Research:Open ambushing

The official GemStone IV encyclopedia.
Jump to: navigation, search

Open Ambushing (ambushing without hiding)

Ambushing without hiding is modified by ranks in Combat Maneuvers and Ambush skill only (barring of course different players having different STATS that could affect this skill).

The more Combat Maneuvers ranks you get, the less Ambush ranks you will need to see the relative maximum benefit of the skill. Even at level 50 with 2x CM ranks you will start to hit a plateau of diminishing results with the Ambush skill on open aiming, even with large weapons.

That plateau is at around 25 to 30 ranks for a level 60ish player with 2x CM. So it's possible that if you are level 25, you might see benefit of up to say 50 ranks, or if you are 75, maybe only up to say 20 ranks.

The last time I did the math, when I was about level 75, I could achieve a say 75% success rate vs legs with a maul with 2x CM and 25 ranks of ambush, and an 80% with 2x CM and 30 ranks of ambush. It then took nearly 30 more ranks of ambush for me to notice and even 1% increase in accuracy.

How much ambush helps is a diminishing effect that seems linked to how close to the designed maximum accuracy you get with CM training, via basically gaining levels and continuing to double train in the skill.

Note: The percentages in this document are not mathematically correct, but are rather designed to make a point. While not accurate, they convey how the ambush-CM relationship works, and give the rough plateaus as one player has experienced them.

If I ever locate my exact figures, I'll post them in a table below.

Saved post regarding this matter, in response to whether or not a plateau exists

Yes, I'm 100% certain that ambush training, used in conjunction with open ambushing has a spot where you start getting diminishing returns.

The problem is, that level at which you get diminishing returns depends on what train you are (because of CM ranks, having Wspec, and having full bond). You can't just say "30 ranks is where the plateau is".

Something like "A level 75 warrior with 2x CM, full wspec and full bond, will only see realistic additional benefit to aiming a claidhmore with 30 ranks of ambush or less" is more realistic.

But it might be that a level 30 warrior sees trackable benefit for up to 55 ranks of ambush.

From my own data with myself, which I just blew a fixskill to test:

Level 80, 2x cm, 15 ranks ambush, aiming at left leg with claidhmore 50 times at earth elementals = .62

Level 80, 30 ambush ranks, aiming at left leg with claidhmore 100 times at earth elementals = .73

Level 80, 50 ambush ranks, aiming at left leg with claidhmore 100 times at earth elementals = .74

The TPs from 15 to 30 made sense to me. The TPs from 30 to 50 did not.

Also note that only hits that successfully struck the critters were used. The "deflected off their side" BS that only happens on elementals wasn't counted in the swing count.

Also interestingly, with a maul, I had nearly or above 10% higher accuracy per category. (Without being bonded to it).

2015 Confirmatory Testing Preface

I generated three data sets ambushing heads with my lance in orc warcamps. I am level 92 with 216 CM ranks using enhancives. Due to the relative ease of killing quickly in warcamps, I was able to more than double the number of data points used by Madmountain in his tests. As an aside, I can't imagine the frustration of trying to do this with earth elementals.

I am also fully specialized in the lance including full paladin bonding, however I do not share Madmountain's belief that these factors play any role in accuracy. I am agnostic regarding the impact of stat bonuses, in part because it would be so difficult to test and in part because it seems like it would have an extremely small impact compared to other factors.

For what follows, I define a "hit" as an attack that gets through E/B/P to the AS/DS roll or an attack that fails to find an opening for a strike.

2015 Confirmatory Testing Data

With 0 ranks of Ambush I scored 218 hits. Of those hits, 123 struck the head, 85 struck another body location, and 10 failed to find an opening for a strike. This yields approximately 56% accuracy and approximately 5% complete failure.

With 30 ranks of Ambush I scored 226 hits. Of those hits, 170 struck the head, 55 struck another body location, and 1 failed to find an opening for a strike. This yields approximately 75% accuracy.

With 50 ranks of Ambush I scored 330 hits. Of those hits, 255 struck the head, 74 struck another body location, and 1 failed to find an opening for a strike. This yields approximately 77% accuracy.

Please note that with respect to the single failures on the second and third tests I believe I was prone or otherwise under the effect of some kind of debuff when I swung.

Analysis and Conclusion

For the purpose of comparing my data with Madmountain's data, I will assume that our numbers are equally accurate even though he used far fewer swings to arrive at his percentages.

It seems obvious that there is a soft cap mechanic involved in open ambushing accuracy. I suspect, based on Madmountain's remarks about his increased accuracy using a maul, that the theoretical maximum accuracy is tied to the weapon base.

Unlike Madmountain, I do not believe that CM and Ambush are interchangeable in the formula. I believe that there are two separate soft caps, one for each skill, and that the modified totals are then added together to arrive at actual accuracy. That is, the portion of accuracy that comes from CM cannot be replaced by any amount of Ambush and vice versa.

In Madmountain's first test he had 162 ranks of CM and 15 ranks of Ambush for a combined 187 relevant ranks. His accuracy was 62%. In my first test I had 216 ranks of CM and 0 ranks of Ambush. My accuracy was 56%.

In Madmountain's second test he went up to 30 ranks of Ambush for a combined 192 relevant ranks. His accuracy was 73%. I also went up to 30 ranks of Ambush for a combined 246 relevant ranks. My accuracy was 75%.

In both third tests the move from 30 to 50 ranks of Ambush produced an increase in accuracy of 1-2%.

The first tests show conclusively that some amount of Ambush is indispensable to achieving maximum accuracy. 54 additional ranks of CM gave me 6% less accuracy compared to Madmountain's 15 ranks of Ambush.

The second and third tests show the extent of the diminishing returns. In the second test my 54 additional ranks of CM provided a 2% increase in accuracy over Madmountain. In the third tests we both received comparatively small increases in accuracy for additional ranks in Ambush beyond 30. The extreme nature of the diminishing returns suggest logarithmic function comparable to Redux.

Additional Research

There was some discussion on the PC that 24 ranks of Ambush might be the magic number. This morning I decided to try it out. While my test was cut short by a grizzled sorcerer, I still got a statistically significant number of attacks in.

There are some slight modifications to my skills and methodology since my last tests that bear mentioning. For this test I had an enhanced 214 CM ranks (two less than my previous tests) and 24 Ambush ranks. Once again I was aiming at the heads of Grimswarm orcs with my lance.

I also took into consideration the widely held belief that aiming at the head when the head already has a rank three wound causes a successful attack to default to the chest. In this situation there is no way to distinguish between a successful aimed attack defaulting to the chest and a miss that happens to hit the chest. To avoid this problem I discounted all subsequent hits on a foe that already had a rank three head wound.

Before getting imploded I scored 210 hits (see above for definition of 'hit' in this context). Out of those hits, 159 struck the head and 51 struck some other location. None of my swings failed to find an opening for a strike. This yields 75.7% accuracy. I attribute the slight increase in accuracy to the more careful treatment of default targeting in the case of existing rank three wounds. I estimate that a repeat of my original test with 216 CM ranks and 30 Ambush ranks would probably yield a 76% accuracy instead of the 75% accuracy I reported previously.

Based on today's test I feel extremely confident that 24 ranks of Ambush is sufficient to reach the soft cap on aiming accuracy derived from Ambush training. In my opinion, this finding also strongly suggests that 24 ranks of CM is sufficient to reach the soft cap on aiming accuracy derived from CM training. I stand by my position that aiming accuracy is the sum of two separate equations, one for each skill where each skill has its own independent soft cap.


  • Madmountan's brain and personal calculations.
  • Also Aequis/Nylis 2015 confirmatory testing