It is exactly what theymos want to get from the bump score: To reduce effects of shit-bumps from spammers, bumping-service-providers, and shill-accounts.
Today we can see topics on those 4 boards in more reasonable orders, and they have not been sorted according to the newest posts like before the bump score was applied.
No one can stop them to make their shit posts (only when they break rules, we can report their posts to moderators). If no rules are broken by them, their posts will only affect people who posted in those topics because they will see those topics listed on the top when they click on "Show new replies to your posts".
A topic's bump score is affected by the above two actions over the last 36 hours. Each action's effect decays linearly over the 36 hours, so an event which contributed 1 at the beginning will contribute 0.5 after 18 hours.
Some simulations from LoyceV and PrimeNumber7.
They both made their simulations from this one:
I think that the auto-stabilization property I described actually is desirable, though, so I changed the system to provide it. Now your super-bump power is: b*(1-0.15*s/b), where b is your base power and s is the total bump score you've contributed via superbumps over the last 30 days. This gives the same result as the previous algorithm except when any 30-day period overlaps with another.
I made a graph to fully understand what this means. With 1 Super bump per day, it takes a couple of months to stabilize to a certain Super bump power.
I simplified your formula to as follows:
b*[0.85^(s/b)]
I am fairly confident the above is the intention, but have no way of testing if it is implemented because I have superbumped threads in the past 30 days. If the 's/b' is an exponent, I can rule out the '0.15' from being the base because if someone has never superbumped any threads, the exponent will be zero, and the parenthesis would be simplified to '1-1', or zero, and if the 0.15 is multiplied by 's/b' the rolling 30 day cumulative superbump score (assuming 1 superbump per day) varies a lot for the first 90 days after the first 30 days of superbumping. Using an exponent to 0.85 results in the superbump power stabilizing within 60 days if someone superbumps one thread per day.
I have charted both the above exponent formula, and 0.15 being multiplied by s/b below:
b*[0.85^(s/b)]
0.15 being multiplied by 's/b':
I have also charted the individual superbump power if someone bumps one thread per day based on the exponent formula:
loyceV
charted individual superbump power based on 0.15 being multipled by s/b.
I have uploaded the underlying data to a google docs
spreadsheet.
I am still experimenting, but my current hypothesis is that the multiplication formula can be gamed by stopping superbumping for several days when the superbump value gets very low, while the exponent formula stabilizes quickly enough so that this strategy of gaming will not work.
Many months passed after the day bump score went in effects, how about a new update?
Some stats on the likely effect of this change:
Number of new posts during the two-week period between August 18-31 (before this change): 131219
Number of new posts during the two-week period between September 22 - October 5: 142907
Increase of 11688 posts or +9% (helped in part by Yobit I would assume).
However the number of new posts on the 4 affected boards (for the same time periods) is...
Before: 58406
After: 46781
Decrease of 11625 posts or -20%.
