Ranged Physical Damage Comparisons

[Magus]

That wiki article, as most wiki articles tend to be, is stocked with sources.

[Guest Graion@Work]

Huh, I overlooked the drama in this topic prior. But yeah, truncated graphs are bad and Morpheus, you did the first personal attack here. This is frankly, low from you. Just a few citations backing that claim up.

Misleading Advertising: The Truncated Graph - SGR Law

Truncating Bar Graphs Persistently Misleads Viewers - ScienceDirect

How Graph Misrepresents Data – Cases and Tools in Biotechnology Management (pressbooks.pub)

Truncating the Y-Axis: Threat or Menace? (northwestern.edu)

If the difference is huge, then there's no problem demonstarting it both truncated and untruncated form and there should be no reason to perceive such a request offensive.

Yes, there are commonplace scenarios in the corporate world, when truncating axis makes more sense, because it simplifies the presentation (I've seen similar examples where the difference between 99.99% and 99.9% uptime was relevant and yeah, truncating made sense to ensure we grasp our own state). It still meant though that my then-manager who also looked at those graphs completely understood what does uptime even means.

This isn't about statistics, or how (much) people work with it, actually, this is moreso psychology and the way of how the decision-making behaviour can be altered. (And since we already threw out appealing to authorities; my first diploma thesis research topic was about the human decision-making behaviour, until I dropped the topic after all the research finished because my draft ended up featuring a lot more psychology than actual IT, which I was supposed to write).

Back to this scenario - I actually don't understand this graph. I haven't played ToB nor lategame IWDEE for years now and I don't actually remember what damage numbers I generated back then.

[morpheus562]

6 minutes ago, Magus said:

That wiki article, as most wiki articles tend to be, is stocked with sources.

And literally from the source itself: This is not surprising given that a restricted y-axis highlights what most researchers are testing for, mean differences, more clearly than a graph with an unrestricted y-axis.

Testing above is to more easily determine the DIFFERENCE between how different classes perform and a restricted y-axis graph, as shown, allows one to more clearly view the data.

Additionally, the article sourced for truncated graphs was based on psychology specifically not statistics as a whole, and it should be noted that truncated graphs were still the predominate type of graph being used: based on our casual observations of journal articles, scientific posters, and presentations at conferences, bar- and line-graphs with restricted y-axes are still the predominant ways in which psychological findings are presented: Across all empirical articles published in 2017 in this journal, 41 used at least one bar- or line-graph with restricted y-axes to present their findings, while only 18 used unrestricted y-axes.

[lynx]

Nobody has been doubting the intention, just your engineerial approach to perception and pigheadedness.

To lighten the mood, here's a German gem I found today, after it has been online for at least two years, from a reputable institute at that:

Spoiler

 

[morpheus562]

58 minutes ago, Guest Graion@Work said:

Huh, I overlooked the drama in this topic prior. But yeah, truncated graphs are bad and Morpheus, you did the first personal attack here. This is frankly, low from you. Just a few citations backing that claim up.

No, it was @lynx who made the first personal attack with their appeal to authority comment. And frankly, out the gate I was accused of misrepresenting the data and "how to lie with statistics". Neither is good form. Integrity matters, and my integrity was brought into question without any reasonable approach.

Edited November 21, 2022 by morpheus562

[polytope]

I don't see the issue with graph axes starting at non-zero values, it's the case for pretty much every stock chart, and I've never seen weather forecasts in Kelvin.

19 hours ago, Magus said:

As for mitigating kensai, "ken" stands for "sword", not dagger nor axe. I think a reasonable lore-friendly tweak would be to limit kensai to 2 pips in non-sword specialities.

Anyway, archers are still more versatile with their variety of arrows.

I fear that with HLAs they'll still be the best ranged DPS dealers, specialization vs grandmastery costs them 3 damage per hit (and thus at lvl 12 and thereafter they'd still be doing more damage per hit than a regular fighter with weapon GM), but no reduction in APR while using GWW. I'd prefer to restrict their damage bonuses to melee weapons (opcode #285 works, but seems unused in game besides on TorGal).

Archers have two other major advantages, their THAC0 in the early game usually surpasses any other character, due to dexterity (even if not an elf) and the stacked bonuses of magical arrows fired from a magical bow, and their called shot special ability is really OP, it grants extra damage at lvl 16, but even before that penalizes the target's THAC0, saves vs spell and eventually strength. Archers do good teamwork with both invokers and avenger druids i.e. ensuring enemies get stuck in webs and effectively turning any Chromatic Orb into a Finger of Death.

[suy]

From my side, as "responsible" of the axis being like it is, I'll file an issue to look at this to see how to improve it. On occasions I thought of controlling the values of the axes (like the "ticks" or other stuff), or provide rubber-band zooming, etc. I've personally liked the behavior that it has right now, where I basically have to fiddle very little with the axes to make it display something meaningful.

In fairness, most of the times the lowest value that it's gonna show with the current implementation it's gonna be 0. Only in extreme cases like here is gonna appear to be "zoomed in" to the upper values. This is the worst offender in the original post that I did about 4 years ago:

I 100% honestly can't tell if that is misleading to people, or it is fine enough. I look at lot at the numbers in the points, and it's kinda obvious to me that it's zoomed in to see the values better (otherwise they are stacked on top of each other much more often). So it is low priority for me to add the feature, but I'll definitely consider it for the benefit of others and to avoid any controversy.

[lynx]

If you want to improve readability, I suggest removing clutter (so called chart junk) first.

If you have all the data points labelled, there's no need for the vertical axis labels. But that's minor. It'd be much cleaner the other way around, without the numerous labels — people supposedly know how to read charts. Either way many ticks could go.

If the labels remain, I suggest at least rounding. I guess the underlying library just sucks and you can't improve placement (overlap, cutoff)? And you could only draw them when the slope changes or there is an intersection.

I'd also suggest a bit thicker connecting lines, so when there's more of them, it's easier to match the colors with the legend, since it's more likely similar ones will get used.

The legend could also use more spacing. Personally I prefer vertical ones, since all the markers are then aligned and easier to compare.

As for the baseline: like I said, I'm not sure what the best would be. Perhaps 0 (for relational integrity), but perhaps not, it's combat after all. So maybe fists at STR 24, which could have respectable damage in theory, but I guess crappy APR and to hit chances that likely lead back to 0.