Jump to content

RNG Simulator


Buff00n
 Share

Recommended Posts

I've been kicking this idea around for a long time, and now I finally have a version of it that you can run directly in your browser:

Warframe RNG Simulator

It works in most browsers I've tested (don't use Edge).

The basic idea is to set up some scenario where success is dependent on RNG, like farming a specific Warframe or rolling a Riven looking for specific criteria, and then simulate that scenario thousands of times to see how long it actually takes.  With a simple drop table with a single item you can directly calculate statistics like the expected average and 90th percentile.  But when you throw in multiple items, multiple drops, or something with complicated criteria like a Riven, then it becomes easier to just do the simulation.

There's a UI for setting up the scenario, and a ton of documentation under the 'Help' section.

The output of the simulation is a bunch of statistics plus a chart where you can see the distribution of how long something takes, either in terms of the number of missions or, with some scenarios, in terms of actual minutes or hours spent.  You can save the chart as an image to share, and you can also copy a link to share that goes directly to your scenario.

There are a ton of examples in there already.  Feel free to send me links to ones you create, and if they're good then I'll add them in there.  If you have a scenario you'd like to see that's not there then let me know in a comment.

Edit (2018-12-19): I added some Profit-Taker examples based on the latest drop tables:

Spoiler

Profit-Taker: All Mods.  Trying to get all twelve new mods without trading is a bit of a grim proposition.

dR9euGU.png

Edit (2018-11-14): I added some Fortuna examples:

Spoiler

Assuming you don't waste a lot of time doing K-Drive tricks between missions, the grind for Garuda is comparable to Gara, at least for getting the blueprints.  Getting the materials is a whole different story.

YbUbq6D.png

The only other things that stand out are how long it takes to get enough Medical Debt-Bonds and Advances Debt-Bonds to fully rank up Solaris United standing.  

IwnZaHl.png

4er9QvH.png

I think the takeaway here is to check Ticker's stock often, at least while you're still in the process of ranking up SU.

 

My original goal with this was to compare various Warframe farms in terms of expected time commitment.  Here's a few of the heaviest ones:

Spoiler

Equinox:

E0hhBz8.png

Ivara (If you're bad at Spy):

ETKzvVc.png

Khora:

lPWnWeC.png

(The second visible layer on the Khora chart reflects the small number of players that get all the rotation C rewards first and then start bailing early after four or six zones.) 

Once I had those in place I could expand the basic concept to cover things like hard-to-get Mods:

I just added some Arbitration scenarios too, to see how long it takes to get all four of the mods:

I also wanted to see how long it took to roll certain kinds of Rivens.  That took a lot of work, but the result is pretty fun to play with:

Spoiler

Basic damage Gun Riven:

WK06rxb.png

Either a Crit or a Status melee Riven:

P7PAFA6.png

"Godlike" Lanka Riven:

PVNVl42.png

I'll admit these don't really meet many players' definitions of GROLLS, but looking at numbers for when you get extra super picky will only serve to depress you.

The examples also includes a selection of some of the heavier grinds.  Mostly, these are here to show how important trading is to getting what you want in Warframe.

I'm not trying to hate on RNG; I'm just trying to provide an understanding of it.  You can't set guarantees with probability, but you can set expectations.

Edited by Buff00n
Profit-Taker examples
Link to comment
Share on other sites

  • 4 weeks later...

Huh, today I happened to validate my numbers on the Braton Vandal/Lato Vandal farm.  

I calculated the expected average number of runs to get all seven parts at just under 103.  For me it took 97 runs, a very average result.

Braton BP 2
Braton Barrel 9
Braton Receiver 5
Braton Stock 3
Lato BP 1 (this was the holdout)
Lato Receiver 2
Lato Barrel 1

Science: It works sometimes.

Link to comment
Share on other sites

7 hours ago, GOOFBALL1 said:

So it takes approximately 100 rolls to get a god roll for riven huh... can't say I like that.

I believe my sims said more towards 500 rolls for god tier.  You can get a pretty decent one with 100 rolls though.  If you do some searching on YT you can find quite a few "1 rolled this 100x" video and yeah....

Link to comment
Share on other sites

On 2018-10-13 at 12:51 PM, Buff00n said:

I've been kicking this idea around for a long time, and now I finally have a version of it that you can run directly in your browser:

Warframe RNG Simulator

It works in most browsers I've tested (don't use Edge).

The basic idea is to set up some scenario where success is dependent on RNG, like farming a specific Warframe or rolling a Riven looking for specific criteria, and then simulate that scenario thousands of times to see how long it actually takes.  With a simple drop table with a single item you can directly calculate statistics like the expected average and 90th percentile.  But when you throw in multiple items, multiple drops, or something with complicated criteria like a Riven, then it becomes easier to just do the simulation.

There's a UI for setting up the scenario, and a ton of documentation under the 'Help' section.

The output of the simulation is a bunch of statistics plus a chart where you can see the distribution of how long something takes, either in terms of the number of missions or, with some scenarios, in terms of actual minutes or hours spent.  You can save the chart as an image to share, and you can also copy a link to share that goes directly to your scenario.

There are a ton of examples in there already.  Feel free to send me links to ones you create, and if they're good then I'll add them in there.  If you have a scenario you'd like to see that's not there then let me know in a comment.

 

My original goal with this was to compare various Warframe farms in terms of expected time commitment.  Here's a few of the heaviest ones:

Equinox:

E0hhBz8.png

Ivara (If you're bad at Spy):

ETKzvVc.png

Khora:

lPWnWeC.png

(The second visible layer on the Khora chart reflects the small number of players that get all the rotation C rewards first and then start bailing early after four or six zones.) 

Once I had those in place I could expand the basic concept to cover things like hard-to-get Mods:

Vengeful Revenant:

iIRDsNC.png

Condition Overload:

G9um9pf.png

I just added some Arbitration scenarios too, to see how long it takes to get all four of the mods:

30 minute survivals:

XLpWGjP.png

1 hour survivals:

H1nIkd1.png

1 hour interceptions:

5jQljW9.png

I also wanted to see how long it took to roll certain kinds of Rivens.  That took a lot of work, but the result is pretty fun to play with:

Basic damage Gun Riven:

WK06rxb.png

Either a Crit or a Status melee Riven:

P7PAFA6.png

"Godlike" Lanka Riven:

PVNVl42.png

The examples also includes a selection of some of the heavier grinds.  Mostly, these are here to show how important trading is to getting what you want in Warframe.

Getting ten Arcane Energizes from Tridolon hunts:

GhlXuH6.png

Full Braton Vandal and Lato Vandal sets from ESO:

FX3GFk9.png

 

I'm not trying to hate on RNG; I'm just trying to provide an understanding of it.  You can't set guarantees with probability, but you can set expectations.

nice Monte Carlo simulations mind if I use them in a school report.  

Link to comment
Share on other sites

2 hours ago, (XB1)Danowat32 said:

Can you also make an RNG RNG simulator, to give me the probability of the results of your RNG simulator being correct?

this is an RNG simulator and they are used in science all the time to predict variance in an experiment.  I am learning about them and bell curves for engineering.   

Link to comment
Share on other sites

1 hour ago, Volinus7 said:

I'm assuming the values provided in the official drop tables are the actual probabilities, after the normalization, attenuation, and bias described in VoiD_Glitch's doc.

As for their description of how RNG actually works, yeah, as as far as I can tell, the Lua scripting language used by Warframe still uses C++'s standard LCG algorithm.  Modern browsers like Chrome and Firefox (using the V8 Javascript engine) have almost all switched to xorshift128+.  However, the point of my work is to take the published drop table numbers and see where they lead, not to dig into the guts of the random number generator itself.

Link to comment
Share on other sites

9 hours ago, GOOFBALL1 said:

So it takes approximately 100 rolls to get a god roll for riven huh... can't say I like that.

1 hour ago, Chappie1975 said:

I believe my sims said more towards 500 rolls for god tier.  You can get a pretty decent one with 100 rolls though.  If you do some searching on YT you can find quite a few "1 rolled this 100x" video and yeah....

It all depends on what weapon you're rolling for and what you define as a "god roll".  You can see the exact choices I made on each of my examples if you scroll down to the stat listing.  They range from 1 in 50ish to 1 in a million and up, depending on how narrowly you decide to define "god roll".

I think the two take-aways from the Riven examples are that you should (1) broaden your expectations, and (2) roll a bunch of Rivens a moderate number of times each rather than roll a single Riven hundreds of times.

Link to comment
Share on other sites

3 hours ago, (XB1)Danowat32 said:

Can you also make an RNG RNG simulator, to give me the probability of the results of your RNG simulator being correct?

Yes, quite easily in one sentence: The longer you run the simulation the more accurate it is.

Link to comment
Share on other sites

1 hour ago, spirit_of_76 said:

nice Monte Carlo simulations mind if I use them in a school report.  

1 hour ago, spirit_of_76 said:

this is an RNG simulator and they are used in science all the time to predict variance in an experiment.  I am learning about them and bell curves for engineering.   

Sure!  The source is right there, it's Javascript.

However, this kind of simulation doesn't really produce a bell curve/normal distribution.  In general, running Bernoulli Trials until success produces a Negative Binomial Distribution, which is not bell-shaped.   It approaches a normal distribution if the degrees of freedom are high enough, but it will never actually get there.

Link to comment
Share on other sites

13 hours ago, Buff00n said:

It all depends on what weapon you're rolling for and what you define as a "god roll".  You can see the exact choices I made on each of my examples if you scroll down to the stat listing.  They range from 1 in 50ish to 1 in a million and up, depending on how narrowly you decide to define "god roll".

I think the two take-aways from the Riven examples are that you should (1) broaden your expectations, and (2) roll a bunch of Rivens a moderate number of times each rather than roll a single Riven hundreds of times.

Well yeah, the way to go with rivens is to roll until you get an acceptable roll and use it yourself or if you luck out with a good weapon roll you can live without, sell it. At least that's how I roll. But that doesn't change what's god tier. I've seen people claiming all kinds of rivens god tier, but to me god tier is when there's nothing above it: three wanted stats and a benign or even a wanted negative, and in the case of Lanka you really want that -zoom for easier scope use, which makes Lanka god tier more restricted and harder to get. Other stats would be cc cd ms, or instead of cc maybe electricity or heat.

But yeah, thanks for the simulator!

Link to comment
Share on other sites

20 minutes ago, GOOFBALL1 said:

Well yeah, the way to go with rivens is to roll until you get an acceptable roll and use it yourself or if you luck out with a good weapon roll you can live without, sell it. At least that's how I roll. But that doesn't change what's god tier. I've seen people claiming all kinds of rivens god tier, but to me god tier is when there's nothing above it: three wanted stats and a benign or even a wanted negative

Probably my closest example to what you're talking here is what I called "Perfect Rubico Riven", which is indeed about a 1 in 500 chance, but even that has a lot of leeway and isn't requiring the full 3 + 1 stats.

 

27 minutes ago, GOOFBALL1 said:

and in the case of Lanka you really want that -zoom for easier scope use, which makes Lanka god tier more restricted and harder to get. Other stats would be cc cd ms, or instead of cc maybe electricity or heat.

What you're talking about here, a Lanka Riven with +CD, +MS, 1 of +CC, +Electricity, or +Heat, and -Zoom, is roughly 1 chance in 50,000.  It's so unlikely that the simulation takes several minutes on a pretty serious laptop just to see it happen a few hundred times.

If you narrow it even further, to just +CD, +CC, +MS, and -Zoom, then that's an example I already have up there, "Riven With Exact Stats".  You have to let it run for a very long time to get a decent number, but the odds are around 1 chance in 150,000.

These results are so outside the realm of bonkers that nobody should be setting any goals on achieving it.  Either you win the space lottery, or someone else does and you have tens of thousands of plat to burn.  They are almost purely hypothetical fantasyland.

I guess, being fantasy they are fun to fantasize about.  However, I prefer running the numbers on a lower, "Godlike" tier that a hardcore player could conceivably set as a goal that's achievable before the sun swallows the Earth.

Link to comment
Share on other sites

7 minutes ago, Buff00n said:

Probably my closest example to what you're talking here is what I called "Perfect Rubico Riven", which is indeed about a 1 in 500 chance, but even that has a lot of leeway and isn't requiring the full 3 + 1 stats.

 

What you're talking about here, a Lanka Riven with +CD, +MS, 1 of +CC, +Electricity, or +Heat, and -Zoom, is roughly 1 chance in 50,000.  It's so unlikely that the simulation takes several minutes on a pretty serious laptop just to see it happen a few hundred times.

If you narrow it even further, to just +CD, +CC, +MS, and -Zoom, then that's an example I already have up there, "Riven With Exact Stats".  You have to let it run for a very long time to get a decent number, but the odds are around 1 chance in 150,000.

These results are so outside the realm of bonkers that nobody should be setting any goals on achieving it.  Either you win the space lottery, or someone else does and you have tens of thousands of plat to burn.  They are almost purely hypothetical fantasyland.

I guess, being fantasy they are fun to fantasize about.  However, I prefer running the numbers on a lower, "Godlike" tier that a hardcore player could conceivably set as a goal that's achievable before the sun swallows the Earth.

Woah, thanks for taking the time to run the simulation for that. And here I thought 1:100 was bad lol. Really wish there was some way to lock riven stats, even if that'd make it untradeable.

Link to comment
Share on other sites

26 minutes ago, GOOFBALL1 said:

Woah, thanks for taking the time to run the simulation for that. And here I thought 1:100 was bad lol. Really wish there was some way to lock riven stats, even if that'd make it untradeable.

I know Path of Exile has a system to lock stats on randomly rolled equipment.  However, IIRC, it's limited, complicated, and very expensive, and it's probably one of the more friendly example you'll find in a F2P game.

Link to comment
Share on other sites

6 hours ago, (XB1)Danowat32 said:

What does the sim say for Mesa Systems then?

Mesa's drop table is just like any other standard boss Frame, which means getting all three parts takes 6 runs on average, and 10 for 90% confidence.  

What makes Mesa annoying is the Mutualist Alad V coordinates.  By far the most time-effective way to farm those is to wait for Infested invasions that drop them.  A distant second is Orokin Derelict Defense, which will average about 3 1/2 hours to get three Nav coords if you don't bring a Speed Nova.  If you stay for the full 20 rounds to get Formas then it's about 4 1/2 hours.  This is lessened by the fact that you can trade coordinates and built keys.  You can also run key shares to effectively quadruple your keys, which you definitely should be doing.

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
 Share

×
×
  • Create New...