Open Ambushing - New Data

[Aequis]

I added the results of my testing to the saved post on the wiki

https://gswiki.play.net/mediawiki/index.php/Open_ambushing

tldr - You need to have approximately 30 ranks of Ambush to achieve maximum accuracy no matter how many ranks of CM you have. Even a lance can pop a head 75% of the time.

EDIT - 24 ranks of Ambush is the magic number

[Goat]

I'm going to read that carefully, but the tl;dr I'm getting is "get 30 ambush ranks." Thanks.

Edit: FWIW, 101 ranks of CM and 151 ranks CM, with 0 rank ambush each time, both yielded right around 50% headshot rate with a lance against legged tritons. (My testing was not especially careful, and my 50% assertion is based on old memories.)

[Aequis]

I'm going to read that carefully, but the tl;dr I'm getting is "get 30 ambush ranks." Thanks.

Edit: FWIW, 101 ranks of CM and 151 ranks CM, with 0 rank ambush each time, both yielded right around 50% headshot rate with a lance against legged tritons. (My testing was not especially careful, and my 50% assertion is based on old memories.)

That's broadly consistent with my testing. I project that the difference between 101 CM and 151 CM would be 1-2% accuracy, so you would need to run a lot of careful tests to detect that difference. That's interesting though, since it suggests that a prone foe provides no bonus to accuracy.

[Jace Solo]

What I'm getting is:

30 ranks to Ambush is the best.

At what point do your CMan ranks out way any benefit?
OR at cap with 303 do you still need the 30 ambush to get that accuracy?

I read the whole thing, I just didn't feel if that was conclusively stated.

[Cereal Killer]

Based on his findings, he's claiming that the 30 ranks of Ambush are necessary for everyone. I would argue for 24 ranks, which is the point at which your ambush skill hits 102 (I've seen it referenced in posts that it is based on skill, not ranks).

[Jace Solo]

Testing!

Also, noticed at the bottom of the OP it says "no matter how many ranks of CM you have."

I suppose it would be nice to know where the diminishing returns really start to drop off. Like, how much per rank, but I know that'd be a slow process. 300 swings per rank through 30, or even 40, ranks.

[Aequis]

Based on his findings, he's claiming that the 30 ranks of Ambush are necessary for everyone. I would argue for 24 ranks, which is the point at which your ambush skill hits 102 (I've seen it referenced in posts that it is based on skill, not ranks).


It crossed my mind that it could be 24 ranks. That would certainly make a lot of sense. However, if you compare the improvement in accuracy from 0 ranks to 15 ranks with the improvement from 15 ranks to 30 ranks, it seems like they are about equal, assuming the difference between me at Madmountain at 30 ranks was due to my increased CM.

An alternative would be that it is purely bonus based and that the diminishing returns are not particularly severe for the first 20 or so points of bonus beyond 100.

Only way to know for sure is for someone to do a test at 24 ranks and see if it's within a percentage point or two of the results at 30 ranks. Maybe I'll do it when I'm down migrating out of the 50 rank build.

[Aequis]

Testing!

Also, noticed at the bottom of the OP it says "no matter how many ranks of CM you have."

I suppose it would be nice to know where the diminishing returns really start to drop off. Like, how much per rank, but I know that'd be a slow process. 300 swings per rank through 30, or even 40, ranks.

Yeah, that's why I shot for 300 swings on the 50 rank test. Madmountain's data suggested that the improvement would be small and I wanted to get the most reliable results I could manage given my attention span.

[Mohrgan]

Regarding the last point in your piece:

"One more thing... And this is based on anecdotal observations because I didn't think to track it until I was almost done. Particularly in my second and third tests I noticed that a significant number of misses were occurring on certain subsequent swings against the same target. Specifically, I noticed that the chance of successfully hitting the head was extremely low (perhaps even zero) if a previous attack had inflicted a non-fatal rank three head critical. There was no messaging that the head was missing; it's just that follow up attacks would not hit the head. For my part, if I could confirm that the rank three head critical was fouling subsequent aimed shots, I would change my aiming location in that situations."


I've for years found this to be the case (anecdotally, i've never collected data). I always rotate my aiming target if I hit but didn't death crit (aim head, aim neck, aim eye, or what have you), because the second shot almost always misses after you failed to crit kill and seems to hit somewhere else at random. For creatures where I'm collecting skins, varying up the target is a must if you didn't one shot, because the quasi-random wonky second shot is liable to ruin your skin, but if you pick a different body part to target, you keep control of what you are damaging.

There's got to be some wild factor in there that is responsible, just to screw with us. Because, you know, crit randomization by itself doesn't already screw with us enough.

[Roblar]

Specifically, I noticed that the chance of successfully hitting the head was extremely low (perhaps even zero) if a previous attack had inflicted a non-fatal rank three head critical.


I always rotate my aiming target if I hit but didn't death crit (aim head, aim neck, aim eye, or what have you), because the second shot almost always misses after you failed to crit kill and seems to hit somewhere else at random.

There is a reason for that last part he speculated on....aiming for a body part that has suffered a rank 3 wound without killing target will default the next swing at same location. It is not random and there is not a low chance, there is no chance of hitting same spot again.

So a non-fatal rank 3 will always try to target the chest for instance when using aiming/amush verb on a target's head that is already rank 3. I'm not sure of other body parts but I expect most, or all, move on to the chest.

[Gnomad]

Yup. If you hunt with a crushing weapon and get the "skull cracks in several places" crit, time to aim for the neck.

[Warriorbird]

I added the results of my testing to the saved post on the wiki

https://gswiki.play.net/mediawiki/index.php/Open_ambushing

tldr - You need to have approximately 30 ranks of Ambush to achieve maximum accuracy no matter how many ranks of CM you have. Even a lance can pop a head 75% of the time.

I suddenly feel the urge to train ambush.

[Aequis]

So, having purchased a new weapon that I need to bond with, I'm poised to run some more tests once my migration is over with. Other than 24 ranks of ambush, is there something I should be looking at?

[Aequis]

This is last call for suggestions regarding open ambush testing. I am currently migrating down to 24 ranks of ambush to test that theory. It should be done in a few days. I intend to do a test aiming at heads with 24 ranks of ambush and see how that compares to the current data sets. If (when?) that proves to be less effective than 30 ranks, I will also try a test aiming at eyes with 30 ranks which should be interesting.

[Aequis]

That was it. Another test at 30 ranks targeting heads and discounting hits to the chest in situations where a rank three head wound has already been inflicted.

[Aequis]

I updated the wiki with this morning's test. Those of you who thought that 24 ranks of Ambush made sense are correct. There is a statistically insignificant difference between 24 and 30 ranks. Since 24 ranks proved so effective, I skipped the part where I was going to redo the 30 rank test to account for default targeting when the location aimed at already has a rank three wound.

Tune in next time for accuracy percentages when aiming at the eyes.

[Aequis]

My preliminary tests for ambushing eyes were extremely discouraging. Again, I am using a lance, and I recognize that it would be ridiculous to have good accuracy trying to lance something in the eye. It appears that the accuracy is around 60%. But figure there's a 5% margin of error at least. I only scored 50 hits before it became apparent to me that there was no way it would be more efficient for me to aim at eyes even after factoring in the propensity of a lance to deal puncture damage. This is in part because of the lower overall accuracy and in part because of the unfavorable crushing damage chart eyes.

[Carsyn]

Excellent work, we thank you!

[Aequis]

The question remains whether ambush ranks above and beyond 24 would improve accuracy while aiming at eyes. Between my tests and MadMountain's tests it seems clear that heads and legs share the same target dependent accuracy factor and that 24 ranks is the ambush soft cap for those locations. Maybe in May I'll burn my free fixskills trying to find out.

[gilchristr]

did you know there is a fixskill coming in 2 weeks?

[Androidpk]

I know ambushing from hiding with a knife works really well but how about open ambush?

[Whirlin]

Ambushing only has additional critical weighting and a stance pushdown component when ambushing from hiding.

Otherwise, daggers are actually 62.5% piercing, 37.5% slashing damage type, so, eye shots are good, 62.5% of the time.

[gilchristr]

"so, eye shots are good, 62.5% of the time"

Unlike puncture mixed with crush (where if you go for the head you might get puncture, and if you go for the eye you might get crush), slash to the eye can still get the guaranteed crit kill (rank 5 slash to the eye is a kill).

[Menos]

Out of interest, I pulled up a couple of my duskruin runs. Looking at aiming a lance for the head of a standing target with 100 levels, 202 CM and 101 ambush. Out of 55 ambushes I got 1 fumble, 3 EBPs, 5 missed location, 46 hits to the head. So excluding the EBPs, we can call that ~10% missed, <2% fumble, ~89% hits to the head. Oddly 4 of the missed shots were abdomen and one was a hand.

Those numbers jive pretty well with my experience in running through zillions of arena battles. While 24 ranks of ambush may well get you to 75% success, I would argue that going from 1/4 missed shots to ~1/10 missed shots is a pretty huge difference. I feel like Menos the warrior with 202/202 ranks did slightly better than that as well, but I am not prepared to dig through my logs from 2003 to prove that.

[Menos]

Oh, that is with an elf with +35 dex bonus and +30 strength bonus.

[Menos]

I tried a bigger data set (109 shots), numbers didn't change much. (87% hits 12% missed location and 1% fumbles)

I may get around to annoying O until he writes a script to check all 500 some odd duskruin runs where I was using my ambush tactics and not just spamming 1615. I suspect the numbers will be similar though. That, or have him check the last 10 years of OTF logs.