.426 Wolf Struggle Tests


Timber Wolf

Recommended Posts

(Note: Here are links to the other tests -  .393 and 1.14)

The test scenario: A 100% stalker character wearing no clothes walking toward a wolf.  Measuring condition loss, injuries and weapon condition loss for 20 struggles with each.  All weapons were at 89% condition.  Used a macro to automate 8.3 mouse clicks per second.

595ba93612cb6_426ConditionLossbyWeaponSheet2.thumb.jpg.37869e104733dfa0dc569ef449097192.jpg

 

59445945963bd_426ConditionLossbyWeaponChart.jpg.520694a308d8ae1be9d85d2fb2767ced.jpg

595ba93145fec_426ConditionLossbyWeaponSDandAVG2.jpg.fa83a40df38e1b56d2c95ecc4991a252.jpg

 

5944594e5dfc9_426InjuriesChart.thumb.jpg.b37033175a2121719990ecf74ba3f46c.jpg

59445950baf45_426WeaponConditionsChart.thumb.jpg.8bb5e8390cebe9989dee16d3b8648bcd.jpg

 

The hatchet is now and has always been the best weapon.  What do you think about these results?

Edited by Timber Wolf
Added link to older tests.
  • Upvote 29
Link to comment
Share on other sites

You, my friend, are amazing.

 

I will now stop grabbing the knife. Thank you for the work, effort, and all the information you provide.  With every single result you post, you provide this community with some amazing information.

 

Ten

 

 

Edited by Tenasi Vol
  • Upvote 5
Link to comment
Share on other sites

It's my pleasure, @Tenasi Vol :)

 

2 minutes ago, Drifter Man said:

I think they are pretty impressive!

Did the wolf always lose? And do you have an idea how many were killed on the spot?

Thanks!  I think the stats would always show I won, but only because I didn't die.  They never died on the spot and all ran off.  

  • Upvote 1
Link to comment
Share on other sites

So, how much difference does it make wearing clothes?  I used the same testing scenario, but this time wearing clothes all at 100% (except the mittens started at 99%).  I had a total of 26% Protection (protection % is shown for each item).  I used the same 89% hatchet.

594558e86c874_426ClothingTestSheet.thumb.jpg.0081947204076cf07ea5968adb0f2394.jpg


Now let's compare between wearing clothes and not wearing clothes.

595ba99803283_426ConditionClothesandNoSheet2.thumb.jpg.da03afd006dc2cbcde324787ad2c9f57.jpg

594555cf3038a_426ClothesandNoClothesChart.jpg.0c4938102ea32105bcae5b53ece721bf.jpg595ba99408da3_426ConditionClothesandNoAVGSD2.jpg.42d95284e98a578d95ced5349d3ab83d.jpg


It looks to me as though having protection doesn't make much difference, in terms of condition loss or injuries sustained.  What do you think?

Edited by Timber Wolf
Revised Charts
  • Upvote 6
Link to comment
Share on other sites

If reduction of damage scales linearly with protection, as I would expect, the result with clothes should be about 16-17. It is possible that 20 attempts didn't capture this precisely.

Maybe that damage with clothes is also more consistent in that it prevents large condition loss as in attempt No. 14.

I'm not saying you should do more tests :) You've done more than enough. Just saying that more numbers might reveal the exact trends.

Great work!

  • Upvote 3
Link to comment
Share on other sites

Wow, thanks a lot for your efforts @Timber Wolf!

That's really some awesome data - I personally always considered the knife to be superior to the hatchet, but your data definitely changed my opinion about that.

16 hours ago, Timber Wolf said:

It looks to me as though having protection doesn't make much difference, in terms of condition loss or injuries sustained.  What do you think?

Well, I think @Drifter Man is right. The Stds of all struggle-related data in TLD are always huge and this ultra-high variability of a struggle outcome (even if you use clicking macros and have perfectly equal starting conditions every try) makes it quite likely that "smaller" n numbers give you some kind of skewed results.  

It's already great that you chose n=20 (for data in real life, that's usually a very solid starting point, at least in my field of work), but due to TLD struggle data's huge variability, one would probably need n numbers in the range of n=100+ to really see a 26% decrease. But I'm optimistic it might exist. :)

But anyway, I personally never chose my clothes to maximise the protection value in the first place and I certainly see no reason to change that. I mean even if it meant an average condition benefit of 16% average condition loss versus 22% average condition loss, you would have so many weight and temperature disadvantages in your combat clothes that it would probably never be worth wearing them anyway.^^

  • Upvote 1
Link to comment
Share on other sites

@Timber Wolf  Nice tests -- thanks for sharing.  A couple questions: 

1. Edit: Never mind, figured it out :D 

2.  What level of rest did your character have?

3.  What level of hunger did your character have?

These last two might be quite important.  Based on hints from support in your similar thread on test_branch, I did a little indirect testing being well fed and rested while looking for steps to reproduce struggle bugs.  My effectiveness during the struggle seemed noticeably better, for what it's worth.

Edited by Ruruwawa
  • Upvote 1
Link to comment
Share on other sites

14 hours ago, Drifter Man said:

Just saying that more numbers might reveal the exact trends.

This is a very good point.  20 didn't really feel like enough, but it has become my standard amount for testing.  Another round or two could reveal a wider variation of results.
 

5 hours ago, Scyzara said:

I mean even if it meant an average condition benefit of 16% average condition loss versus 22% average condition loss, you would have so many weight and temperature disadvantages in your combat clothes that it would probably never be worth wearing them anyway.^^

Excellent observation!  And I completely agree.  I wasn't expecting the protection from clothing to make a significant difference, but I had hoped it would at least be an appreciable difference.  As it is (or at least as it seems to be), a character may still be best served by removing clothing prior to an expected wolf struggle - which should not be a viable strategy, in my opinion.
 

2 hours ago, Ruruwawa said:

What level of rest did your character have?

I always attempt to remove as many variables as I can in the tests I run.  I slept until fully rested, then walked toward a wolf on the lake and lured him toward the fishing hut (on TWM).  Then I had to past 1 hour of time, in order to save the game in the exact state I wanted to run the tests.  So, I was fully rested (minus 1 hour of passed time) when I performed all of these tests.
 

2 hours ago, Ruruwawa said:

3.  What level of hunger did your character have?

  I was nearly full of both water and food.  I made a point of filling up, for the reasons mentioned above.
 

My temperature gauge was about half way, because I had to pass 1 hour of time without clothing to setup the testing scenario.  Other than that, my character was pretty much at - all systems go. :)

  • Upvote 1
Link to comment
Share on other sites

37 minutes ago, Timber Wolf said:

Then I had to past 1 hour of time, in order to save the game in the exact state I wanted to run the tests. 

I just re-read my responses and feel the need to mention something off-topic.  I still find it ridiculous that I can "pass time" and the animals do not move from their current locations.  I can pass several hours of time and all of the animals will still be in almost the exact same spot as before I passed the time.  This needs to be addressed, as it is an incredibly easy way to exploit the game.  The weather does change when passing time, so I don't see why the rest of the game world shouldn't follow suit.

I would have slept for one hour to get my character at no fatigue, except when I sleep the animals actually move to different locations - as it should be when passing time.

Edited by Timber Wolf
Link to comment
Share on other sites

2 hours ago, Timber Wolf said:

I always attempt to remove as many variables as I can in the tests I run. 

Looks close enough in my book, even with the hour of pass time -- thanks.   Anecdotally (since I didn't actually measure), I didn't notice a big difference until "in the red" for food or rest. 

Link to comment
Share on other sites

On 6/17/2017 at 9:25 AM, Timber Wolf said:

 

594555cf3038a_426ClothesandNoClothesChart.jpg.0c4938102ea32105bcae5b53ece721bf.jpg594555d216628_426ClothesandNoClothesSDandAVG2.jpg.e135fbff9426847fd74cf94f92652e8a.jpg


It looks to me as though having protection doesn't make much difference, in terms of condition loss or injuries sustained.  What do you think?

Look at instances where Condition loss > 30% and this is where protection seems to factor in.  

No protection 30% or more condition loss = 7/20 ... 40% or more = 3/20 

With protection, 30% or more condition loss = 1/20 ... 40% or more = 0/20

I am confident this dataset is sufficient sample size and should very closely match either a poisson or negative binomial distribution.  It's been a decade or so since I've dealt with getting an equation for such a dataset, but 10-15 years ago I could have done this comparison in a few minutes.  Unfortunately the muscles that are not exercised lose their strength, and I have forgotten a lot of my early career statistics.

You're correct in that it doesn't change the fact that you can reliably seek out a wolf struggle naked with intent to kill it, but the protection does offer a clear difference, and I'm fairly certain someone with more than a passing knowledge of statistics can figure out a very good approximation of the formula in 10-15 minutes based on your existing data.  Probably what would help more than more samples at an existing protection value, would be another dataset of 20 at some protection value in the middle, but I don't think that would be necessary to reverse engineer the equation used with 80-90% confidence.

I'm not sure how you change people struggling in their underwear without significantly increasing the difficulty of the early game.  The game already has a steep learning curve and high early game difficulty, that then levels off once you get a foothold.  I think the only thing really punishing is a significant risk of 90-100% condition loss with the most favorable weapon at a pretty significant click rate, and if that is the case, then what is the chance of 90-100% condition loss for a new player who hasn't found a hatchet?  Being punishing for new players is one thing, but the game favors ways of learning in-game rather than pushing people to the forums, and I think this doesn't fit the game design directions.

Wolves need to be less threatening than bears.  It's an accomplishment of progression when a player gets to the point where they're not scared of a single wolf attack because they have a struggle weapon in good condition and plenty of wound treatments.  Having that kind of step in progression is good for player experience.  Bears are already there as the "you can never struggle with them" predator, and a 3-pack of wolves is also in the same category.  I'd much favor seeing more bears / packs on maps vs. buffing single wolves (which they have done.)

I think even if you buff them, you still eventually get to the point where you can fight them naked, you just have to plan for extra recovery time... but what really prevents the recovery time when you kill a wolf and have meat to survive plenty of resting time?  Only the other afflictions that have been added like cabin fever, no way to sleep and recover condition when fully rested, parasite risk and now the scent risk of carrying home the spoils of your victory at low condition.  Hinterland has consistently added these as ways to curb players stripping down when hunting wolves vs. where the game was a year ago, and I think it's quite a lot less beneficial to fight wolves naked vs. when I first bought the game.

Edited by Concillian
Link to comment
Share on other sites

  • 2 weeks later...
On 6/18/2017 at 4:02 AM, Scyzara said:

one would probably need n numbers in the range of n=100+

Alright, let's see what that looks like!

595af47a4e945_426100StruggleTests.thumb.jpg.b0da33fd0bf745db1e49d161f17964ac.jpg

595af47b71594_426100StruggleTestsPart1.thumb.jpg.f2dd16ece26950ec645c613866773d73.jpg

595af47d1a228_426100StruggleTestsPart2.thumb.jpg.02fcf7b6bb978f22846d237cacfbb3cc.jpg

595af47de3e51_426100StruggleTestsAVGSD.thumb.jpg.d0c6b63ef3a66b539ebee6c5606e128c.jpg


While testing it occurred to me that the outcome might be affected by how long it took me to select a weapon.  So, for the last 50 tests I waited until I was almost out of time to select a weapon before beginning the struggles.  I was pleased to find it didn't seem to make any difference.

  • Upvote 7
Link to comment
Share on other sites

You must feel weary now :)

So that's an average of 18.4% condition loss with 26% protection (based on 100 tests) compared to 22% condition loss with no protection (based on 20 tests, in the OP), all with hatchet.

But I put the numbers from the first table in this topic into excel and calculated an average of 26.7% condition loss instead of 22%... can you check how you calculated that average?

If my average is correct, it would mean a reduction in average condition loss by 31% with clothes.

  • Upvote 1
Link to comment
Share on other sites

7 hours ago, Drifter Man said:

But I put the numbers from the first table in this topic into excel and calculated an average of 26.7% condition loss instead of 22%... can you check how you calculated that average?

Crap!  I should have double-checked all of my calculations.  You're right, my formula was only accounting for the first half of the results.  An error that was repeated in my Clothes vs No Clothes tests.  These sheets and charts have been revised.  Good eye, @Drifter Man!

Link to comment
Share on other sites

Great work, Timber Wolf!

At a quick glance, I get the impression that the differences between the averages you find are considerable from the sequence graphs, but insignificant from the column plots with average and standard deviation. This trick of the eye is due to that I am used to seeing such columns with error bars that indicate the uncertainty of the average, not the standard deviation of a single measurement. I think you could benefit from plotting what is known as "standard error of the mean" in stead of the estimate of the standard deviation. The standard error is the standard deviation divided by the square root of the number of samples. In this way you would exploit the often large number of samples to indicate that the uncertainty of the average value is reduced.

See e.g. https://en.wikipedia.org/wiki/Standard_error

  • Upvote 1
Link to comment
Share on other sites

How did I miss this post? Very interesting. I haven't had a wolf fight since the cartographer update, so don't know anything. I remember before the update we could sometimes fight off the wolf without any damage by starting clicking before the actual attack. I presume that trick has been removed with the 'weapon selection'?

Link to comment
Share on other sites

1 minute ago, Hotzn said:

I remember before the update we could sometimes fight off the wolf without any damage by starting clicking before the actual attack. I presume that trick has been removed with the 'weapon selection'?

Nope, you can still get away with no damage on occasion.  In the most recent graph I posted, you can see there were a few struggles that resulted in no damage (both with and without clothes).

As far as I can tell, it's a matter of clicking at just the right time before the struggle actually begins (and just after selecting a weapon).

Link to comment
Share on other sites

Nice graphs @Timber Wolf^_^

You seem to be refining your method for teasing out the game's mechanics ;)

Still, it's too bad that there is still a "best" weapon for struggles. The notes for the latest release indicated that there should have been more variation. For instance, the knife (if memory serves) was supposed to have a higher chance to kill the wolf and the hammer was supposed to end struggles quicker. Clearly that is not the case though. Otherwise the hammer would result in the least condition loss 

  • Upvote 2
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now