HoundsTooth
-
Posts
84 -
Joined
-
Last visited
Content Type
Profiles
Forums
Events
Store
Articles
Patch Notes
Posts posted by HoundsTooth
-
-
Can Illusion/Rad pump out decent damage without containment? I'm thinking through APP's and none of the attacks seem worth taking without the extra damage from containment. Controller damage modifiers are so low.
-
I remade my Ill/Rad, and I'm having a lot of fun with it. I'm not sure how to up my late-game damage, aside from making Phantom Army perma. Other controllers can lock down mobs to get containment, then hit them with APP's. Is there any way to get equivalent damage with Illusion? Is spamming Flash the best/only way to set up containment, since Illusion doesn't get an AOE immobilize or stun?
-
Play content without IOs and come back and say the same thing but you can't. It takes a lot of INF, IOs, and time to make any AT trivialize content. A brute or scrapper cannot solo +4 x8 without IOs, they would absolutely get demolished.
As much as I love the madness unleashed by IOs and Incarnates, I may start doing a series of ‘no sets, no incarnates’ characters. -Maybe- SOs only, though lvl 50 IOs and HammiOs do open up possibilities.
Not that Im -against- Game-Shattering Power, but I feel the ATs and powersets start to lose some of their distinction under the weight of incarnates and set bonuses, so it might be cool?
You could try having two builds? One OP build with lots of sets and other build with just SO's. You can switch between the two.
I change my mindset as I play. Levels 1-45 are when I can have a regular build. Once I get closer to 50 and get veteran levels, I like the challenge of getting the most OP build I can afford.
-
I am a statistics PhD. Your understanding is correct. The rule of thumb is that when you have >30 sample, you can apply the CLT. This has nothing to do with a sample representing the population. It is used to calculate uncertainty, e.g. margin of error
Fascinating. So for a population of 100K and provided you are doing random sampling, how do you determine a valid sample size? I know there are several variables to consider.
I have a PhD in Terryology. So, I can tell you definitively that the best way to determine the best sample size is to prove that 1x1=2. But that'd be obvious if you were a genius like me.
Impressive. I failed my Terrology 101 class.
-
I am a statistics PhD. Your understanding is correct. The rule of thumb is that when you have >30 sample, you can apply the CLT. This has nothing to do with a sample representing the population. It is used to calculate uncertainty, e.g. margin of error
Fascinating. So for a population of 100K and provided you are doing random sampling, how do you determine a valid sample size? I know there are several variables to consider.
I suspect you chose 100K so you can get a nice, clean number but it's not so simple. Nothing in life is simple :)
You need to decide on a few things:
[*] You choose a model you want to fit. Linear regression is the most common model.
[*] You say the smallest relationship you are interested in seeing. Ex: If my new medicine works, I want to be able to see an increase of at least 3 weeks remission time.
[*] You define the statistical power you want. Power is the probably of seeing a relationship if there is one. Ex: If there my new medicine treats cancer better than a placebo, I want an 80% chance to see this improvement
[*] Confidence level. Ex: 95% confidence
Then there are formulas to get you the required sample size. You can look those up if you're interested. Here's a link to a sample size calculator for linear regression https://www.danielsoper.com/statcalc/calculator.aspx?id=1.
Sample size calculations are surprisingly hard to do. People routinely publish papers about sample size calculations in new settings.
EDIT: If this is something you're interested in or something you need to learn I can DM you some resources.
-
I'm sorry, but this is really, really wrong.
A sample size of >30 is a totally unrelated concept. The rule of thumb is that when the sample is >30, you can use normal based confidence intervals because your sample size is large enough for the Central Limit Theorem to apply. You need large sample sizes so you can correctly sample everyone of interest in your population, and so you can add weights to your survey to account for low responses in certain groups.
Again, you are misunderstanding the text presented, which means it is not "really, really wrong."
To reiterate, a sample size of greater than 30 is VALID when a population is known. A sample of larger numbers of a population renders finer results, but the outcome is largely the same. Your proclamation that you need a large sample size to account for everyone of interest in a population is somewhat misleading, we are interested in the whole population, and must attempt to fairly represent them through sampling MEANS.
I do appreciate your insight. There are no incorrect statements, analogies, inferences, descriptors or concepts in my responses as you have mentioned.
My understanding for n=30 is that it's tied to the CLT, which can only be applied for data sets with successive, random sampling. Neither of which is present in a forum thread. Again, I am not a statistics PhD, but part of my background is in marketing, and I never heard the CLT applied there (just discovered it due to this thread). I've also seen quite a bit of debate about the utility of n=30 (in fact that it is an accepted standard due to it being what fits best in textbooks - though that seems spurious reasoning). As I said, feel free to PM me more information (on the web, or start a dialogue) if you are concerned about threadjacking - I'd really like to learn more. And for what it's worth, if you want people to buy into what you are saying you have to do more than say - I am an expert, trust me...you need some reasoning. I don't doubt you have it, please share!
I am a statistics PhD. Your understanding is correct. The rule of thumb is that when you have >30 sample, you can apply the CLT. This has nothing to do with a sample representing the population. It is used to calculate uncertainty, e.g. margin of error
-
Actually, I ran a statistical inference and regression anova analysis on this from the responses in the thread. Because there were more than 30 distinct sample responses, it is a statistically viable pool. With the alpha at 5%, confidence level of 95%, it is statistically accurate that the community supports this, as demonstrated by the sample numbers.
Being very familiar with statistical analysis, a sample size of 30 or more has been shown to be statistically representitive of the population. Data, and proof.
I’m not arguing but for a population in 50-90k range, 30 seems like too small a sample size
When I first began learning statistics, I felt the same. Mathematical evidence has shown that 30 or more is accurate, at around a 95% confidence with a margin of error of +/- 3%.
Ipsos and all other polls use the same mathematical standards proofed over the last 200 years or so.
Again, not trying to argue, just learn - most Ipsos polls and certain all of Gallup polls I've seen track around 1,000 responses - which is the confidence level and margin's you've expressed and covers a population larger than the planet earth has. For a population of 90K, with the same margin and confidence level, I'm calculating 988 responses needed.
Here's the tool I always used for marketing - https://www.surveysystem.com/sscalc.htm and what I used to derive the numbers above...
Again, I'm not trying argue, feel free to explain here or send me a PM. I'd love to learn more
Thread hijacking may be occuring when trying to delve into statistical mathematics. Simply, a sample size of 30 or greater is sufficient. Whether the sample is 30, 300 or 1000, the difference will continue to refine, but the results will largely be the same. Only when a a population is unknown is it necessary to calculate sample size. If it is unknown, the we use the afore mentioned formula.
I'm sorry, but this is really, really wrong.
A sample size of >30 is a totally unrelated concept. The rule of thumb is that when the sample is >30, you can use normal based confidence intervals because your sample size is large enough for the Central Limit Theorem to apply. You need large sample sizes so you can correctly sample everyone of interest in your population, and so you can add weights to your survey to account for low responses in certain groups.
-
The population can be considered known. The sample size is the participants in the thread.
The mistake you're making is that taking more samples fixes always fixes your inference. It does not. If you take a lot of sample but they don't represent the population, you get a wrong answer that is calculated very precisely.
You are making an assumption, whether you realize it or not. You have to assume your sample is representative of the population. You assumed the responses in this thread represent the responses of everyone playing the game. I have no idea if this is true or not, but I suspect it is not.
The population can be considered known. The sample size is the participants in the thread.
The argument that forum participants are not representative of the whole population can be accurate is the sample size is under 30. As sample size grows over 30, this is a non issue.
The sample size of 30 is a totally unrelated concept. It has nothing to do with how representative a sample is. People say it's safe to trust [linear] regression hypothesis tests (confidence intervals and p-values) with a sample > 30.
This goes beyond a simple poll of data, in a vacuum. Why? A set data poll within itself is a population, not a sample. A sample is used to model a population and must be run through statistical formulae.I'm sorry, but I have no idea what this means. A poll is a sample of a population. It has uncertainty in it which you must account for.
In essence, the thread has enough data to accurately model the population with a confidence level of 95% and a margin of error of +/- 3%, supporting that the population in majority (51% or greater) is in favor of the change.This is just not true. Your sample is not random. It is not represented of the population. You may be able to get the desired margin of error, but the sample is incorrect. You need a truly random sample or you need to apply weights when you perform an analysis.
I am a statistician. I am happy to help you make your poll better or to improve your understanding. Respectfully, the analysis is not as insightful as you believe.
-
You can still casually play and get a top build, and you can do it faster now than ever before. But it's a bit much to ask that you be able to do it in the short time these servers have been running.
Curious. Why is a "bit much to ask" that a person be able to do it in such a short period of time?
I'm not being combative. I'm really trying to understand the mentality that "X is too fast, and should be eliminated."
I'll qualify my question here: PVP, I get your point. That's competitive. But for a co-operative game that also allows solo play with built-in mission difficulty sliders so you can effectively solo content meant for full teams if you can build for that......and there's no financial reward for anyone for keeping the player base moving slowly or at least all at the same speed/pace.....
Why is this a bit much to ask for?
I don't understand either. I would like the understand the though process.
This game has been offline for 10 years and could be shut off at any time. This game is no longer subscription based, so there's no monetary reason to slow down player's growth. You should be able to do things quickly considering the current state of the game. On top of that, nothing stops you from slowing down the experience for yourself.
-
Hey, thanks for the reply. It is a valid argument. But I've always viewed CoH as a game for those who want light play and not power game. Imho, there is no reason I shouldn't be able to get the builds I want or the special IOs that go for bazillions simply because I don't constantly play 50s and get tons of influence. My argument is why restrict and force the playstyle? I don't' want to play the same character over and over again. I do want to be able to tweak my characters. It is just too much to expect me to do three TFs worth of content to get one IO. CoH has always been the game you could just jump in and jump out of and I love it for that, I just wish its advancement system matched that. It doesn't.
The big thing is, 34 merits (using a number you noted in your earlier post) buys you 102 Enhancement Converters. These are selling at about 125,000 inf, and slowly dropping, so let's call it 100,000. That means 102 x 100,000 = 10,200,000 Inf. As someone who does like to start using set IOs from level 20 onward, I can tell you that will set you up with recipes and the salvage to make them at low level (and you'll find many enhancements are cheaper to buy than to craft). Those few very expensive IOs (relative to the current market, at least), such as the Performance Shifter proc, you just buy a cheap end mod set recipe or enhancement and save a few converters to turn it into one. That can get iffy, though my worst luck to date with that has been burning through 12 converters.
While converters are dropping, boosters were selling for 1.75 mil last night and still had a low supply. Unslotters are also doing well at 150,000 each, but are also heavy on supply. The point is, to date, one of these 3 options will net you the money to slot your character, and if you keep spending your merits on whichever has the best inf to merit ratio at the time, you'll continue raking in the dough as you level up to the next group of sets you want to slot.
You kind of unintentionally proved their point. To make your argument you needed to know how much items are selling for on the market, and you had to do some math to calculate relative costs. Their arguing (and I am as well) that CoH is not a min-maxing game. You shouldn't need a deep knowledge of the market to build a great character.
If you have a lot of merits you can mostly ignore the market. It is not as effective as farming and playing the market, but it works. Before, we had an easy way to get enough merits to get good recipes.
-
I think reward merits should be increased across the board a good deal. At least twice what they are. MSR was definitely high, but honestly, IMHO it was in the right direction.
I agree with this 100%. You took the words out of my mouth.
I don't have a ton of time to play, and I don't farm. High end builds were always unobtainable to me. Doing MSR was the first time I've felt I could make a great build too. I understand that it was unbalanced, but there should be a comparable way to get merits.
-
I'd love to run it. My highest toon is level 40 and not IO'd. Hope that's okay.
I have a weird schedule, but I can make most days work if I plan ahead enough.
-
Dark Servant is stupid. It will pull mobs if you let it. You have to stay further back from mobs now. I always slotted extra recharge into Dark Servant. It comes in handy if it gets killed or if you want to drop it to take the first strike from a new mob.
Dark pit should be used whenever it's up. It's a set defining power that makes you hit harder.
I never used Black Hole. I think it's an emergency power that let's you escape and recover.
-
Yes and it took me forever to get it working. You can use Wine to run Pines. Link to Wine: https://www.winehq.org.
For some reason that no one can explain to me, if you install Wine through Homebrew it will not work run Mids or Pines. I think this is because of a problem with .Net, but that's just a guess.
-
I mean the Incarnate Hybrid Slot ability. There's "Melee Radial Embodiment" which gives scaling +Regen, +def and +CC protection per nearby enemy. It's not a permanent buff and has a 2 minutes uptime and a 2 minutes cooldown.
Acrobatics isn't really CC protection, it only gives mag 2 hold protection and some resistance. So in most cases you'll still get held, just for a slightly shorter period of time.
Thanks! I'm doing some of the endgame content for the first time, so I don't know all the lingo.
-
Another post.
Have you considered getting CC protection from Acrobatics? I would drop Mutation and Super Speed and replace those with Super Jump and Acrobatics. If you want to keep the stealth from Super Speed, you could drop Fallout. I never use Fallout, but it may be a mule for IO bonuses in your build.
-
Nice build! Thanks. I may work towards this before I roll a /Time corruptor.
EDIT: What do you mean by Melee Hybrid? Can you explain a bit more? I'm comfortable in melee.
-
I want to try out /Time, but I don't want to run another Fire blast corruptor as I'm also leveling my Fire/Rad corruptor. Could you recommend a second best primary? I'm thinking Sonic/Time or Ice/Time.
-
Anyone care to post this build in the "plain " for those that haven't yet downloaded a builder.?
Noob question here:
Murcielago do you need the Body Mastery pool for the endurance help? If you slot 2 Endurance Modification and a Performance Shifter in Stamina plus Numina's Convalesence in Health, you have a lot of endurance regeneration. You also have a fast charging Dark Consumption to help out. Pines/Mids doesn't play nicely with Mac so I can't check the exact numbers right now.
I'm considering taking another EPP/APP, either Dark or Fire for extra AoE damage.
-
I feel like the current game is focused on clearing big mobs quickly. To me, powersets with strong single target damage seem like a holdover from a different time. Why should I build a toon with good single target damage when a Fire AT can clear the bulk of the mob in the time it takes me to whittle down the bosses and LT's?
I understand there are no bad builds in this game, in the sense that everything is viable. From a numbers point of view, is there any reason to take Stone or Energy on any AT when you can get better AoE damage elsewhere? Am I looking at this the wrong way?
-
Will you keep extra XP Halcyon to encourage more players to join, or will it be reverted immediately?
-
How do you feel about taking Spring Attack as a filler until you get Shield Charge?
-
Affirmative, adding to that Against All Odds for a whole lot of +DMG. Between SD and AAO you'll be able to steamroll most +4x8 groups. The only weakness /Shield has vs SR is it's defense debuff protection being lower but it makes up for a lot more offense.
I run a Mac and I can't get Mids working yet. I can't look at your build.
How do you pump out enough AoE to take on 4x8?
-
Thanks. I'll give it a try. I've been wanting to roll my first Scrapper for a while.
Warshade or Peacebringer?
in Peacebringer & Warshade
Posted
I would go with a PB. WS's really benefit from a full IO build. Perma eclipse is a thing of beauty.
I also have not been able to remake my WS. Something about seeing him struggle through the low levels just makes me sad.