Moderator: Clan Directors
is it quite messy to match the actual file you are using, to say the least. however i can much more easily export a file from the db in your format and try that for comparison.Leehar wrote:Yeah, the dataset I uploaded onto the dropbox is from a while ago, so it probably hasn't been updated since early december. Ice should have the latest version.
But you do make a good point, it is going to be very tough to reach a perfect match, so we just need to make sure we reach a close approximation of what we're looking for & remove any significant deviations.
Is there any way to match what we have in the database to the F400's dataset without a manual line-by-line analysis?
The crucial elements are the 2 clans scores, and the date?


I'd prefer in the region of 90%, so lets work on identifying the main differences between the tables, and trying to isolate where exactly they're coming from, and what we can do to fix them.IcePack wrote:I can't change Lindax there, but any admin can via submitting and e ticket.Can I ask how you were able to become the mod for the usergroup? I don't seem to recall the mechanism for processing such a change, so would be interesting to note for future reference.
On that score, I do agree that Lx is the best man for the job now that crown is gone, so could you add him as the mod for the time-being while we brainstorm how to move the project forward?
As posted here & discussed with Lx over Skype, I do not wish to be the moderator over there long term.
It's really difficult to know exactly what's going on with CC's side of things. I can only guess. I know I spent some time on it awhile ago and things weren't adding up, but it was difficult to pin point why / where the issues were. The last time results were compared in February there was some very odd results on the CC side, but I suppose that all depends on the definition of "close enough approximation". If I recall there was 3-4 very odd results in the top 10, so if 60-70% of the top 10 matching is close enough then I guess your ready to go. I'd have to look again and you / BW are going to have to decide how close is close enough.Good to know about the CL results being one of the likely elements that are screwing up the on-site scoreboard, any other ideas on what is causing major discrepancies between the 2 systems? Otherwise if it's some of these relatively minor areas since I don't think the small sample sizes can cause huge issues, we can perhaps move forward with the current scoreboard if it provides a close enough approximation of the F400. As BW mentioned above, it doesn't really make sense to hold up the useful update for smaller technicalities.


Thats because it was done in early 2014?Keefie wrote:BW,
Looking at that list of wars and the first thing that screams at me is that there are loads of 2014 completed wars missing.
K

OK.Leehar wrote:Thats because it was done in early 2014?Keefie wrote:BW,
Looking at that list of wars and the first thing that screams at me is that there are loads of 2014 completed wars missing.
K

Yeah it was done a while ago... and also the data used may not have been right up to date at that time. If you look back in this thread a little you will see the discussion. I believe there was general similarity, but anomalies, and it wasn't clear why. Unfortunately right now the details elude me beyond that it was not especially easy to implement what I did... my general thinking is that if we get caught up on details this may never happen so we need to try and find a path of least resistance that gets the clan world a result that will be better than now, even if not perfect.Keefie wrote:OK.Leehar wrote:Thats because it was done in early 2014?Keefie wrote:BW,
Looking at that list of wars and the first thing that screams at me is that there are loads of 2014 completed wars missing.
K
I'm quite happy to help out here. Have the two sets of data been thoroughly checked ? I can do that if they haven't.
K
If you can perform an analysis of the differences that could cause the large deviations between the data sets that would be very helpful.Keefie wrote:OK.Leehar wrote:Thats because it was done in early 2014?Keefie wrote:BW,
Looking at that list of wars and the first thing that screams at me is that there are loads of 2014 completed wars missing.
K
I'm quite happy to help out here. Have the two sets of data been thoroughly checked ? I can do that if they haven't.
K


This system is coded and connected to the live database. In the link below you can change the start and end dates to be what you want.Keefie wrote:BW,
Could I ask you to run your stuff again with a cut off date of 31st October. If you could post a snap of the data again I'll use that to compare against Ice's F41 for the same date.
Cheers
Keith




You can change the number included in the url as below:Keefie wrote:Would it be possible to see the full table for all clans, before you make any changes to exclude non-active clans.
Thanks











The F41 uses the same algorithm as the F400. The only differences being that the weight is set to 125 and only clan wars of 41 games and over are included. So the two datasets match.bigWham wrote:So the above score for HDN from IcePack is different again, can someone clarify?
The algorithm I used was "Algorithm of 400", which I think is different to "F41". It was sent to me in HTML/javascript format, and did what I could to convert the process.


