- Adjust profit calculations for Yiimp-based pools to actual performance based on past payouts compared to estimates. Configured in Options dialog, Statistics section.
awesome!What's good is that you don't have to run your own Awesome Miner for a period of time to collect this data yourself. I can also make changes and optimizations and everyone will get it right away. Because I store the original pool statistics in the cloud all the time, any new calculations can use all old pool statistics as well.
I'm very much open to any feedback and discussions about how to improve this. At least the infrastructure for the feature is in place now.
This is definitely a great step - its good to have this information, however it begs the question of how to properly use this information.
I've never thought the pools were necessarily stealing or anything, just that the delay between the mining and exchange and the occasional difficulty spikes and drops accounted for most of the variation between them. So looking over the instances where all the pools are mining the same algo, there isnt a clear pattern where one pool has consistently better performance than the other.
Anyway, from the description given, it sounds like all AM is doing is taking the reported 24hr actual, the reported 24hr estimate, and then comparing the two over time to give a score for how close the estimate is getting to the actual.
If this is correct, there are then two obvious questions:
First...this feature was highly requested because of mistrust of the pools. If all that's being done is comparing statistics reported by the pool, does this really solve the problem? How do we know the pools are actually reporting an honest 24/hr estimate or actual?
Second - what does this say about which option we should choose now? Does this correction factor modify the reported actual rate to come to a theoretically more accurate estimate - meaning that we should now choose actual to get this more accurate estimate? Or does it correct the 24hr estimate to a now more accurate estimate - meaning that we should choose 24hr estimate? Or is choosing current more viable now, because the stats collected by AM now account for the variation - so if a coin is being reported as currently being super profitable, if it remains more profitable after the AM correction, it's probably legitimate and worth taking advantage of?
Overall I think it would be helpful if you could be more specific about how this calculation is made, what the time window is, etc. It's a step in the right direction but it's not really clear at all how to best take advantage of it.
Also, along the same lines, I think it would be potentially even more useful if AM were to calculate some sort of variance factor for pools/algos - in other words how much profitability varies throughout the hour/day. There are some coins/algos that are very stable, and some that constantly fluctuate between high and low profitability on what seems like a minute by minute basis. I've had to disable a few algorithms simply because they keep causing constant switches for no good reason. Even with a 1 minute interval, I could watch revenue go from an extremely high number to a ridiculously low one. It'd be nice to put a number on that so I could have some data as to which algos/pools are too inconsistent to be worth my time. I want to be mining the most profitable coin I can be at any given time - and 24hrs is too long a time window for most of the coins on yiimp pools.
This information is critical because these pools are reporting averages, and outliers throw off averages. (Using simple and unrealistic numbers to make the point) Two algos could report 10 mBTC/GH for 24 hours - but one could be getting to this number by being within +/- 5% of 10 mBTC/GH for 24hrs straight, and the other could be .01 mBTC/GH for 23.9 of those hours with a few minutes or even seconds of 1000 mBTC/GH every day. Mining the latter is a waste of my time when even a 1 minute interval leaves my miners on the wrong side of that for up to a minute - those minutes add up when you've got lots of algos acting like this. I mean I certainly wouldn't want to use 24/hr estimates or averages if it meant my miners were potentially making nearly nothing 99% of the time and hoping that 1% of extreme profitability actually delivers enough to make up for it.
Anyway, this is still good data once it's clear what to do with it. Eventually I'd hope for it to get to a point where AM can identify between consistent earners and inconsistent earners, and use that information to help us exploit legitimate profitability spikes while mitigating the harm of profitability drops. Even something as simple as the profit switching being much more responsive and cutting a poor earner off early would be useful. Like as far as I can tell the stats update and switching intervals are not even in sync - there is absolutely no good reason it should continue to mine something for another second when the stats update showed profitability fell off a cliff, but there's still up to a minute left before it hits the switching interval and does something about it.
