NOTICE: NEW FORUMS

We are excited to share with you our brand-new and improved Disruptor Beam forums!

We've chosen new forum software that brings added functionality to improve your experience, including increased spam management. Your username details have been carried over, so you should be able to login with your old forum details without issue. For our veteran community members, you'll see that we have many of the thread categories you’re used to, so we hope you'll feel right at home there.

Our new forums are located here: https://forum.disruptorbeam.com/stt/

 
   
 1 2 3 >  Last ›
STT Voyage Estimator
Posted: 24 September 2017 02:30 PM   [ Ignore ]  
Avatar
Cadet
Rank
Total Posts:  29
Joined  2017-07-30

I had some free time last night, so I worked out a way to estimate voyage duration based on skill/starting AM input:

https://codepen.io/somnivore/full/Nabyzw

Switch to editor view to see the source code with the formulas, or click here:

https://codepen.io/somnivore/pen/Nabyzw

My original spreadsheet that also has this calculator as well as some raw data:

https://docs.google.com/spreadsheets/d/1IS2qEggZKo1P1kBJq-qoDxJvxtKfQXpfVna9z4E_dNo/edit?usp=sharing

It would help to have more data. I’ve created a form where anybody can submit their skill/ship voyage stats along with the actual voyage time. I can use this to refine the variables used by the estimation formulas:

https://docs.google.com/forms/d/e/1FAIpQLSdWEJWABCVZiBKWeTnFevCYeg9f9DCG_x49KdlSnGTOh3acEQ/viewform?usp=sf_link
Please contribute all your data points - not just the outliers (i.e. don’t just contribute when the estimate is off for you)

Here is another form for hazard check totals to help us narrow down the exact chance of primary vs secondary vs other skills being checked for hazards:

https://docs.google.com/forms/d/e/1FAIpQLScAWaefv0h1Rc2m-plYWeA6KkV_M-qDMBUbMLN1FORvm8J5gA/viewform?usp=sf_link

Quick crew allocation tips:
* Don’t give up skills for traits. Each trait only adds about 1 minute (per refill)
* Prioritize primary and secondary skills. The chance of primary, secondary, and other skills being tested for hazards are (roughly!): 1/3, 1/4, and 1/10
* Go for crew that have the highest total skills
* Avoid any one skill being very high (even primary/secondary skills). The goal is to reach a point where all hazard checks fail. If one skill is too high, the voyage will end while that skill is done making successful hazard rolls. I will try to add automated warnings to my estimator later, but try to keep your highest skill to be under 1000*hours estimated.
* Although I’m not entirely sure yet, I believe the optimal distribution is to have primary and secondary skills at the aforementioned “cap”. So, something like 8000/8000/1000/1000/1000/1000. Ultimately, it depends on the crew you have available. Basically, to compute the value of each crew, add the product of each skill multiplied by the skill check chance (I may at a little calculator panel for this later)
* Slots only restrict what crew can go there. So out of the crew you want to use, arrange them such that you get the most trait bonuses (without sacrificing total skill levels for it)

P.S. If you’re near the end (failing all hazard checks) and want to estimate the time left, just divide AM by 21 for an accurate estimate (in minutes), or by 22 for a safe estimate. I plan to add this to the estimator page later when I have a bit more free time. If the difference between your highest and lowest skills is very large, this estimate will be less accurate.

This is not Spam. This is not Spam! This is not spam.

 Signature 

Engineering Officer in the fleet Subspace Eddies

Profile
 
 
Posted: 24 September 2017 03:31 PM   [ Ignore ]   [ # 1 ]  
Avatar
Yeoman
RankRank
Total Posts:  81
Joined  2017-05-03

and its about 21Antimater/Minute after that point on refills..

Profile
 
 
Posted: 24 September 2017 03:40 PM   [ Ignore ]   [ # 2 ]  
Avatar
Cadet
Rank
Total Posts:  29
Joined  2017-07-30
[ISA] Raven - 24 September 2017 03:31 PM

and its about 21Antimater/Minute after that point on refills..

Good idea - I’ll add a refill estimate too. Yeah, it’s about 21 (with my current constants, I get 21)

Update: revised from 20.5 to 21

 Signature 

Engineering Officer in the fleet Subspace Eddies

Profile
 
 
Posted: 24 September 2017 03:47 PM   [ Ignore ]   [ # 3 ]  
Avatar
Yeoman
RankRank
Total Posts:  81
Joined  2017-05-03
Chewable C++ - 24 September 2017 03:40 PM
[ISA] Raven - 24 September 2017 03:31 PM

and its about 21Antimater/Minute after that point on refills..

Good idea - I’ll add a refill estimate too. Yeah, it’s about 21 (with my current constants, I get 20.5)


Rounded up….. cause you don’t want to Miss it and have to pay if you don’t plan to…

Profile
 
 
Posted: 24 September 2017 03:59 PM   [ Ignore ]   [ # 4 ]  
Avatar
Cadet
Rank
Total Posts:  29
Joined  2017-07-30

It’s important to realize that this is just an estimate. It represents (in theory) the average time the voyage will take if you were to run it hundreds of times. There’s no guarantee. It’s not a minimum.

For the 6 test runs I have data for, the estimate is within about 7% now. That’s pretty good. But you shouldn’t entirely rely on it. Always assume the estimate is too high to be safe.

Perhaps once I have more data, I can also show the error of the estimate (e.g. show a confidence time interval).

I guess I’ll add a -10% estimate for now. But I don’t know the degree of confidence - the percent chance that the estimate actually falls within 10%. 6 sample points is not a lot.

 Signature 

Engineering Officer in the fleet Subspace Eddies

Profile
 
 
Posted: 24 September 2017 05:06 PM   [ Ignore ]   [ # 5 ]  
Avatar
Lieutenant
RankRankRankRank
Total Posts:  277
Joined  2016-04-19

This is cool, thanks for putting the work into it. Where would you like the data from our voyages?

 Signature 

Adult Swim Fleet

130 crew immortalized because voyages. 37% to Kahless (because voyages again).

“The truth is usually an excuse for a lack of imagination” - Elim Garak

Smudge. Never forget.

Profile
 
 
Posted: 24 September 2017 05:21 PM   [ Ignore ]   [ # 6 ]  
Avatar
Cadet
Rank
Total Posts:  29
Joined  2017-07-30
Bigstupidgrin - 24 September 2017 05:06 PM

This is cool, thanks for putting the work into it. Where would you like the data from our voyages?

Good question!

I threw together a Google Form you can use to submit your data to me (this is linked directly to the sheet I linked above - but in a separate tab; I’ll transfer it to the correct tab manually):

https://docs.google.com/forms/d/e/1FAIpQLSdWEJWABCVZiBKWeTnFevCYeg9f9DCG_x49KdlSnGTOh3acEQ/viewform?usp=sf_link

You can also post your data here if you like, I suppose, or link to it, etc.

Thanks!

 Signature 

Engineering Officer in the fleet Subspace Eddies

Profile
 
 
Posted: 25 September 2017 05:34 PM   [ Ignore ]   [ # 7 ]  
Avatar
Cadet
Rank
Total Posts:  1
Joined  2017-05-16

First of all, thanks for your work. I’m triyng to understand your program, but what’s maxSkill? what’s the influence of the maximum skill in the odds of lose a hazard?

Profile
 
 
Posted: 26 September 2017 03:09 AM   [ Ignore ]   [ # 8 ]  
Avatar
Cadet
Rank
Total Posts:  29
Joined  2017-07-30
guest_824679002824704 - 25 September 2017 05:34 PM

First of all, thanks for your work. I’m triyng to understand your program, but what’s maxSkill? what’s the influence of the maximum skill in the odds of lose a hazard?

I’m using some math trickery to make the formulas relatively “simple”.

The way I have it may not be the best way to go about it - I just sort of wrote quickly threw it together.

maxSkill - the highest of your 6 skills, yep (the max function simply selects the highest value)

There is no special mechanic involving your highest skill. Actually, the way I use it here means that if you have a skill that’s very high compared to other skills, the estimate will be less accurate.

To make the formulas “simple”, I split the voyage into 3 parts, very similar to how STT indicates difficulty in away missions:

Easy - “guaranteed” to pass every hazard roll
Medium - win some, lose some hazard rolls
Hard - lose all hazard rolls

Now, “guaranteed” is in quotes for good reason - there seems to be some randomness for hazard skill checks. It would certainly be nice to understand exactly how random it is, but so far I don’t have enough data for that, and getting that data is very tedious.

So the formulas pretend that there is no randomness (that the range of variation per skill is actually 0).

Having split up the voyage into those 3 parts, I calculate separately each part:

Easy: I don’t have a separate variable for this, since it’s so simple: maxSkill/hazSkillPerHour = time spent in the “easy” section
Medium: that’s hazAmUsedInterim - the amount of AM you use in that section of the voyage
Hard: the “rest” of the voyage - expressed as amLeft - the amount of AM left after the easy and medium sections, easily convertible into timeLeft

The formulas have a glaring flaw in that they assume that when your voyage is done, the difficult is so high that you can’t pass any hazard rolls. That’s obviously not always going to be true, though, and because of this the estimate can be slightly higher than it should be - due to assuming that you will get all the AM “possible” from winning hazard rolls.

Hope that helps clarify what’s going on. Feel free to ask more questions / suggest ideas / fork my code / etc.

 

 Signature 

Engineering Officer in the fleet Subspace Eddies

Profile
 
 
Posted: 26 September 2017 08:23 AM   [ Ignore ]   [ # 9 ]  
Avatar
Commander
RankRankRankRankRankRank
Total Posts:  2763
Joined  2015-12-05

Very nice thing initiative smile

It is surprisingly precise for having such a simple mathematics (one can actually do a lot with simple averages). I inserted the data about the run that just finished for me, and the prediction was just 17 min off.

I’m personally toying with the idea of writing a short Monte Carlo simulation of the whole thing, and this would be a very good starting point wink

 Signature 

Minor contributor to the STT Wiki

Profile
 
 
Posted: 26 September 2017 09:01 AM   [ Ignore ]   [ # 10 ]  
Avatar
Commander
RankRankRankRankRankRank
Total Posts:  1658
Joined  2016-12-12

Thank you. This is phenomenal.

Profile
 
 
Posted: 26 September 2017 10:55 AM   [ Ignore ]   [ # 11 ]  
Avatar
Yeoman
RankRank
Total Posts:  97
Joined  2017-03-06

Plus a million!  Awesome work.  Any thoughts of linking this in the voyage section of the wiki?  Great place to bookmark it for people.

Profile
 
 
Posted: 26 September 2017 11:54 AM   [ Ignore ]   [ # 12 ]  
Avatar
Commander
RankRankRankRankRankRank
Total Posts:  2763
Joined  2015-12-05
[TFA] GTMET - 26 September 2017 10:55 AM

Plus a million!  Awesome work.  Any thoughts of linking this in the voyage section of the wiki?  Great place to bookmark it for people.

The wiki now has a link to this thread, unless Chewable C++ wants otherwise.

 Signature 

Minor contributor to the STT Wiki

Profile
 
 
Posted: 26 September 2017 01:33 PM   [ Ignore ]   [ # 13 ]  
Avatar
Cadet
Rank
Total Posts:  29
Joined  2017-07-30
Centurion Matt - 26 September 2017 11:54 AM
[TFA] GTMET - 26 September 2017 10:55 AM

Plus a million!  Awesome work.  Any thoughts of linking this in the voyage section of the wiki?  Great place to bookmark it for people.

The wiki now has a link to this thread, unless Chewable C++ wants otherwise.

No objections here! I was thinking of updating the Voyages info in the wiki to reflect what I’ve learned but haven’t gotten around to it and thought it may be better to wait until my confidence in the numbers solidifies further.

Thanks for all the data submissions already! Please continue to submit your run data through the form. We still could use more to refine the values.

Except for the last 2 submissions, the previous ~15 were all within 2% of the current values! Unfortunately, the latest 2 submissions seem to be big outliers. They both had a very uneven skill distribution, especially the last one (1500-10k skill range) - and this runs up against the problem with the formulas I described earlier.

I may mark those as outliers for now, but later I will try to refine the calculations to better take these situations into account (both directly and possibly by calculating a dynamic confidence factor based on the skill distribution).

 Signature 

Engineering Officer in the fleet Subspace Eddies

Profile
 
 
Posted: 26 September 2017 02:34 PM   [ Ignore ]   [ # 14 ]  
Avatar
Commander
RankRankRankRankRankRank
Total Posts:  2763
Joined  2015-12-05
Chewable C++ - 26 September 2017 01:33 PM

Except for the last 2 submissions, the previous ~15 were all within 2% of the current values! Unfortunately, the latest 2 submissions seem to be big outliers. They both had a very uneven skill distribution, especially the last one (1500-10k skill range) - and this runs up against the problem with the formulas I described earlier.

I guess that one solution for you might be do try to estimate error bars. The result of your algorithm should probably be more of an interval than a precise number anyway, imho, and studying these outliers might help you in establishing good boundaries for these intervals.

(Pragmatically, that is. An ab initio estimation of the intervals could be very, very tricky.)

 Signature 

Minor contributor to the STT Wiki

Profile
 
 
Posted: 26 September 2017 03:26 PM   [ Ignore ]   [ # 15 ]  
Avatar
Cadet
Rank
Total Posts:  29
Joined  2017-07-30
Centurion Matt - 26 September 2017 02:34 PM
Chewable C++ - 26 September 2017 01:33 PM

Except for the last 2 submissions, the previous ~15 were all within 2% of the current values! Unfortunately, the latest 2 submissions seem to be big outliers. They both had a very uneven skill distribution, especially the last one (1500-10k skill range) - and this runs up against the problem with the formulas I described earlier.

I guess that one solution for you might be do try to estimate error bars. The result of your algorithm should probably be more of an interval than a precise number anyway, imho, and studying these outliers might help you in establishing good boundaries for these intervals.

(Pragmatically, that is. An ab initio estimation of the intervals could be very, very tricky.)

Yeah, I’m working on going in that direction as I find time.

Refining this further gets very tricky indeed.

I already provide a -10% estimate on the calculator now, which is pretty simple, but also ignores the nature of the inputs (low pri/sec leads to lower confidence, etc)

In the shorter term, I might at least add a warning if your primary or secondary skills are too high

 Signature 

Engineering Officer in the fleet Subspace Eddies

Profile
 
 
   
 1 2 3 >  Last ›
1 of 4