Jump to content

RNG Experiment: Attack Coinflips


Sakura150612
 Share

Recommended Posts

What difficulty AI was the coin-flip data collected on?

 

I only ask because when I was completing the Trainer Challenge back in the day using the Plasma Vanilluxe deck I remember coin flips being more lopsided in the favour of the AI when the difficulty was set to maximum (attacking with 5-10 energy on Vanilluxe resulted in about 35% flips for heads after 200~ data samples) - I didn't mind because I was still stomping the AI into the ground and figured the game modified those flips to simulate a more difficult challenge.

 

However; the PTCGO was still in it's infancy at the time (Plasma Freeze had just been released and the Frost Ray theme deck finally gave me the motivation to finally collect my 45 free packs) so the coin flip system has likely been changed since then.

Link to comment
Share on other sites

What difficulty AI was the coin-flip data collected on?

 

I only ask because when I was completing the Trainer Challenge back in the day using the Plasma Vanilluxe deck I remember coin flips being more lopsided in the favour of the AI when the difficulty was set to maximum (attacking with 5-10 energy on Vanilluxe resulted in about 35% flips for heads after 200~ data samples) - I didn't mind because I was still stomping the AI into the ground and figured the game modified those flips to simulate a more difficult challenge.

 

However; the PTCGO was still in it's infancy at the time (Plasma Freeze had just been released and the Frost Ray theme deck finally gave me the motivation to finally collect my 45 free packs) so the coin flip system has likely been changed since then.

Well i didn't go against the AI in Trainer's Challenge where you can set difficulty, but rather in Deck Test Mode in Deck Construction, so i figure the difficulty level is standard. Also i don't feel inclined to believe that the difficulty leven in TC affects the RNG not even for coin flips.

Link to comment
Share on other sites

What difficulty AI was the coin-flip data collected on?

 

I only ask because when I was completing the Trainer Challenge back in the day using the Plasma Vanilluxe deck I remember coin flips being more lopsided in the favour of the AI when the difficulty was set to maximum (attacking with 5-10 energy on Vanilluxe resulted in about 35% flips for heads after 200~ data samples) - I didn't mind because I was still stomping the AI into the ground and figured the game modified those flips to simulate a more difficult challenge.

 

However; the PTCGO was still in it's infancy at the time (Plasma Freeze had just been released and the Frost Ray theme deck finally gave me the motivation to finally collect my 45 free packs) so the coin flip system has likely been changed since then.

 

I did it with the "Test your deck" option in the Deck Manager, so I don't get to chose the difficulty setting. It could default to the highest level, but to be honest I have absolutely no idea what the difficulty level is. Although I cannot speak for the state of the game 4 years ago in that particular scenario, I know that in its current state the game does not have any issues with its coin flips in any of its forms.

  • Upvote 1
Link to comment
Share on other sites

Well i didn't go against the AI in Trainer's Challenge where you can set difficulty, but rather in Deck Test Mode in Deck Construction, so i figure the difficulty level is standard. Also i don't feel inclined to believe that the difficulty leven in TC affects the RNG not even for coin flips.

I think, from experience, that the RNG in the TC is skewed on the higher difficulties, and maybe the Test Deck automatically puts you against the highest difficulty AI, though I'm not sure about this. If the RNG is rigged on higher difficulties, then your data would show that the Test Deck puts you against a Normal AI.

If you have time, could you look into the TC further? Though the TC RNG may have changed in the update.

Edited by theotherguytm
Link to comment
Share on other sites

If you have time, could you look into the TC further? Though the TC RNG may have changed in the update.

What do you mean? Make a test for the coin flips in TC? That would be harsher and take an amount of time i don't have to invest right now, as pokémon's HP are lower and the attacks flip less coins. The thing with Malamar-EX or the Gyarados Sakura used is you can actualy flip a lot of coins in one single attack, thus reducing the experiment time.

I feel inclined, though, to make a 100/200 coin flip test in Versus Mode with this Archie's Blasto/Malamar-EX deck, adding a few twists to make it more reliable and 'competitive', because the deck i took to Test Deck Mode is quite unreliable. When i have the time to do it i'll post the results here. Going against real life players as Sakura did is the ultimate sample we can get.

Link to comment
Share on other sites

This is a very good Post. I enjoyed reading it even though I literally had to read it slow because I'm not good at math. Here the thing though. It isn't the RNG that I'm concerened about. I'm concerned about the shuffling and the matchups in PVP. The shuffling system needs to be adjusted. You know how many times I started off with 2 or more mega pokemon in my hand in the first turn? Way to many for it to be a fluke. Also I promise you I played with the same deck for like 40 games. Then I made a new deck without EX in it................when I clicked on PVP for some reason it took a minute to find my opponent.........which it usually finds my oppoenent right away...............and low and behold my oppoenent is literally playing the counter to my deck. Cmon!!! Seriously! You know how many times that happened when I played the same deck like 10+ times and never run into a certain type, then I switch decks and I run into my complete counter. I know for a fact it bases my opponents on the type of deck I'm running. I wish I had the patience and time to do a study like Sakura did but I'm too busy trying to survive in pvp facing a type I'm weak against. Anyway.

 

By the way Sakura..................I challenge you to a match. You must be good and I love a challenge. Hit me up. I'll send friend request.  

Edited by bulbasaur32277
  • Upvote 2
Link to comment
Share on other sites

This is a very good Post. I enjoyed reading it even though I literally had to read it slow because I'm not good at math. Here the thing though. It isn't the RNG that I'm concerened about. I'm concerned about the shuffling and the matchups in PVP. The shuffling system needs to be adjusted. You know how many times I started off with 2 or more mega pokemon in my hand in the first turn? Way to many for it to be a fluke. Also I promise you I played with the same deck for like 40 games. Then I made a new deck without EX in it................when I clicked on PVP for some reason it took a minute to find my opponent.........which it usually finds my oppoenent right away...............and low and behold my oppoenent is literally playing the counter to my deck. Cmon!!! Seriously! You know how many times that happened when I played the same deck like 10+ times and never run into a certain type, then I switch decks and I run into my complete counter. I know for a fact it bases my opponents on the type of deck I'm running. I wish I had the patience and time to do a study like Sakura did but I'm too busy trying to survive in pvp facing a type I'm weak against. Anyway.

 

By the way Sakura..................I challenge you to a match. You must be good and I love a challenge. Hit me up. I'll send friend request.

 

Hey it's great to see you're back bulbasaur xD I'll be happy to take the challenge, so whenever we catch each other online we can play.

 

Regarding the matchmaking, that's probably the one thing I won't be able to properly investigate. The main issue it has which the other forms of RNG lack is that there are too many unknowns and that I have essentially no parameters to compare the results with. The reason I did coin flips first is because they are very straight forward; you expect a 50-50 split in the long run and each subset of coin flips can be determined as either normal or anomalous with a simple formula (1 divided the square root of the sample size). With matchmaking, while the game randomly picks from the player pool, said pool is restricted by MMR, and while it's technically not a factor, it's restricted by what people are using at the moment (as in, a very hyped deck will normally see more play).

 

As I have no theoretical way of knowing the actual proportion of people playing each deck type, even if I gathered results they wouldn't mean much. What I CAN do to test is use 2 or more different decks in the same format, play them for at least 30 matches each and compare. But to do this, I wpuld have to do it under similar conditions, and ideally play all the games in a single day, as that would give me the least amount of variance in the data. I might tackle that one in my winter vacations (I live in the southern hemisphere) when I can actually sit down and play 100+ games in a single day xD

 

Shuffling, on the other hand, has no real unknowns. It is possible to calculate theoretically the chance of pulling any given card at any given specific situation. The reason I haven't done it is because the math behind that is a nightmare xd

  • Upvote 1
Link to comment
Share on other sites

This thread was getting a little off topic and it had to be cleaned up a bit.

So this is a quick general reminder to everyone: to stay on topic.

Link to comment
Share on other sites

I ran a small sample last night using the PlasmaFreeze "Frost Ray" deck against 4-star difficulty AI opponents:

 

                                 KarpGrinder                AI Opponent
Total Coin Flips               172                               36
Heads Flipped                 71                                18
Heads %                       41.2%                             50%

 

 The flip ratio definitely felt better than I remember from years past and the AI flip rate seemed to be far more balanced - but I'll collect more data this evening.

 

EDIT: BBCode Table writing is weird.

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

  • 4 weeks later...
  • 4 weeks later...

I have a rule for item cards that require a coin flip to activate; I'll only use them once they have been in my hand for one turn. I have MUCH higher success rate with these cards when used in that way. Which makes me think that the RNG calculation is based on time in play. e.g. a numeric value representing the time the card has been in play (either in seconds or number of turns), is used in the sum to calculate the heads/tails, odd/even result.

Link to comment
Share on other sites

I have a rule for item cards that require a coin flip to activate; I'll only use them once they have been in my hand for one turn. I have MUCH higher success rate with these cards when used in that way. Which makes me think that the RNG calculation is based on time in play. e.g. a numeric value representing the time the card has been in play (either in seconds or number of turns), is used in the sum to calculate the heads/tails, odd/even result.

It's not.  Sorry, but it just isn't.

Link to comment
Share on other sites

I have a rule for item cards that require a coin flip to activate; I'll only use them once they have been in my hand for one turn. I have MUCH higher success rate with these cards when used in that way. Which makes me think that the RNG calculation is based on time in play. e.g. a numeric value representing the time the card has been in play (either in seconds or number of turns), is used in the sum to calculate the heads/tails, odd/even result.

Whatever helps you sleep at night mate :D.

  • Upvote 1
Link to comment
Share on other sites

  • 2 weeks later...

So...

Today it happened for me to be at a friend's house and i was present in the room while he was playing some PTCGO matches.

Long story short, i witnessed a Tapu Lele GX who it was using a Fury Belt getting knocked out by 8 consecutive heads coins flips from a Pyukumuku.

The outcome of the attack was 240, which was the exact amount of damage needed in order to knock out the card.

 

Don't get me wrong but Pyukumuku isn't the case.

Opening hands are the case.

Coin flips are the case.

RNG totally ignoring a deck's consistency is the case.

 

Maybe TPCi is fine with Sakura's experiment as it serves their agenda of the RNG being totally fine, maybe rest of the PTCGO community is fine with Sakura's experiment as they aren't into performing such a dedicated experiment themselves, but i'm not fine either with Sakura's experiments or TCPi's agenda.

 

The RNG is broken, i see it everyday within matches and if you think what i say is absolutely wrong then prove me wrong by showing the RNG code publicly.

Unless Sakura had access to the RNG's code, then i can also say this experiment is absolutely wrong too.

 

I am sick and tired of the RNG as PTCGO is all about luck and winning isn't based on deck consistency.

Edited by The_Real_Bug
Link to comment
Share on other sites

Come on, man; you know statistics don't work like that.  This would only be a problem if that happened every time with pyukumyuku, but it doesn't.  Your friend got unlucky, but that means the RNG is working right- not that it's working wrong.  The chances of heads are 1/2 to the nth power, in this case 1/256.  It's a low chance, but it exists, and it happens.  Unless you can show that these cases happen significantly more than once very two hundred fifty two times, there isn't a problem.

 

Innocent until proven guilty, no?

  • Upvote 1
Link to comment
Share on other sites

So...

Today it happened for me to be at a friend's house and i was present in the room while he was playing some PTCGO matches.

Long story short, i witnessed a Tapu Lele GX who it was using a Fury Belt getting knocked out by 8 consecutive heads coins flips from a Pyukumuku.

The outcome of the attack was 240, which was the exact amount of damage needed in order to knock out the card.

 

Don't get me wrong but Pyukumuku isn't the case.

Opening hands are the case.

Coin flips are the case.

RNG totally ignoring a deck's consistency is the case.

 

Maybe TPCi is fine with Sakura's experiment as it serves their agenda of the RNG being totally fine, maybe rest of the PTCGO community is fine with Sakura's experiment as they aren't into performing such a dedicated experiment themselves, but i'm not fine either with Sakura's experiments or TCPi's agenda.

 

The RNG is broken, i see it everyday within matches and if you think what i say is absolutely wrong then prove me wrong by showing the RNG code publicly.

Unless Sakura had access to the RNG's code, then i can also say this experiment is absolutely wrong too.

 

I am sick and tired of the RNG as PTCGO is all about luck and winning isn't based on deck consistency.

I agree with this. I'd say I lose %98.76 of coin flips (random number, just means I lose most of the time), and it really is annoying. I can't even go first if I want to, as the coin flips never agree with me.

 

I'm not saying that the RNG is broken (or am I?) but I mean seriously, what are these hands with 4 VS Seeker + Sycamore? These hands with 4 DCE? Or just bad hands in general, there really hadn't been a game where I haven't had this happen.

Link to comment
Share on other sites

I agree with this. I'd say I lose %98.76 of coin flips (random number, just means I lose most of the time), and it really is annoying. I can't even go first if I want to, as the coin flips never agree with me.

 

I'm not saying that the RNG is broken (or am I?) but I mean seriously, what are these hands with 4 VS Seeker + Sycamore? These hands with 4 DCE? Or just bad hands in general, there really hadn't been a game where I haven't had this happen.

Yeah it is annoying when you have a bunch of DCE/vs seeker and a sycamore as your only draw support. Makes you really salty even when the RNG is supposedly working right.

Link to comment
Share on other sites

Anyone who thinks there's a problem needs to prove it. Anecdotes are not reliable data. Record every game.

  • Upvote 1
Link to comment
Share on other sites

Fixed that for you ;)

So...

Today it happened for me to be at a friend's house and i was present in the room while he was playing some PTCGO matches.

Long story short, i witnessed a Tapu Lele GX who it was using a Fury Belt getting knocked out by 8 consecutive heads coins flips from a Pyukumuku.

The outcome of the attack was 240, which was the exact amount of damage needed in order to knock out the card.

 

Don't get me wrong but Pyukumuku isn't the case.

Opening hands are the case.

Coin flips are the case.

RNG totally ignoring a deck's consistency is the case.

 

Being biased towards a small sample size is the problem

Ignoring normal outcome, due to selective memory is the problem.

Believing that a single game out of ten-thousand is somehow a viable stat is the problem.

 

Maybe TPCi is fine with Sakura's experiment as it serves their agenda of the RNG being totally fine, maybe rest of the PTCGO community is fine with Sakura's experiment as they aren't into performing such a dedicated experiment themselves, but i'm not fine either with Sakura's experiments or TCPi's agenda.

 

The RNG is broken, i see it everyday within matches and if you think what i say is absolutely wrong then prove me wrong by showing the RNG code publicly.

 

Or someone could dedicate some time to test the RNG.... you know, with a large sample size that uses methods of statistics, rather than gut feeling ( which is the worst method of measurement know to mankind)... yeah, wish someone would do that....

 

Unless Sakura had access to the RNG's code, then i can also say this experiment is absolutely wrong too.

 

I am sick and tired of the RNG as PTCGO is all about luck and winning isn't based on deck consistency.

 

Oddly enough I still maintain a 80-85% win rate with almost all of my decks and I know a bunch of players in the same range. The RNG was never broken, it just boils down to the same things every time:

Players like to boil down an entire match to a single instance of RNG “Oh well, if I had only hit that coin flip there I'd have won. Stupid RNG.” That's not how the game works.

 

Players tend to throw away games rather carefree, if they get a bad start, thus loosing any opportunity to get better at handling those situations. “Shaymin-EX is my starting Pokemon ? Stupid RNG, concede”.

 

Players love to ignore any instance where RNG favours them, as well as any instance of regular RNG. “That Pyukumuku flipped heads 8 times in a row!” → proceeds to ignore the last 20 times you played against the card, where the attack didn't do any major damage.

 

Ranting is fine, but at this point in time it is up to the RNG critics, to provide any kind of usable data, that'd proof their point of view..

 

 

  • Upvote 2
Link to comment
Share on other sites

No Felidae, you didn't fix anything for me.

Turning Sakura's experiment into godlike material that we should all bow down to and never question isn't the case.

 

The RNG is broken and no one will ever admit.

Sakura's experiment didn't prove anything as much as i can't prove what i'm saying.

Link to comment
Share on other sites

Innocent until proven guilty, no?

That is the whole point.

 

Since it hasn't been proven otherwise, the RNG is innocent of being broken.

 

BUT

 

That doesn't prove that it's working fine! Or that is perfect. Or that it wouldn't need any improvement.

 

Sakura was patient enough to conduct that experiment and that is admirable. Until a next one is conducted that shows different results we have to accept it as a benchmark to all discussions.

  • Upvote 4
Link to comment
Share on other sites

Since it hasn't been proven otherwise, the RNG is innocent of being broken.

 

BUT

 

That doesn't prove that it's working fine! Or that is perfect. Or that it wouldn't need any improvement.

 

 

Read this.

Repeat.

Repeat.

Repeat.

 

Now you are wiser.

Edited by The_Real_Bug
Link to comment
Share on other sites

If anyone would like to, we can set up coin-flip based unlimited deck and collect data, I'll be on around 19:00 PDT (approximately 12 hours from the time of this posting).

 

Refined coin-flip data can be collected on the following (but not limited to):

•Coin flips on certain item cards.

•Coin flips based on Turn #.

•Coin flips based on certain status conditions.

•Coin flips on the remainder of a game dependent on the opening (first turn) flip.

•Coin flips based on the 'Advantage' tag.

•etc.

 

Otherwise: carry on with unfounded argument.

  • Upvote 2
Link to comment
Share on other sites

If anyone would like to, we can set up coin-flip based unlimited deck and collect data, I'll be on around 19:00 PDT (approximately 12 hours from the time of this posting).

 

Refined coin-flip data can be collected on the following (but not limited to):

•Coin flips on certain item cards.

•Coin flips based on Turn #.

•Coin flips based on certain status conditions.

•Coin flips on the remainder of a game dependent on the opening (first turn) flip.

•Coin flips based on the 'Advantage' tag.

•etc.

 

Otherwise: carry on with unfounded argument.

 

Unless we get a chance to look at the RNG code itself, this conversation will go on and on.

Link to comment
Share on other sites

 Share

×
×
  • Create New...