2 replies to this topic

[BitterSprintUser]

Please excuse me if you're currently exercising any of these practices...

While looking at the Sprint index its obviously apparent at one point in time someone with a crazy connection (out of the norm) posted some series numbers. Well that is skewing the results IMO. You should omit the highest and lowest scores then average. To that point it would also be nice to see some sort of normal distribution (bell curve) as well as standard deviation in order to give users a much more in depth / accurate look at their results and comparisons to the index.

I sincerely appreciate the effort put forth. This site has been invaluable to me to prove my side of the "he said she said" battle with my ISP to the fact their service is sub par at best.

Regards
BSU

[CA3LE]

Please excuse me if you're currently exercising any of these practices...

While looking at the Sprint index its obviously apparent at one point in time someone with a crazy connection (out of the norm) posted some series numbers. Well that is skewing the results IMO. You should omit the highest and lowest scores then average. To that point it would also be nice to see some sort of normal distribution (bell curve) as well as standard deviation in order to give users a much more in depth / accurate look at their results and comparisons to the index.

I sincerely appreciate the effort put forth. This site has been invaluable to me to prove my side of the "he said she said" battle with my ISP to the fact their service is sub par at best.

Regards
BSU

Welcome to the site! I love the suggestion. I need to think about how to turn that into a mySQL query now... that's actually a very complex statement to program. But it makes sense for sure. I would have to drop the top 5% and bottom 5% (for instance) then average the remainder. I doubt that mySQL has command to do that... but if it does an addition like this would only take me a few seconds to work in. Although, if the query is too heavy then I'll have to cache the results otherwise people will be waiting ... ... ... if I can do it, practically, I will.

One thing I really need to do is narrow the hosts down to their region instead of as a whole. Currently you're comparing to everyone on the host... nationwide. I have allot to do before I can work that level of complexity into the database... but I am working towards that. With this addition will come better host detection which will help with the large number of hosts being detected as 'Unknown'. It's allot of work and I rarely look forward to database work. But it will make this site a much more powerful tool.

So many layers to this site that I have to continually update and improve on... I think need 10 more keyboards and 100 more fingers.

Kinda sad where Sprint is on the list...

http://testmy.net/mobile/hoststats/

[BitterSprintUser]

I'm glad you to see your interest. I'm no SQL expert by any means but I've been in IT for about 12 years specializing more in hardware and sys/network admin.

Here's a dirty way you may be able to pull off the deviation.

Take your data - see how many results you have (lets say 1000 for example sake) - figure out what 5% is (50) - now remove the max value from the list X times (50 in this case). As far as standard deviation I'm pretty sure there are statements for that (version dependent) its usually STDDEV (expr) or STDDEV_POP (expr) respectively VAR_POP would give you a standard variance.

Best of luck and again thanks for the site - its invaluable.

p.s. no need to remind me how terrible Sprint is - they do a fine job doing that LOL