Ambush verb's affect on critical rank

[Terzl]

I've been recording data on the AMBUSH verb (when aimed and used from hiding by a rogue) and how it modifies final critical rank. I wanted to share an observation.

A commonly held belief is that AMBUSH'ing from hiding adds some amount of phantom damage (based on training, stats, race, or whatever) to the raw damage calculation. When dividing the increased raw damage by some critical divisor, you end up with a higher critical rank. This is also how weapon critical weighting is often explained.

If this was true, then AMBUSH should have a greater effect against creatures with softer armor because their critical divisors are lower (and after division, the final critical rank would be higher).

For example, assuming the mechanics work as stated and my training/stats give me +22 phantom damage when AMBUSH'ing from hiding:

- Against robes this "extra" +22 phantom damage should increase the final critical rank by 4 ranks (+22/5).
- Against plate it should increase the critical rank by 2 (+22/11).

My data shows something very different.

Whether I'm AMBUSH'ing something in robes, chain or plate, the average increase in critical rank that I experience is nearly identical across all armor sub-groups (read: divisors).

I've never tested weapon critical weighting, but if it works as believed (i.e. bt adding phantom damage) then AMBUSH'ing from hiding is different and mechanically superior.

Has anyone else ever done any testing on this or have any comments?

[thefarmer]

Randomization

[Danical]

I'd like to see the raw game output, skills, and stats.

[Terzl]

Randomization
I took the average increase in critical rank on each data set. Unless I'm way off, that should offset whatever randomization takes place (as long as critical randomization is uniform across armor sub-groups). I did not take the average critical rank itself.

I'd like to see the raw game output, skills, and stats.
I have two files, one is the raw game output of me ambushing with non-crit weighted weapons against different armor groups, the other is the excel file with my calculations. I'll try to clean them up and post them if you think it'd be worthwhile.

[Danical]

Please do so!

[Terzl]

A small sample of data that illustrates my methodology. Feel free to shoot holes in my thinking.

Creature: Hill Troll

Level 45 rogue, from hiding, non-weighted weapon
2x Ambush, 2x OHE



Broadsword vs. Plate
Endroll DF Raw Crit Rank Actual Phantom Crit Inc.

231 0.2 26.2 2.4 4 +22 +2
163 0.2 12.6 1.1 3 +22 +2
215 0.2 23 2.1 6 +44 +4
213 0.2 22.6 2.1 2 +0 +0
222 0.2 24.4 2.2 6 +44 +4
192 0.2 18.4 1.7 5 +44 +4
155 0.2 11 1.0 4 +33 +3
138 0.2 7.6 0.7 0 +0 +0


Broadsword vs. Leather
Endroll DF Raw Crit Rank Actual Phantom Crit Inc.

228 0.25 32 5.3 9 +24 +4
221 0.25 30.25 5.0 6 +12 +2
233 0.25 33.25 5.5 5 +0 +0
215 0.25 28.75 4.8 6 +12 +2
201 0.25 25.25 4.2 7 +18 +3
267 0.25 41.75 7.0 8 +12 +2
286 0.25 46.5 7.8 7 +0 +0

Endroll is straight from game output.
DF is damage factor of broadsword vs. creature armor.
Raw is calculated raw damage.
Crit Rank is the calculated critical rank based on critical type and raw damage.
Actual is the critical rank displayed in game output.
Phantom is the amount of phantom damage that would have been added for this critical to display.
Crit Inc is the actual increase in critical rank from calculated to actual game output.

Capturing the data was tricky because the broadsword dealt two types of damage, I had to make sure I discerned the critical TYPE based on the messaging to record the proper critical level. I need to learn to write a script. Ugh.
I'm now swinging a mace so it will be easier to record one type of damage. May be worthwhile to re-tackle this project if I'm not proved wrong right away.

[Danical]

Most trolls have crit padding which will randomize past somewhat.

[Terzl]

Most trolls have crit padding which will randomize past somewhat.
Hah! I chose a creature that wore different types of armor thinking it would eliminate variance.

Can you recommend a critter (up to 45ish) that would be a good candidate to pull data from?

[Drew]

arch wights in the landing wear half plate, and in the icemule wear nothing or maybe leathers.

[Riltus]

1. The full weapon crit weighting is added as phantom damage to each attack.

2. Randomization results are based on the critical rank regardless of armor.

3. If your character has 44 points of weighting with aimed ambush attacks from hiding, all attacks against robe wearing targets will be rank 9 crits before randomization. They will also all be rank 9 crits against leather with as little as 10 raw damage. Rank 9s will randomize with an equal distribution to ranks 5, 6, 7, 8 or 9 with an average result of 7.

Against scale, the raw damage required for a rank 9 is 19 with 44 points of weighting. With a minimum of 19 raw damage the average would still be 7.

Even chain (37 raw damage) and plate (55 raw damage) would be a guaranteed rank 9 crit with 44 phantom weighting from ambush. Therefore, with 55 raw damage and 44 points of weighting the results would be the same for all armor sub-groups, an average critical rank of 7.

Mark

[Latrinsorm]

Broadsword vs. Plate
Endroll DF Raw Crit Rank Actual Phantom Crit Inc.

231 0.2 26.2 2.4 4 +22 +2
163 0.2 12.6 1.1 3 +22 +2
215 0.2 23 2.1 6 +44 +4
213 0.2 22.6 2.1 2 +0 +0
222 0.2 24.4 2.2 6 +44 +4
192 0.2 18.4 1.7 5 +44 +4
155 0.2 11 1.0 4 +33 +3
138 0.2 7.6 0.7 0 +0 +0


Broadsword vs. Leather
Endroll DF Raw Crit Rank Actual Phantom Crit Inc.

228 0.25 32 5.3 9 +24 +4
221 0.25 30.25 5.0 6 +12 +2
233 0.25 33.25 5.5 5 +0 +0
215 0.25 28.75 4.8 6 +12 +2
201 0.25 25.25 4.2 7 +18 +3
267 0.25 41.75 7.0 8 +12 +2
286 0.25 46.5 7.8 7 +0 +0Your "Actual" column is somewhat problematic, because your observed crit rank almost always represents a range of possible values. Consider the fourth endroll of broadsword vs. full plate: you observed a rank 2, but this could be the result of anything from a rank 2 max to a rank 4 max, inclusive. This means that your phantom damage could have been anywhere from 0 to 31 (giving you a total of 54 damage for the purposes of crit calculation).

God bless you if you try and search it, but there has been research done on ambush before. Rather than using an averaging methodology, the researcher looked for "there can be no more than" and "there can be no less than" hits and scrutinized where they overlapped. My recording was regrettably poor in that era, but I am confident it was Mark/SPYRIDONM (who posts here) that did this work.

Also worth considering is the interaction of weighting from Dex bonus (as researched by GILCHRISTR) with other sorts of weighting (potentially including AMBUSH). The aforementioned Mark is to my knowledge the only person to look at this interaction - if I may summarize, it's a pain in the ass, but not an unmanageable one.

[Terzl]

3. If your character has 44 points of weighting with aimed ambush attacks from hiding, all attacks against robe wearing targets will be rank 9 crits before randomization. They will also all be rank 9 crits against leather with as little as 10 raw damage. Rank 9s will randomize with an equal distribution to ranks 5, 6, 7, 8 or 9 with an average result of 7.
Mark,

So my chart was missing a column that said "Maximum Potential Crit Rank" for each swing given the "phantom damage" mechanics assumption and then I would have saw how the randomization effected the final critical output.

Your explanation fits my numbers perfectly. Thanks for the clarification.

Vin

[Terzl]

Your "Actual" column is somewhat problematic, because your observed crit rank almost always represents a range of possible values.

You're right.

I needed a column that said "Maximum Potential Crit Rank" - given phantom damage assumption of 44, or whatever. Suddenly, all the observed critical values make sense given that new column + the explanation of randomization.