is Total Credit, and that appears fine on your account.

I had thought I was using the appropriate "half life" calculation now for RAC (since we calculate it instead of BOINC). it's something I will eventually get right! I believe the problem may be applying the "half life" to the total credit for a result rather than the incremental fraction coming in for a time (i.e. the 98 credits per trickle and not the cumulative for being at, say, phase 3, timestep 54000 etc)

Carl wrote: <I> I had thought I was using the appropriate "half life" calculation now for RAC (since we calculate it instead of BOINC). it's something I will eventually get right! I believe the problem may be applying the "half life" to the total credit for a result rather than the incremental fraction coming in for a time (i.e. the 98 credits per trickle and not the cumulative for being at, say, phase 3, timestep 54000 etc)</I>

I'm assuming you have "fixed" it as I'm now getting a more realistic RAC (90+/-).

&gt; I'm assuming you have "fixed" it as I'm now getting a more realistic RAC
&gt; (90+/-).

yeah, I just tried it, it's a semi-kludge but seems to give more normal #'s (CPDN, with trickles etc for credits has always been odd for RAC) -- what I do now is use their "half life exponential decay" applied to the average trickle date for trickles across a result, and then multiply that with the max credit the result has returned so far (i.e. the credis for the latest phase/timestep). so it seems to be sensible.

&gt; yeah, I just tried it, it's a semi-kludge but seems to give more normal #'s...

Thanks