Draskon, post: 2607, member: 808 wrote:
Appreciated!
I agree that a visual coloured text system should be implemented, and it'd be nice to have the colours for each threshold configurable to match a user's liking, but not the actual thresholds themselves if at all possible, since that then will turn what's supposed to be a more objective grading system fundamentally into a purely user-by-user subjective system, overriding the basis of the thresholds set according to the data the grades are based on, if that makes sense.
Using probability as a means to determine theoretical limits is a fair approach but still a bit dicey because of statistical outliers. They can and will exist with the system as it is which makes the grading system trickier for another reason:
What do we want to base the grading system on, precisely? Do we want to base it on rarity, or part performance, or the results after RE'ing? (In the case of VsA/VsS/EPS/Refire on weapons, these all round to the nearest .010 quite predictably unlike any other stat.) A lot of the issue I see with the grading system on other servers, and even in my own experience is that up until around C+ and B tier components, the mentality of the grading system starts to shift from usage and "strength" to rarity and potential trades/sales which blurs the line of what the grading system is supposed to represent, if that makes sense. It can represent both, but I think any kind of a grading system should be pretty clear on what exactly it means when you see a letter grade next to a part.
You definitely have a case of diminishing returns with parts in this game as well, so I guess it's natural for that overlap, but I think having some way of showing the grade as it relates to performance is better than rarity, but it could be included in the same grade letter as a percentile, or something.
... Though then what percentile do you reference? The overall probability of that stat across all the averages on the server? Or percentile based on simulated probability of getting a drop in that range? Because if we ran simulations of multiple hundreds of thousands/millions of rolls, you'd get a different average every time; Enough to skew the grading system or even percentile in a significant enough way as there is no absolute cap (aside from zero, like I mentioned) to stats.
Makes my head hurt just thinking about it haha.
In my opinion it should be based off of a solid single source metric. Say the pre-RE'd parts minimums listed in the RE Guide on the guides sub-forum, which is I think pretty close to the metrics from live. Maybe grade it as Red tier for stuff under the described minimums to RE, Yellow for On Number, And Green for what my guild collectively calls a "Cherry" part. If it wasn't a headache it might be worth it to color code them based on each statistic the part had; coloring each as checked by the agreed on literature master list. But I understand that would be a HUGE undertaking.