Ok.

Peistes

Consideration: Low number of crits, so no guarantee of hitting true limits, so allowing for that.

Lvl 73-74, max crit of 511; 4.0 should be 487. Minimum recorded was 488 and 489 in the two parses. If 488 is min, max should be 512.

Lvl 75-76, max crit is 507; 4.0 should be 483. Matches observed.

Extract 8% crit bonus

73-74: If 4.0 crit was 487, this would be 451, which is not divisible by 4. If 4.0 is 488, this is 452. Probable 452.

75-76: With 4.0 of 483, this is 448. Divisible by 4.

Base damage:

73-74: 452/4 = 113

75-76: 448/4 = 112

Lvl 73 with no +att had the same max damage as lvl 74, so 73 is probably not at true max pDif. With +att, its max went higher. Similarly with the others... I don't see anything to indicate that the observed damages were not at capped attack, just varying due to relatively low sample size. Will just use the lower of the two mins, and the higher of the two maxes.

First, factor out the 1.05 spread:

Lvl Max /1.05
73: 474 452 (note: == 4.0)
74: 466 444
75: 455 434
76: 449 428

Note that those are the lowest that the max pDif can be; it's possible for it to be higher.

Now look at the base pDif range

Lvl Range Diff pDif Range
73: 368 - 452 84 0.743
74: 360 - 444 84 0.743
75: 354 - 434 80 0.714
76: 349 - 428 79 0.705

Given that the max pDif value can potentially be higher than listed, we should tend towards the maximim pDif range values as being closest to correct. That points us towards 0.75 as the likeliest value for the full range.

That gives us the trivial look at things. Next, pulling up the details...

Lvl 73 -- The parse with +att pushed towards the top end of 4.0 * 1.05, which implies the max pDif possibly went slightly above 4.0 in order to get a fair numbers of samples in that range, but at the very least hit exactly 4.0. The min damage on that was 371, with other observed min damage values nearby that are sufficienctly close-packed that it seems unlikely for true min to be more than a point or two below it. 452 - 371 is an 81 point difference; much closer to the level 75-76 data.

The parse without +att also had a significant gap between its lowest value (368) and its second-lowest value (377), which disturbs me, as it feels out of place. However with the low number of samples, there's the risk of such gaps.

Lvl 74 -- The parse without any +att had a min-max difference of 79. Given that just about every single damage point was hit between the two extremes, the lack of any values between 360 (min observed in the +att parse) and 365 disturbs me. 79 puts it in the same range as the level 75-76 parses.

Lvl 75 -- Did not have any oddities at either extreme of either parse. Min/max on both parses was also almost the same.

Lvl 76 -- Parse with +att was a bit sparse on the upper end, where its top two values were 445 and 441. The parse without +att had several values between 443 and 449. Sparsity is somewhat acceptable due to the low number of samples.

If we put aside those values which seem problematic, we get:

Now look at the base pDif range

Lvl Range Diff pDif Range
73: 371 - 452 81 0.717
74: 365 - 444 84 0.699
75: 354 - 434 80 0.714
76: 349 - 428 79 0.705

Which puts it in line with a 0.72 max range.

Now the question is, are those problematic data points 'real'? If they are, then that implies a full pDif range of 0.75, though also giving us an uncomfortable gap between the 73-74 data and the 75-76 data (though it could be explained by low sample sizes). If they are not then we have a nice pDif range that is consistent across levels, but we need to find out how those anomolous values could come about.

Vultures

Lvl 83

Base damage: 114

Min outlier: 287. Otherwise, mins of 298 and 301.

Max 403 => 384

Diff: 86, or 97 with outlier

Range: 0.754, or 0.851 with outlier

Lvl 84

Base damage: 113

Min outlier: 289. Otherwise, mins of 294 and 296.

Max 394 => 376

Diff: 82, or 87 with outlier

Range: 0.726, or 0.770 with outlier

Both of these lend themselves towards a pDif range of 0.75, though they also have outlier issues.

Expected cRatio values, if we assume +0.06 per level:

Peiste

73: 3.56

74: 3.50

75: 3.44

76: 3.38

Vulture

83: 2.96

84: 2.90

Expected min pDif and min damage values for -0.36 and -0.375 offsets:

Peiste Min -0.36 Min Dmg Min -0.375 Min Dmg Obs.Min Outlier
73: 3.20 361 3.185 359 371 368
74: 3.14 354 3.125 353 365
75: 3.08 344 3.065 343 354
76: 3.02 338 3.005 336 349
Vulture
83: 2.60 296 2.585 294 298 287
84: 2.54 287 2.525 285 294 289

And these don't match up. Predicted for the peistes is significantly lower than observed (~10 points), though predicted for vultures is not too far off.

On the other hand, we can show that the 289 outlier on lvl 84 vultures is well within expected range. With the difference in level correction plus the difference in base damage, a 9 point difference in the min between the 83 and 84 vultures should be expected. The outlier 289 is 9 points lower than the non-outlier 298 from the lvl 83 vultures. The 287 outlier on level 83 vultures is still untenable, though.

We can use the observed mins, and an assumed pDif offset (either -0.36 or -0.375) to back-calculate an estimated level correction for each of these.

Peiste Obs.Min Outlier pDif Lvl Corr. Raw Lvl Corr. 0.36 Lvl Corr. 0.375
73: 371 3.2832 0.049 0.0632 0.0638
73: 368 3.2566 0.048 0.0623 0.0628
74: 365 3.2301 0.049 0.0636 0.0642
75: 354 3.1607 0.048 0.0634 0.0640
76: 349 3.1161 0.049 0.0642 0.0648
Vulture
83: 298 2.6140 0.038 0.0609 0.0618
83: 287 2.5175 0.032 0.0548 0.0558
84: 289 2.5575 0.037 0.0612 0.0622

And here we come across something quite peculiar. When doing the calculations I forgot to add in the assumed pDif offset before dividing by the level correction on one of the numbers, and ended up with something very close to our normal 0.05, remaining stable across those levels. That's the Lvl Corr. Raw column. However that only worked for the Peistes.

I would be most curious to see if that change occurred at level 79 mobs (-20 levels of correction). That could have some interesting implications.

On the other hand, most of the other numbers have problems. Adding the assumed pDif offset yields a level correction value that's too high, for the peistes, and none of the numbers for vultures returned a decent level correction value -- the raw value is too low for anything, and the +pDif value is higher than it should be.

The vultures could get a 0.05 level correction if the pDif offset was -0.20 for the min, but that's heavily skewed towards the positive side, given the full pDif range.

I have a couple ideas about what could be going on, but I'm going to need more data. There's a couple strange possibilities, but I need another data axis to triangulate from.

TEST: More data sets on any/all of the same mobs, but level capped at level 90 instead of 99, while keeping capped attack. Take a rdm along for Dia III or something, if you need to.

TEST extra: More data on the peistes just to help fill out the edges.