Black Widow 3% Evade? Cool Story Bro

https://youtu.be/XScv3vGKWXw

3% evade chance working as intended

Comments

Icecold2099 wrote: »

My point is that the 3% chance of BW evading is on a single item. So the sample size is 1 so it's either going to happen or not going to happen, but there is no way to predict how often it does happen or not happen because each sample size is only 1.

I'm afraid this is mathematical gibberish. The actual *definition* of "percent chance to occur" is value that is congruent to the average occurrence rate taken asymptotically over sufficiently large sets. In other words, you cannot simultaneously say that an event actually has a percent chance to occur and also say there's no calculable average rate of occurrence. That defies the definition of "chance," and "average."

To state this in a more practical way, when a game developer makes an event have a "3%" trigger, they mean they are using a PRNG to generate a value such that three times out of a hundred that value meets a certain criteria. And the way PRNGs themselves are created and tested mathematically is by generating large amounts of output and testing to see if the distribution of values meets certain criteria. "No bias" is one of those criteria, meaning if you generate values intended to pass a certain criteria 3% of the time, over long enough runs the generator will always do so within statistically defined margins for error.

Which means when a game developer says a single event has a 3% chance to occur in his or her game, they *mean* that if this situation is observed a large number of times the event will occur in 3% of those situations, plus or minus a statistical deviation that gets proportionately smaller as the number of observations increases. That is the definition of "statistical average."

Spurgeon14

Kabam Miike wrote: »

Hey OP,

This looks normal to me, unless I'm missing something. 3% Chance to Evade per hit does not mean she will only evade 3 out of 100 attacks, or even a guarantee that she would evade 1 in 1000. This is not how probability works. With a 3% chance per attack, there's a slim chance that you could throw 100 attacks, and she could evade each and every one of them.

It's best not to treat Black Widow like a Champion that can't evade. This is something I see come up a lot, but treat her the same way you treat Spider-Man, or Nightcrawler. Champions that can suppress Ability Accuracy, or negate Evasion (Ice-Man!) are great choices. I know that this isn't always possible in Dungeons, however, as you have your team of 3 and can't spy ahead.

So treat 3% like 20%+? Should we apply this same logic to crystal drop rates...? 😉🤔

Helicopter_dugdugdug

Icecold need to go to school ... have no concept of how probability is expressed

unknown

This content has been removed.

DNA3000

Verzz wrote: »

DNA3000 wrote: »

Icecold2099 wrote: »

My point is that the 3% chance of BW evading is on a single item. So the sample size is 1 so it's either going to happen or not going to happen, but there is no way to predict how often it does happen or not happen because each sample size is only 1.

I'm afraid this is mathematical gibberish. The actual *definition* of "percent chance to occur" is value that is congruent to the average occurrence rate taken asymptotically over sufficiently large sets. In other words, you cannot simultaneously say that an event actually has a percent chance to occur and also say there's no calculable average rate of occurrence. That defies the definition of "chance," and "average."

To state this in a more practical way, when a game developer makes an event have a "3%" trigger, they mean they are using a PRNG to generate a value such that three times out of a hundred that value meets a certain criteria. And the way PRNGs themselves are created and tested mathematically is by generating large amounts of output and testing to see if the distribution of values meets certain criteria. "No bias" is one of those criteria, meaning if you generate values intended to pass a certain criteria 3% of the time, over long enough runs the generator will always do so within statistically defined margins for error.

Which means when a game developer says a single event has a 3% chance to occur in his or her game, they *mean* that if this situation is observed a large number of times the event will occur in 3% of those situations, plus or minus a statistical deviation that gets proportionately smaller as the number of observations increases. That is the definition of "statistical average."

Thank you for explaining it better than I could ever. I felt like bashing my head against a piano at points here trying to explain it.

It helps if you realize that when Kabam says that BW has a 3% chance to evade, that literally means that somewhere there's an Excel spreadsheet that someone put a "3" in a particular cell (or more likely a 0.03). And why they put that three there is because they wanted a three out of a hundred *average* rate, and 3% *chance* gives that 3% *average rate* with statistical random variation. So while it isn't strictly speaking mathematically true that the two statements are identical in meaning, as far as game content development is concerned they are essentially synonymous.

It can be frustrating, but even the game developers get this wrong. Crystal odds in the game are described as "drop rates" but those are actually "drop odds." Those two phrases mean subtly but fundamentally different things, and the phrase the game uses is the wrong one. If the game said "statistical average drop rates" that would be correct, but also very long. Drop odds is the correct term.

Helicopter_dugdugdug

DNA3000 wrote: »

Verzz wrote: »

DNA3000 wrote: »

Icecold2099 wrote: »

My point is that the 3% chance of BW evading is on a single item. So the sample size is 1 so it's either going to happen or not going to happen, but there is no way to predict how often it does happen or not happen because each sample size is only 1.

I'm afraid this is mathematical gibberish. The actual *definition* of "percent chance to occur" is value that is congruent to the average occurrence rate taken asymptotically over sufficiently large sets. In other words, you cannot simultaneously say that an event actually has a percent chance to occur and also say there's no calculable average rate of occurrence. That defies the definition of "chance," and "average."

To state this in a more practical way, when a game developer makes an event have a "3%" trigger, they mean they are using a PRNG to generate a value such that three times out of a hundred that value meets a certain criteria. And the way PRNGs themselves are created and tested mathematically is by generating large amounts of output and testing to see if the distribution of values meets certain criteria. "No bias" is one of those criteria, meaning if you generate values intended to pass a certain criteria 3% of the time, over long enough runs the generator will always do so within statistically defined margins for error.

Which means when a game developer says a single event has a 3% chance to occur in his or her game, they *mean* that if this situation is observed a large number of times the event will occur in 3% of those situations, plus or minus a statistical deviation that gets proportionately smaller as the number of observations increases. That is the definition of "statistical average."

Thank you for explaining it better than I could ever. I felt like bashing my head against a piano at points here trying to explain it.

It helps if you realize that when Kabam says that BW has a 3% chance to evade, that literally means that somewhere there's an Excel spreadsheet that someone put a "3" in a particular cell (or more likely a 0.03). And why they put that three there is because they wanted a three out of a hundred *average* rate, and 3% *chance* gives that 3% *average rate* with statistical random variation. So while it isn't strictly speaking mathematically true that the two statements are identical in meaning, as far as game content development is concerned they are essentially synonymous.

It can be frustrating, but even the game developers get this wrong. Crystal odds in the game are described as "drop rates" but those are actually "drop odds." Those two phrases mean subtly but fundamentally different things, and the phrase the game uses is the wrong one. If the game said "statistical average drop rates" that would be correct, but also very long. Drop odds is the correct term.

But any normal person can understand what they mean by drop rate ...

Icecold2099

Verzz wrote: »

Obviously if your sample size is one, it is either going to happen once or zero times if that is what you are asking

.

If this is your prediction of something with a 3% chance happening then i guess you're correct, you can predict it. It's going to happen once or zero times.

I think you're under the impression that i don't understand what 3% is and how it correlates into the chance for something. That's not correct, i understand that just fine.

You're also taking the 3% probability and using that as a determination that it equals about 1/33. This is a correct conversion of decimal to fraction, but has no barring on the discussion.

The OPs point, to my understanding, is how BW 3% chance to evade triggers at a seemingly higher percentage. Then another poster stated that a 3% chance she should evade about once every 33 attacks.

I think the issue is that you don't understand that, regarding BW chance to evade, the sample size is 1. So you can't predict if she is going to evade that one attack or not. Let's say she does evade that attack though, that does not give you any information or indication if she is going to evade the next attack or not.

Hence, you cannot predict it.

You can, however, take an educated risk that because there is a 97% chance she is not going to evade she probably won't, but you absolutely cannot say with certainty that she will or will not evade any given attack.

It does not translate into her evading about once every 33 attacks.

If you still don't understand that, having played the game, watched the OPs video of the game, read others accounts of the game, then you're just not going to understand it.

DNA3000

Helicopter_dugdugdug wrote: »

DNA3000 wrote: »

Verzz wrote: »

DNA3000 wrote: »

Icecold2099 wrote: »

My point is that the 3% chance of BW evading is on a single item. So the sample size is 1 so it's either going to happen or not going to happen, but there is no way to predict how often it does happen or not happen because each sample size is only 1.

I'm afraid this is mathematical gibberish. The actual *definition* of "percent chance to occur" is value that is congruent to the average occurrence rate taken asymptotically over sufficiently large sets. In other words, you cannot simultaneously say that an event actually has a percent chance to occur and also say there's no calculable average rate of occurrence. That defies the definition of "chance," and "average."

To state this in a more practical way, when a game developer makes an event have a "3%" trigger, they mean they are using a PRNG to generate a value such that three times out of a hundred that value meets a certain criteria. And the way PRNGs themselves are created and tested mathematically is by generating large amounts of output and testing to see if the distribution of values meets certain criteria. "No bias" is one of those criteria, meaning if you generate values intended to pass a certain criteria 3% of the time, over long enough runs the generator will always do so within statistically defined margins for error.

Which means when a game developer says a single event has a 3% chance to occur in his or her game, they *mean* that if this situation is observed a large number of times the event will occur in 3% of those situations, plus or minus a statistical deviation that gets proportionately smaller as the number of observations increases. That is the definition of "statistical average."

Thank you for explaining it better than I could ever. I felt like bashing my head against a piano at points here trying to explain it.

It helps if you realize that when Kabam says that BW has a 3% chance to evade, that literally means that somewhere there's an Excel spreadsheet that someone put a "3" in a particular cell (or more likely a 0.03). And why they put that three there is because they wanted a three out of a hundred *average* rate, and 3% *chance* gives that 3% *average rate* with statistical random variation. So while it isn't strictly speaking mathematically true that the two statements are identical in meaning, as far as game content development is concerned they are essentially synonymous.

It can be frustrating, but even the game developers get this wrong. Crystal odds in the game are described as "drop rates" but those are actually "drop odds." Those two phrases mean subtly but fundamentally different things, and the phrase the game uses is the wrong one. If the game said "statistical average drop rates" that would be correct, but also very long. Drop odds is the correct term.

But any normal person can understand what they mean by drop rate ...

Quite a few "normal" people have interpreted "drop rate" to mean that's how often those things will drop, which is incorrect. That is the statistical long term average but no one person ever actually sees those rates. It is a point of confusion that is only making the general confusion about drop odds unnecessarily worse.

unknown

This content has been removed.

DNA3000

Icecold2099 wrote: »

The OPs point, to my understanding, is how BW 3% chance to evade triggers at a seemingly higher percentage. Then another poster stated that a 3% chance she should evade about once every 33 attacks.

Removing the word "average" from my words changes the meaning of them sufficiently that you're no longer referring to what I said at all. Nowhere did I say that a 3% chance means that evade should happen at or about once every 33 attacks, and I would correct anyone who said that. In fact, I've corrected people who have said that more times than I can count. But it will, over time, end up being about one out of thirty three attacks on average, because that's the actual definition of percentage chance.

winterthur

I am 'found' and I am lost.

Let's change the evade % to 97 instead of 3.

How should the fight work out?

DTMelodicMetal

winterthur wrote: »

I am 'found' and I am lost.

Let's change the evade % to 97 instead of 3.

How should the fight work out?

Easy for Iceman 😂

DNA3000

PolarisCV wrote: »

Also worth mentioning that 12 of those 33 hits came while she was stunned. So evaded 3 times out of 21 hits (including multi hit specials). Roughly 14%

I did not initially keep track of this, so I will take your word for it. Since no one else did it, just to close the loop on this (and as penance for that time I accidentally did the wrong calculation**) the odds of seeing at least three evades out of thirty three attacks is about 13.7% or a little less than 7 to 1 odds against. The odds of seeing at least three evades out of twenty one attacks is about 5.9% or a little better than 17 to 1 odds against.

If you're seeing that all the time, it would suggest the evade chance was bugged. But at those odds, everyone should be seeing that level of evading at least occasionally.

A hundred evades in a row would definitely be bugged. It isn't mathematically impossible, but the universe won't last long enough to see that sequence of events stochastically.

** the correct calculation is to calculate the odds of exactly zero, exactly one, and exactly two occurrences, and then subtract that from one. For 21 attacks, the odds of zero, one, and two are about 52.7%, 34.3%, and 7.1% respectively, and therefore the odds of getting either zero, one, or two evades is about 94.1%; the odds against this happening are about 5.9% which is the odds of getting three or more.

unknown

This content has been removed.

CFree

Icecold2099 wrote: »

Verzz wrote: »

Obviously if your sample size is one, it is either going to happen once or zero times if that is what you are asking

.

If this is your prediction of something with a 3% chance happening then i guess you're correct, you can predict it. It's going to happen once or zero times.

I think you're under the impression that i don't understand what 3% is and how it correlates into the chance for something. That's not correct, i understand that just fine.

You're also taking the 3% probability and using that as a determination that it equals about 1/33. This is a correct conversion of decimal to fraction, but has no barring on the discussion.

The OPs point, to my understanding, is how BW 3% chance to evade triggers at a seemingly higher percentage. Then another poster stated that a 3% chance she should evade about once every 33 attacks.

I think the issue is that you don't understand that, regarding BW chance to evade, the sample size is 1. So you can't predict if she is going to evade that one attack or not. Let's say she does evade that attack though, that does not give you any information or indication if she is going to evade the next attack or not.

Hence, you cannot predict it.

You can, however, take an educated risk that because there is a 97% chance she is not going to evade she probably won't, but you absolutely cannot say with certainty that she will or will not evade any given attack.

It does not translate into her evading about once every 33 attacks.

If you still don't understand that, having played the game, watched the OPs video of the game, read others accounts of the game, then you're just not going to understand it.

Can you respond to the 4:02 pm post from DNA3000?

Icecold2099

I was still on page 2 before my last reply.

When I originally read @DNA3000 first post that I was disputing, I read it as if he was stating that on average every 1/33 attacks she's going to evade, meaning if you've attacked her 32 times and she hasn't evaded, the next attack will trigger an evade. Or you'll get roughly 3 evades every 100 attacks, etc...

This is the thought process that all of my arguments were based on.

So when @Verzz stated I was wrong, I thought they were arguing the point of what I initially undersrood @DNA3000 to be saying. So I was stating you could not predict when she WAS going to evade. Basically, that in any given fight of 33 hits she could evade anywhere from 0-33 times. Very unpredictable...

Based on @DNA3000 clarification of what they meant, I can see how what they were saying makes sense if they were basing it on what the mentioned regarding excel spreadaheets.

I also am not a game designer and was under the impression that to program a 3% chance of something happening when something else occurs it was just something like "if this, then this at 3%" or whatever. So that everytime BW is attacked, while not stunned, there is basically an electonic roll of the dice deciding if the evade is going to happen or not. And thats not something that is predictable.

unknown

This content has been removed.

DNA3000

Icecold2099 wrote: »

Based on @DNA3000 clarification of what they meant, I can see how what they were saying makes sense if they were basing it on what the mentioned regarding excel spreadaheets.

I also am not a game designer and was under the impression that to program a 3% chance of something happening when something else occurs it was just something like "if this, then this at 3%" or whatever. So that everytime BW is attacked, while not stunned, there is basically an electonic roll of the dice deciding if the evade is going to happen or not. And thats not something that is predictable.

Just to clarify, it is typical in game development for a game like this for a lot of the development to actually occur in Excel: a champion is basically a bunch of rows of numbers, because champions are like Mad-libs: you fill in the blanks for every property, and that includes the properties of abilities. There's generally a bunch of scripts (often called a "toolchain") that converts those spreadsheets into text files, and then from text files into compacted binary files that get loaded into the game servers (and clients).

Ultimately, the game engine does what you describe: somewhere in there is a giant loop in the code that says, for every tick of the clock, check to see what's going on now, and for every ability that is executed, check its properties and do what those say to do. Every ability probably has a "chance to occur" field (it may in fact be called "AbilityAccuracy") and by default it is set to 1.0. BW's evade is (probably, for discussion purposes) set to 0.03, so the game does do a random roll to see if that ability should trigger. But this doesn't directly happen in code, it happens in "meta-code" that uses the data the content developers fed to the game engine.

winterthur

Verzz wrote: »

Ok this is coherent. Let’s agree on this. I agree that nobody can tell with 100% certainty on any sample size of one that has any random chance for more than one outcome.

Sinking into quicksand.
What is the difference then if the evade is 97%?

If confined to per hit, it should be yes or no only, 50%, no?

unknown

This content has been removed.

Helicopter_dugdugdug

So much facepalm in this thread ... wish i had evaded reading it

DNA3000

winterthur wrote: »

Verzz wrote: »

Ok this is coherent. Let’s agree on this. I agree that nobody can tell with 100% certainty on any sample size of one that has any random chance for more than one outcome.

Sinking into quicksand.
What is the difference then if the evade is 97%?

If confined to per hit, it should be yes or no only, 50%, no?

Just because there are only two possibilities, doesn't mean they are equally likely. There are only two possibilities, Sun comes up tomorrow, Sun doesn't come up tomorrow. But that's not 50%.

Probability, in both a colloquial and technical sense, is a measurement of uncertainty. It describes what we know about an outcome that hasn't happened yet and we can't yet observe. But once the thing happens, it isn't proper to talk about the odds of that already happening, because those odds are 100% - it did happen. We can only talk about the odds of something happening before it happens, or what the odds were prior to it happening and we didn't know.

When we attack Black Widow, before the attack happens we know there's a 3% chance she will evade before it happens. That means that, given identical circumstances, in approximately three out of a hundred times when we observe that situation BW will evade, on average, across sufficiently large numbers of observations. But any one particular attack either will or will not include an evade. Averages only make sense for large sets of measurements. There's no such thing as the average of one thing.

But once the attack happens, and the result is known, there are no more odds anymore. There is no 50% or even 3%. There is just history.

But to summarize, "odds" or "probability" is a way to express the fact that we don't know exactly what will happen, but we do know *something* about what will happen. We don't know if Black Widow will evade the next attack or not, but are aren't completely ignorant about the situation. We know it is far more likely that the attack won't be evaded than it will. The odds are just the numbers-way of expressing that knowledge with precision.

winterthur

DNA3000 wrote: »

winterthur wrote: »

Verzz wrote: »

Ok this is coherent. Let’s agree on this. I agree that nobody can tell with 100% certainty on any sample size of one that has any random chance for more than one outcome.

Sinking into quicksand.
What is the difference then if the evade is 97%?

If confined to per hit, it should be yes or no only, 50%, no?

Just because there are only two possibilities, doesn't mean they are equally likely. There are only two possibilities, Sun comes up tomorrow, Sun doesn't come up tomorrow. But that's not 50%.

...

But to summarize, "odds" or "probability" is a way to express the fact that we don't know exactly what will happen, but we do know *something* about what will happen. We don't know if Black Widow will evade the next attack or not, but are aren't completely ignorant about the situation. We know it is far more likely that the attack won't be evaded than it will. The odds are just the numbers-way of expressing that knowledge with precision.

Thanks for replying. Just amused that you use the sun as an example. I was thinking if the sun rises ... wait the Sun is in a 'fixed' position?

Anyway, my view (or understanding) is that over large data, the 'averages' is x%. So, if there is 'inconsistent' occurrences, then something else is in the game mechanic (a bug?)? Is this correct?

DNA3000

winterthur wrote: »

DNA3000 wrote: »

winterthur wrote: »

Verzz wrote: »

Ok this is coherent. Let’s agree on this. I agree that nobody can tell with 100% certainty on any sample size of one that has any random chance for more than one outcome.

Sinking into quicksand.
What is the difference then if the evade is 97%?

If confined to per hit, it should be yes or no only, 50%, no?

Just because there are only two possibilities, doesn't mean they are equally likely. There are only two possibilities, Sun comes up tomorrow, Sun doesn't come up tomorrow. But that's not 50%.

...

But to summarize, "odds" or "probability" is a way to express the fact that we don't know exactly what will happen, but we do know *something* about what will happen. We don't know if Black Widow will evade the next attack or not, but are aren't completely ignorant about the situation. We know it is far more likely that the attack won't be evaded than it will. The odds are just the numbers-way of expressing that knowledge with precision.

Thanks for replying. Just amused that you use the sun as an example. I was thinking if the sun rises ... wait the Sun is in a 'fixed' position?

General relativity says the laws of physics honor no preferred frame of reference in the universe.

Anyway, my view (or understanding) is that over large data, the 'averages' is x%. So, if there is 'inconsistent' occurrences, then something else is in the game mechanic (a bug?)? Is this correct?

That depends on your definition of "inconsistent." Almost by definition, "randomness" implies inconsistency. For example, if you're observing something like a coin flip that is supposed to have a 50% chance to land heads or tails, then the sequence H T H T H T H T H T H T H T H T H T H T H T H T has a 50% distribution, but is completely non-random. A genuine random generator will tend to generate almost any sequence with equal likelihood relative to other sequences of equal length. That means, counter-intuitively for most people, that H H H H T T T T happens just as often as H H T H T H T T. The second "looks more random" but it isn't. It is just that H H T H T H T T looks to human eyeballs similar to H T H H T H T T so we tend to lump those sequences all together, and because there are a lot of them they collectively come up more often.

This isn't strictly speaking correct, but for our purposes, randomness implies unpredictability. If we saw that every single set of 33 attacks against Black Widow included an evade, that level of consistently would actually be proof that the game wasn't properly triggering BW's evade randomly, and had been rigged to generate that rate of evade, using something like a streak breaker or a quota bucket. Genuinely random means sometimes you'll see a lot less evades than the averages predict, and sometimes you'll see a lot more. But the distribution of how often you see these deviations is itself something that is statistically predictable.

Welderofortune

So what Kabamike is saying is that a champ with 97% chance to cause bleed on every hit could go 100 hits without causing a bleed once as the odds do not matter according to his post so what is the point of posting them if they are not verifiable. If someone could post a summoner with a black widow as profile I'll run thru five duels and keep track of evades and hits, if a couple of us would do that we should get a pretty large sample size, and possibly put the matter to rest.

DNA3000

Welderofortune wrote: »

So what Kabamike is saying is that a champ with 97% chance to cause bleed on every hit could go 100 hits without causing a bleed once as the odds do not matter according to his post so what is the point of posting them if they are not verifiable. If someone could post a summoner with a black widow as profile I'll run thru five duels and keep track of evades and hits, if a couple of us would do that we should get a pretty large sample size, and possibly put the matter to rest.

The statistical margin for error for such a test would be about the square root of the expected result. In other words, to have a reasonable confidence level (statistical measurements are never "certain") that you were roughly within 10% of the correct value, you'd need to perform a test where the expected result was about 100 evades (which would then have an estimated margin for error of about 10), which would require a test of about 3000 attacks.

BigDaddyJoe

When BW evades once every 5 hits and she’s got only a 3% chance 🙄

unknown

This content has been removed.

unknown

This content has been removed.

KobeLives

The probabilities also are that we might go 200 hits without getting an evade. But that has never ever happened to anyone