Reporting ONLY 99% accuracy?

[philipma1957]

so first off here are my stats


845 reported
817 good
  14 bad
  14 not handled


so 817/831=0.9831

and yet I am marked for 99 percent

even if all not handled were good

I would be 831/845= 0.9834

yet I have 99 percent

I am fine with a drop to 98.31 or 98.34

but the question is how do we handled not handled when we score.

[hopenotlate]

Almost every time I flag a post, my eye always falls on those unhandled reports: I don't understand why they exist.
A report is either good or bad, but why does it remain unprocessed indefinitely? I don't think the moderator(s) in charge could have ignored it from the beginning and never had time to look at it again after weeks/months/years.
They certainly examined it and decided not to process it : in which case it's a bad report and not an unhandled one.

What exactly causes a report to be unprocessed?

[LoyceV]

Almost every time I flag a post, my eye always falls on those unhandled reports: I don't understand why they exist.

The Your reports-page explains it:

A report that stays unhandled can be thought of as "soft-bad", since no moderator positively thought that it warranted action.

A report is either good or bad, but why does it remain unprocessed indefinitely?

In some cases, I get that no Mod wants to handle it because it can be debated, but that doesn't doesn't mean it's Bad either. That makes it a gray area.
In other cases, like this one, I don't get why it's not deleted.

I don't think the moderator(s) in charge could have ignored it from the beginning and never had time to look at it again after weeks/months/years.
They certainly examined it and decided not to process it : in which case it's a bad report and not an unhandled one.

As long as the report is Unhandled, other Mods can see it too. If it's still Unhandled, that means none of the Mods think it deserved action.

[Welsh]

To prevent users caring too much about 'damaging' their accuracy maybe we should have this instead of percentages:

• If anything is perfect its you...
• You're doing a cracking job!
• Not bad at all...
• Getting there...

And for 0% maybe we should have:

• Don't give up your day job.

I'm kidding off course. We are all far too egotistical to let go of our precious percentages . Also, they do tend to motivate for those of us who are chasing the 99.99%

One final thought: I don't think many people realize the extent to which theymos believes that users shouldn't lose any sleep over their accuracy. For example, I thought I had internalized the whole "one accurate report is worth many inaccurate reports" thing, but, I obviously hadn't, because bumping into the below quote was still something of a surprise to me:

C'mon PowerGlove. Come up with a brilliant system that would both motivate and reward users for reporting without them worrying about damaging their scores .

[ibminer]

To prevent users caring too much about 'damaging' their accuracy maybe we should have this instead of percentages:
~

Sorry for the wall of text, but my AI response had a lot to say (below).


Yes — there are a number of systems that would likely work better than Bitcointalk’s current “good/bad report percentage” model, depending on the forum’s goals.

Right now, Bitcointalk’s system mostly measures accuracy, not necessarily value. A user can maintain a very high “good” percentage by only reporting obvious spam, duplicate topics, or easy moderator actions. That does not necessarily reward users who:

• identify coordinated abuse,
• catch scams early,
• report ban evasion,
• document plagiarism thoroughly,
• or spend significant time improving forum quality.

The current system also creates a few behavioral distortions:

• users may avoid borderline reports to protect their ratio,
• users may mass-report trivial violations,
• moderators become the sole “truth source” for whether a report was “good,”
• and the public stat becomes more about reputation than actual contribution.

Some alternative or hybrid approaches could work better.


1. Weighted Report Reputation (Probably the Best Overall)

Instead of counting every “good” report equally, reports could earn weighted value based on:

• severity,
• uniqueness,
• accuracy over time,
• and moderator agreement consistency.

Example:

• obvious duplicate spam = 1 point
• plagiarism with evidence = 10 points
• catching an account farm = 25 points
• identifying a phishing campaign before damage spreads = 100 points

This rewards impact, not just quantity.

It also discourages users from farming easy reports.

A reputation system like this could even decay over time so older contributions matter less than recent accuracy.


2. Delayed Accuracy Scoring

One problem on forums is moderators are not always consistent.

A smarter system could:

• track whether later moderator reversals occurred,
• whether admins later banned the user,
• or whether multiple moderators independently agreed.

This reduces:

• moderator bias,
• inconsistent rule interpretation,
• and “luck” depending on who handled the report.

3. Evidence Quality Scoring

Currently, many reports are just:

“spam”

A much stronger system would reward:

• links,
• archive references,
• plagiarism proof,
• behavioral patterns,
• alt-account correlation,
• transaction evidence,
• or timeline documentation.

This would favor investigative work rather than simple button-click reporting.

Bitcointalk especially has a culture where detailed evidence matters heavily, particularly in Scam Accusations and Reputation sections.


4. Community Validation Layer

Instead of moderators alone determining “good” or “bad,” trusted community reviewers could validate reports.

For example:

• moderators make the final action,
• but veteran reporters could endorse the quality of reports,
• similar to peer review.

This distributes trust somewhat more decentralizationally.

The downside:

• clique formation,
• retaliation,
• popularity bias,
• and politics.

Bitcointalk historically struggles with those issues already in areas like DefaultTrust debates.


5. Incentive Tokens / Merit-Style Rewards

Bitcointalk could theoretically reward valuable moderation assistance similarly to how Merit rewards valuable posts.

Examples:

• periodic recognition,
• special badges,
• moderator nomination pathways,
• temporary perks,
• or even forum privileges.

The danger is obvious:

• users start reporting for rewards instead of quality,
• competitive behavior emerges,
• mass-report abuse,
• or targeted harassment through reporting campaigns.

Any direct financial reward system would almost certainly become gamed quickly.


6. Transparency Metrics Instead of Percentages

A raw “98% good” stat actually hides important information.

A more useful profile might show:

• spam reports,
• plagiarism catches,
• scam investigations,
• moderator reversals,
• average response time,
• confirmed bans resulting from reports,
• unique reports vs duplicate reports.

This paints a fuller picture of contribution quality.


7. Anti-Gaming Mechanisms

Any advanced system would need protections against:

• report brigading,
• retaliatory reporting,
• report farming,
• moderator favoritism,
• and social alliances.

Bitcointalk’s simplicity is partially intentional because simple systems are harder to exploit at scale.

That’s one reason the current system has survived so long despite its flaws.


The Bigger Philosophical Issue

Bitcointalk historically leans toward:

• minimal structure,
• minimal automation,
• moderator discretion,
• and social reputation systems rather than algorithmic governance.

So even if technically better systems exist, they may conflict with the forum’s culture.

The forum often prefers:

• imperfect human judgment,
• decentralized social interpretation,
• and visible reputation battles

over highly engineered systems.

That’s also why debates around moderation, Trust, flags, and reporting never fully disappear there.


A Hybrid System Would Probably Work Best

A strong modernized approach might combine:

Component	Purpose
Accuracy %	baseline reliability
Weighted severity	reward meaningful reports
Evidence scoring	reward investigative effort
Time decay	prevent ancient stats dominating
Moderator agreement consistency	reduce bias
Public transparency categories	clearer contribution history
Anti-abuse heuristics	stop farming/manipulation

That would likely produce better incentives than a simple “good/bad percentage” while still remaining understandable to users.