Silk Road forums
Support => Feature requests => Topic started by: bpe on December 16, 2011, 02:45 pm
-
Hello community,
I have a relatively simple idea to fix the kind of flawed reputation system. At the moment it's possible to boost ones reputation, both buyers and sellers, by trading low-priced (i.e. digital) goods and receiving a good feedback in the process. A scammer can even setup multiple fake accounts by himself and buy the reputation for a very small price.
The solution to make potential buyers more transparent for sellers is very easy: Besides the total number of transactions also display the total volume of all transactions combined in BTC and dollar in the buyers stats. This is much more meaningful. A seller can than state in his terms, that all people with a total successful transaction value i.e. under 300$ have to finalize early. So he doesn't have to fear, that the "27 transactions" displayed were just ebooks for 0.10 BTC each.
The solutions for sellers is a little more complicated. First of all it would be helpful, if the review list displayed on a sellers profile would also have a column for the value of the transaction. This would make it easy to sort out much faster, if all the good reviews are from cent-purchases(you can also get that information by clicking the item link, it would just be much easier this way). The second idea is to change the equation that calculates the trust-level of a vendor(the number behind the user name, currently "percentage of positive feedback"). Here is my suggestion:
Let "I" be the set of transactions a vendor has finalized. Let "a_i" be the dollar volume of a single transaction "i", and "b_i" a weight for that transaction, equivalent to the feedback points given minus three. (So 5 points result in a weight of 2, 1 point in a weight of -2). The trust-level "t" of a vendor would then be calculated like this:
t = (sum(a_i*b_i)/sum(a_i))*50 for all transactions in I
This would result in a sellers rating between -100 and 100 which takes into account the amount of each transaction as a weight to determine the trust level. The actual equation is of course just a suggestion, you could easily map the output between 0 and 100 or even use a log on a_i to smoothen it out for high value transactions.
This should just give an idea and start a discussion :) Hope you get my point on this.
Regards
bpe
-
You should consider posting this in the "Feature Requests" subforum: http://dkn255hz262ypmii.onion/index.php?board=9.0
-
Could not agree more BPE. Excellent idea.
Too many 5/5 for who knows what (0.00001BTC?) sales just to artificially increase a sellers rating.
Buyers need to know what the rating numbers really mean.
-
Mr Dank started a thread and I responded, however it's still last night for me and I can't find it, but I linked to where SR explained how it works.
More weight is given to buyers who buy from a wide variety of vendors, and higher priced items. Lots of feedback for low amounts may fill the slots, but won't send a vendor up the ranks. It's pretty well thought out and hard to game, and I'm sure it's also like Googles search algorithms, constantly in a state of flux.
Multiple ratings from a buyer who doesn't also deal with multiple vendors carry less weight, as do low cost purchases.
Except for Mr Dank, I don't think anyone here, once they actually read and study how it's done now, thinks the system is being abused. 5 point feedbacks for finalizing early definitely is a prob, as it's been used by every vendor here who's done a big rip-off and runner.
mseller just posted a thread with a good solution to that in this forum.
-
True VJ.
I noticed and that should be set maximum what buyer purchase stats can have effect on vendor rating.
Also, purchase rate should be determined (rate effect) when buyer ordered and not when he finalize or left feedback because meanwhile many things can happen.
That happened to me. One buyer who had 2 purchases ordered item from me worth $20. He unjust gave me 1/5 and because that, my rating droped from 99,6% to 90,5%. Meanwhile from time when he ordered to finalize he accumulated more then 40 transaction. ( I do not know how much those transaction worth).
I have lucky and sorted this unfair feedback but imagine that buyer can ruin seller reputation just because one 20$ worth item?
Another example is what happened to Nucleo.
He has 40 5/5 transactions and last were 4/5. Because that buyer has around 35 transactions he droped by more then 2% - from 100% to 97,7% - because 4/5 feedback!!?
Imagine how much would he drop if that buyer gave him 1/5? He would have maybe 89%.
That why I think there should be set max limit on purchase rating.
-
+1 OP.
I was going to suggest about the same thing. Imagine when 2/3rd of your transactions are $500+ and you've consistently gotten 5/5 then a buyer trashes your rating on a $20 sample size order giving you 3/5 because they were too dumb to read that regular shipping was going to take three weeks or whatever.
Also the seller *rankings* don't seem to consider the $$ volume at all. I see sellers ranked in the top 20 that do nothing but $0.00 transactions meanwhile seems like my ranking keeps dropping the more my volume increases. :-\ I do a transaction for $5000 in escrow takes me over two days just to package it all, pay over $300 fees to SR and my rank doesn't budge a single point. Meanwhile sellers doing $200 a day are passing me. ???
Ratings/Rankings should definitely be proportional to $$$ amount. And this would also encourage more people to work through escrow for large transactions which means more $$$ for Silk Road.
Additional suggestions: Maybe put less weight towards ratings by buyers who've posted a ridiculously large proportion of low ratings... ie: scammers, feedback trolls, etc.
-
Or the system could jsut ignore any transaction less than say $40 USD. One below that could exist to show history but have no weight on anything other than the text that was written there. That makes the cost about $2.5 USD for each rating. A significant deterrent to false buyer spam, and makes things like digital purchases not count toward buyer rating #.