Author

Topic: Changes to how Scammers' trust scores are calculated? (Read 1580 times)

copper member
Activity: 2870
Merit: 2298
You have double posted 3rd para in 4th. Smiley



What if theymos has updated the trust formula but hasn't posted here?

   -MZ
Thanks fixed. I was really tired when I made that thread and found a few grammar mistakes when I first posted it.

I am pretty sure the trust calculation for scammers has probably changed somewhat to prevent scammers from continuing to troll to try to get the 'worse' trust score
hero member
Activity: 560
Merit: 506
I prefer Zakir over Muhammed when mentioning me!
You have double posted 3rd para in 4th. Smiley



What if theymos has updated the trust formula but hasn't posted here?

   -MZ
copper member
Activity: 2870
Merit: 2298
I think it decreases a little over time. When someone is flooded with loads of negatives their trust score will be very high but will fall down a bit after a short period (not sure on the exact figures/times).
All of GoldBits's negative trust reports were made today. As per the above formula, unless you are a long term con man (e.g. have more then 4 negative trust reports) then reports less then 7 days old are counted 1.5x what they would otherwise be worth
global moderator
Activity: 3794
Merit: 2615
Join the world-leading crypto sportsbook NOW!
I think it decreases a little over time. When someone is flooded with loads of negatives their trust score will be very high but will fall down a bit after a short period (not sure on the exact figures/times).
copper member
Activity: 2870
Merit: 2298
Has the way that the trust ratings are calculated changed?

While looking at a scammer's trust rating ( GoldBits), I see that his trust score is "-50" or -50: -9 / +0(0) when looking at it from my account that uses a "corrected" trust list, and -39: -7 / +0(0) when looking at is from an account with a trust depth of 2 and only trusting DefaultTrust (which is the default).

Additionally I see that TF'strust score is -830: -19 / +21(21). I know for a fact that in the past TF's trust score was both negative and in the four digits, and he has not received any positive feedback recently (this is looking at his trust from an account with a trust depth of 2 and only trusting DefaultTrust).

I know in the past that once a user gets 4 or 5 negative trust reports their negative trust score increase exponentially (the absolute of their score increases; the actual number goes down a lot). After doing a bit of research, it looks like this is the most recent formula for calculating trust scores:

Code:
scam_reports = total unique users who report scamming
reports = total positive ratings
btc = total BTC from positive ratings
unique_reports = total unique users who posted positive ratings
oldest = timestamp of the oldest positive trust rating
latest_scam = timestamp of the latest scam rating

scam_multiplier = 1
# recent scam = bigger negative score
if(current_time - latest_scam < 7 days)
    scam_multiplier += 1.5

# new member = bigger negative score
if(current_time - oldest < 60 days)
    scam_multiplier += 3

months = months since oldest
total = reports + btc/50    # 50 BTC = 1 report

# long-term con men
if(total > 0 && scam_reports > 4)
    scam_multiplier += (scam_reports / total) * 50

# Approximately limit the number of points each person can contribute each month
subtotal = min(2*unique_reports * months, total)
# Newer users have smaller point limits
limited_total = min(subtotal, 10*months) / 10

score = min((limited_total * months - scam_reports * scam_multiplier), 150)
So by my manual calculation, GoldBits should have a trust score of -4,050 or -4,050: -9 / +0(0) with 9 negative trust reports, and -2,450 or -2,450: -7 / +0(0) with 7 negative trust reports and zero positive trust reports (as is reflected by default).

I am not going to attempt to calculate what TF's score should be manually, however I am confident that it was lower very recently.

If this is not the most recent formula for calculating one's trust score, can someone please point me to a more recent formula?
Jump to: