I am not ready to do that. I'd really want to independently confirm the results, either with another tester, or simply do it myself before I took this to the general public. The OP is welcome to put this on reddit, but personally I'd hesitate right now because this kind of a post would cause a firestorm, and thanks to the contradicting data, I'm not even convinced yet. GoboRab's data set might be smaller, but t=100 is already plenty for asymptotic results to hold. We don't need 1700 data points for this kind of thing. This is just a mean calculation, it is not a test that lends itself to poor results in terms of predictive power.

To test the distribution itself I might need way more points, but we aren't there yet. And if the RNG is truly biased, who cares what distribution it has.

UPDATE: I'm in the midst of collecting data for a 26 land deck with no deck manipulation outside of "scry" and "looting". I'm around 20 games in now, all against random opponents on the ladder. From what I've noticed so far, it would seem that there's a chance that any kind of deck manipulation (besides scry and looting) has the potential to make the RNG short circuit, as I haven't had this little of trouble with land drops in weeks (not kidding). Most of my other decks contain 25 lands (including 4 evolving wilds and other card fetchers/manipulators) + plenty of card draw, and I either flood out or get stuck on land drop 3 for eternity. A "lucky" streak like this has happened to me before though, but it's always been short lived, and then a couple days later its back to flood city or stuck on land drop 3 for eternity. If this holds up for around 50-100 games, it would be a FIRST for me.

No it wouldnt.

And if it did that would be a good thing since that would demand some attention from the devs.

Someone post it on reddit please, and report it as a bug to the customer support.

Here's an interesting fact for those who like this sort of stuff.
The number of unique sequences(or shuffles) for a duels deck ( 60 factorial ) is actually more than the estimated known observable atoms in the universe ( between 10^78 to 10^82 )

Wrap your head around that !

Just to confirm, this is is premised on Vt2wa playing 1600 games with a deck with zero deck manipulation and accepting auto losses by refusing to mulligan zero land hands? 1600 games like that? Or, at 4 games per day, 400 days straight doing that?

And yes, posting a thread about the RNG will usually get you insta-banned and/or sharply criticized by the subjective folks on reddit (this is suspicious to me). BUT, I don't know if anybody has ever came prepared with a bunch of numbers.

can we get back to religion please?

Absolutely. What do you think about the people who worship Magic: The Gathering as though it's some kind deity, rather than just a game? I personally think it's ridiculous.

GoboRab did his 100 in like 2 days, I don't see a problem. There were some that wouldn't have made vt2's list, but no, it wouldn't take 400 days.

why dont you just do mulligans vs AI? draw 7, how many lands... draw 6, how many lands.. etc

That's 50 games of Duels a day. Even assuming every match ends in 5 minutes, that's more than 4 hours of consecutive gameplay, just spamming some hyper fast deck.

It's not impossible, but it sounds ****'ing terrible.

Weren't those the tests with the 26 land deck? If so it probably wasn't something hyper fast. Then again, IDK so I can't say for certain. It does sound face smashingly boring though either way.

I feel like that is sort of the problem. I wouldn't half mind recording some data for a 26/25/24 land deck for DJ and Gobo to process. The problem with this is creating decks that can function without any deck manipulation that don't just die on the spot to the current meta, or are just face numbingly boring to play more than a few games with.

Maybe later tonight I will muck around with trying to put together a couple decks that fit the criteria so I can start recording data for the math wizards.

well the mulligans counted toward the sample, and I'm not afraid to toss hands back typically. But yes, that + noting the data took roughly 4 hours (possibly longer since I was typically alt tabbed out only to tab back in and make my play), and it wasn't pleasant

But if someone does half an hour of this every 2 days, you can get another 100 within two weeks

And then in two weeks when I have the info, the topic will be dead and my data ultimately useless.

One day, a terrible fire broke out in a forest : a huge woodlands was suddenly engulfed by a raging wild-fire!
Frightened, all the animals fled their homes and ran out of the forest.
As they came to the edge of a stream they stopped to watch the fire and they were feeling very discouraged and
powerless.
They were all bemoaning the destruction of their homes. Every one of them thought
there was nothing they could do about the fire, except for one little hummingbird.

This particular hummingbird decided it would do something. It swooped into the stream and picked
up a few drops of water and went into the forest and put them on the fire. Then it went back to the
stream and did it again, and it kept going back, again and again and again.
All the other animals watched in disbelief; some tried to discourage the hummingbird with comments like, "Don't bother,
it is too much, you are too little, your wings will burn, your beak is too tiny, it’s only a drop, you
can't put out this fire."

And as the animals stood around disparaging the little bird’s efforts, the bird noticed how hopeless
and forlorn they looked.
Then one of the animals shouted out and challenged the hummingbird in a
mocking voice, "What do you think you are doing?"
And the hummingbird, without wasting time or losing a beat, looked back and said, "I am doing what I can."

****, that's motivational. Okay, I'll try to record my info. I suspect most of my info will be for a 24 land deck, since that's my tourney deck I'm testing anyway (which is lucky enough to have no actual deck or mana manipulation).

It's actually been done before, and the results pretty well showed there was nothing obviously wrong with the shuffler results. Something like 400 trials that time, which is plenty.

I think you answered your own question. There have been plenty of people commenting about strange RNG, but non of them (that I am aware of) have had any data to back it up at all. There are plenty of people like this: https://www.reddit.com/r/magicduels/com ... o/dbxeqk8/

If you don't have any numbers, you can't really say much about the duels shuffling system. If there was a great conspiracy to keep the shufflers flaw hidden, I would expect some of those facts to have ended up either here or on the wiki at this point. It's not like the reddit is owned by WOTC.

As for when this is reddit worthy, I mostly want it published so I have somewhere to link to when there are more people with crazy ideas of how the shuffler works showing up. P > 2e-20 seems fine.

Just for the record, do you understand that the 1700 trials in the OP, are 1) data, and 2) show conclusively that the shuffler is heavily biased toward land flood in the posted results? The whole reason for this thread is that someone did gather a large amount of data, and it's pretty damning.

I am checking if we can corroborate his data, and asking some people to rerun trials. So far we have 1 extra trial which conflicts with the 1700 trial data, but that's not enough. If I were to post on reddit, right now, the post I would make is a) the shuffler is obviously broken, based on 1700 trials, and b) here is the data that proves it conclusively. My only hesitation is that it isn't my data, and for something this earth shatteringly surprising I want further confirmation.

The reason for all of this: This is not how a RNG shuffler is expected to fail - unless the programming is beyond screwed up. Repeated hands? sure (but you'd be highly unlikely to see it even once within just 1700 hands). Possible bias in the tail results? maybe, this could happen from poor calibration. Complete statistical divergence from the mean of the population? no, that's just simply not supposed to happen. That last one should be impossible.

I should also mention the OP's data had 112 incidents of 100% land out of 1700 trials (with varying amounts of land drawn). That never happened in GoboRab's data. By comparison OP's data had only 24 events at 0% land (which is the higher probability event), and Goborab had only 1 (which is about what we'd expect in 100 trials, or at least it isn't an odd amount or anything, 0, 1, or 2 would all make sense and not surprise me too much).

I just specifically built a deck for this and had an error wreck it so I couldn't save it. Which is a bit annoying. But I'll just build it from scratch and memory when I get home I suppose.