# Could the flag warning and ban formulas be improved?

This is a sort of programming / logic question based on this answer on meta:

Flaggers with a recent (past 7 days) flagging history consisting of at least 10 handled flags where >= 10% of flags were declined will see the following when they flag:

IIUIC when I today raise my very first flag and it gets declined, I do not get the warning (I may get a different one, but that's a different story).

Now, I raise nine more flags, all of them accepted. And voila, now I get the warning. Do I really get the warning about doing something wrong only after I do it right nine times in a row?

Isn't this clearly wrong?

The formula is

`````` (W+R) >= 10 && W / (W+R) >= 0.10
``````

where `W` and `R` denote the number of wrong (declined) and good (accepted) flags in the last week. The problem is that it isn't antitone in `R`, i.e., more accepted flags may trigger it.

Only look at folks who cast at least 10 flags a week. There isn't much point in doing much to knock the priority of flags from people who rarely flag anything.

is not wrong, but could be easily replaced by something like

Only look at folks who raised at least two wrong flags a week.

leading directly to a similar (but antitone in `R`), formula

``````W >= 2 && W / (W+R) >= 0.10
``````

Similarly for the temporary ban. Instead of

``````(W+R) >= 10 && W / (W+R) >= 0.25
``````

use e.g.,

``````W >= 3 && W / (W+R) >= 0.25
``````

A simpler alternative would be

``````W - R / 10 >= 2    for the warning
W - R /  4 >= 3    for the temporary ban
``````

IMHO these formulas much more directly express the intentions (like "2 wrong flags are worth a warning" or "10 good flags compensate one wrong flag") and are simpler to tune.

• One thing that might be worth pointing out is that in your example, if the OP flagged 1 thing that was declined followed by 9 things that were accepted and got no warning and then flagged two that were both declined they'd now be flag banned for a week with zero notice at all since they'd have 3/12 flags declined or 25%. – Catija Jun 6 '18 at 22:23
• @Catija That's interesting, indeed. However, we can modify you example: 1 wrong and 8 right, then 2 wrong and 1 right simultaneously. Again, no warning. – maaartinus Jun 6 '18 at 22:34
• I think the motto is: "If you're going to max your flag quota be wrong less often". The squeaky wheel gets the grease. True that two flags with one wrong is 50% but that's what they came up with. I'm too low on the Totem Pole to know the answer, just throwing out a guess. I believe there's three choices: not helpful, no judgement, and helpful. The Mods get a lot of flags, no doubt reducing the number factors into the equation, regardless of correct or wrong. People actually in the know can't say how it really works. – Rob Jun 6 '18 at 22:43
• But you're arguing that the warning makes no sense because their last nine flags were helpful... so that the warning is unnecessary... and then you say that we don't warn people enough in your comment... so I'm somewhat confused. I don't know if it helps but ... very few people flag posts ten times in a week (with the possible exception of SO)... and disputed flags don't count against you and comment flags aren't included at all. – Catija Jun 6 '18 at 22:45
• @Catija With "Again, no warning", I meant that showing the warning after the 10th flag doesn't really help as there's still a way to get banned without a warning. I guess, the warning is just a way to say "be more careful and saves us needless work", rather than "be more careful or else...". `+++` I find it wrong to warn after a series of good events. Similarly, the ban rule could get you temporarily banned just after raising a right flag. I'm not saying, it's important, I'm just saying that such a formula is principally wrong. – maaartinus Jun 6 '18 at 23:59
• I guess, a declined flag is bad as it causes needless work. Logically, an accepted flag must be good (as with all flags being bad they wouldn't have been introduced :D). So we want to minimize the former and maximize the latter, which is something my last two formulas directly express. The problem is linear. Anyway, it's an interesting point about the squeaky wheel. – maaartinus Jun 7 '18 at 0:08