You are viewing a single comment's thread from:

RE: SteemWorld ~ Weekly Support ~ #32

in #steemworld5 years ago

Sorry to hear about the EMF. I'm sure you've looked at possible solutions, but feel free to contact me on Discord as there may be more options.

Great work, as always, and especially under the circumstances.

One Q, nice touch to see the grand total curation efficiency as the rewards are being calculated, but is it a simple average of all curation %s or is it the global sum of rewards divided by upvotes values?

Sort:  

It is just a simple average, not weighted with upvote values (VP). A few hours ago I had a discussion with him where I had the same question^^ and this was his answer.

Thanks :-) One odd thing tho, ran the coming curation rewards last night and ended up with about an effectiveness of 85% - ran it again, on a different machine, today and scored 105%. I just assume a bit of tweaking is taking place, but the two averages can be different.

I used to run it daily but takes way too long for my rewards fund - am seriously impressed it doesn't crash or halt, given the huge number of upvotes per week.

I bet you were testing when he was arguing with me which calculation to use. Because he was trying out both calculatons an testing them on multiple accounts.
Mine was 56% (weighted) and after our discussion, it was 67% (non-weighted).

If I am right about this assumption, it should not change anymore on any machine you are using.

:-) you may need to have that discussion again! What you call "weighted" is the "correct" value in the sense that it is the sum(curation-rewards)/sum(upvotes), both using the same currency. Hence why I would have thought it a much easier calculation to perform recursively as just adding numbers to a cumulative quotient. The "simple mean" of percentages gives a distorted picture as it inflates high yields on very low upvotes. eg upvote=1steem, curation=0.5, plus upvote=0.1, curation 0.4, can give an average of either 225% or 82% ;-) The real return, financially is 0.9/1.1, which is 82%.

just my opinion, but so long as I know how to interpret it.

Hehe, here we are again... I'm no mathematician, but I think it depends on what we want to measure.

The simple average (current logic) gives us the average efficiency in terms of percentage per vote. The 'weighted' average (efficiency * [vote_percentage] / 100) would show us the average in terms of how much we really earn. Of course, if we would be as efficient as the simple avg. shows (for each of our votes), we would have the efficiency in terms of earnings.

I'm still not sure, what would make more sense here.
What about displaying it as:


Avg. Efficiency | 67.12%
Weighted Avg. | 56.87%


'Weighted Avg.' might not be the best name for it though.
I'm open for any ideas :)

Problem solved, just show both :D
I only need the simple average, but I understand if people want the weighted one which is closer to reality but worse for improving curation results.

Maybe just call it avg. efficiency including(or factoring in) vote percentage

Hi, hope you're well. One query from my side as a mathematician but not coder, isn't the weighted average easier/faster to calculate in that it is a recursive addition of the numerator/denominator of the quotient?

And... yes, if there is space, can show both :-) I agree the names are not immediately obvious, perhaps, to many users.

the avg.eff. is the "average return per vote"
the weighted avg. is the "average return per stake" (or per SP)

What do you think?

Yeah, I know that the weighted calculation is the "correct" one, but I noticed that my small upvotes on average provide better efficiency, so if I want to optimize my curation efforts, the weighted calculation would distort my progress.

Based on the efficiency numbers shown on steemworld, I bet that I can improve my curation by using more small upvotes (which I will try now). Maybe you as a mathematician can explain why this is going to improve my efficiency and give me higher curation rewards in total. Because I don't understand 100% why it seems to work this way.

Yes, I discovered this some time ago: increasing smaller votes gives better returns... It's a consequence of binomial distribution, assuming one has an edge to start with, that edge increases ;-)

If you're mixing strategies with the same account then, yes, you may wish to separate the earning-strategy from just upvoting friends and good posts.