It’s been almost a year since the initial introduction of our ranking system. We want to thank the community for the amount of feedback we received.
Based on the feedback, one issue was pretty clear, which was the determination of seniority. Seniority basically means that if your gateway was the first in an area then you should get priority in that area to deliver data packets that the network will ultimately rely upon as a source of revenue. It also discourages others to establish excess capacity in that area and move that capacity to an area that’s less covered.
Since establishing a good ranking system is still viewed by us as one of the most important factors in Crankk’s future success we recently refocused our attention to make improvements on the initial version.
Once we fixed the seniority issue and added extra analysis to the results, another important aspect came to light. Before explaining what it was, let us get back to the original design itself. If you’ve read the original post it is clear that we are relying on the counting of unique packets and how many gateways receive that data packet. As it turned out, some of those data packets were not unique enough and we got a much higher number of receivers per packet than the actual. This lowered the scores. We believe that we managed to fix this as well and now the picture is much more accurate.
We’d like to reiterate that nothing has changed in the original design of the rank. We just made it more accurate. No personal or proprietary data is used in the calculation. It still relies solely on the analysis of data packets received and forwarded by the gateways to Crankk’s LoRa servers.
Why do we assume this version is better now than the previous version? One way we validated was the examinaion of the likely closeness of one gateway to another. Closeness here means simultaneous reception of the same unique data packet and not based on assumed location of the gateway. You can assume closeness or colocation by looking at the “neighbors” of a certain gateway and cross-check it from the neighbor’s point of view. These should form certain groups of simultaneous receivers and, if correct, not yield irreconcilable results. Based on that we’re fine with this version as a step forward.
So, as we hope, it’ll be mostly good news for most of the community. The overall average score now hovers around 52% which is pretty good. You can interpret it as roughly the redundancy that we were aiming for as being acceptable (1+0.5+0.333+0.25)/4 = 0.52. Refer to the original post for some details on the calculation if needed. Of course this is the average with half of the gateways doing better and half doing worse. It means that even though the score has not had an impact on rewards most people deployed their capacity responsibly.
The new scores will be visible shortly, replacing the old ones. It still won’t have an impact on your earnings but the goal, once we gain feedback from the community again, will absolutely be to get there as soon as possible. The rank is and will remain a relative scale. It will not be used to directly multiply your rewards. Instead, it gives us an opportunity to come up with a formula that enables a certain differentiation in rewards. We’ll put this formula up for public discussion as well to get the maximum buy-in from the community. The goal is to find the optimal impact, not more, not less that provides the best gain in the overall score.
Thanks again, for being on this journey with us, we’re doing our absolute best to make and keep you happy with the project.