I personally feel it's better to look at a coin's average... maybe 24 hours is a good indicator, but I prefer longer averages to the tune of 3 days to a week. You will get much more consistent profits by mining a coin that stays popular at #3 all week than trying to chase coins that tank before you can see the profits. It's kind of counter-intuitive. I'm hoping Patrike can add some more customization in determining the parameters on how the profit switching works. Whattomine.com already has parameters to look at 24 hour, 2 day, and 3 day averages.
I started looking into this a little while ago when it was discussed last time, and I found a way to support 24h avg statistics for Nicehash, zpool and WhatToMine - but not for Mining Pool Hub.
I will go ahead and make the implementation for these sources that supports it, because it's only a small implementation. I will add a new settings in the Options dialog (probably in the Statistics sub section for Coins&Profit) where you can select between "Current" and "24h average". This will be used for the information you see on the Coins tab, the Online Services tab and for the profit switcher.
You are correct that WhatToMine is very flexible here, but the other sources are not.
Yes, the only way you could do it with MPH is to have AM keep it's own running average every time it checks WTM or Coinwarz. Which kind of lends itself to a couple of questions I've had.
When using profit switching, AM is strictly relying on what NH, MPH, and ZPool say is the most profitable, correct? the Coins tab data is really just there for convenience? Is there a way to match what WTM says is "most profitable" to one of the profit-switching services? Or does that happen only when you use a custom pool grouping with a managed profit miner?
Another question I have deals with the hashrates defined in the Algorithms or Profit Profiles. When we manually put in the hashrates or use the really awesome new benchmarking feature... should the rates defined there be a single card or the entire rig? Does AM extrapolate that I have 6 cards and use the multiple rate when determining if switching to a different algorithm should be done?
As long as your benches are all the same--pegged for one card or pegged for six--it doesn't much matter. Personally I have a separate profile for my 1070 rig as they actually hash slightly different from my 1080 Ti rigs. Each set of my benches are pegged to 1x of that card.