BIGGEST EDISON FORMAT TIER LIST EVER - Over 100k Matches Analyzed!

[@xxsaberwarrior564]

As a dragon turbo player with locals success, people only see your good hands and complain about it being unfair, but completely disregard the games where you legit brick twice in a row. It is a true gamer/pain experience

[@redgadgetyugioh]

This is also how I feel about it honestly. The times you lose to yourself are just too frequent. Sucks 😪

[@SC2BETTER]

Imo, i think this deck play solo mode, win or loose, there is skill expression in this deck for sure, but i don't remember many interactive games vs dragon turbo lol

[@AB-ds3zl]

i feel like DRAGON TURBO is the edison version of REASONING GATE OTK, high variance, high brick potential but can easily OTK if everything goes right

[@saqlaq96]

This is some next level analysis, great work!

[@redgadgetyugioh]

Appreciate the kind words. 😊

[@TristynRuiz]

Insane level of content. You’re gonna be big. Keep going.

[@redgadgetyugioh]

I was like: "I dont have time lets just show this one video and focus on my studies... 😂", but I think now I have no choice but to continue! Thanks!

[@Ron-di6zy]

Wow, well done Mak.

[@redgadgetyugioh]

Thanks Ron! 🤙🤙

[@Kirin_the_Bouncing_Unicorn]

Damn, I've been saying that Machina feels like it's the most underrated deck of the format. Obviously it's also one of the easier decks to play which pushes the WR up but I'm glad to see that my gut feeling was somewhat correct, the deck is STRONG.

[@redgadgetyugioh]

Machina feels extremely strong. I have been also saying it for a while now and I am also happy the data confirms it lol. Great minds think alike 😉.

[@wonky1k]

I think a big reason DT was so low is random DB players playing the deck. Shit is next level difficult when u have 10 cards in hand and if you discard the wrong 4 you might lose in 3 turns cus of it and you’re just like what did i even do wrong

[@redgadgetyugioh]

I agree with this. This is probably one of the most likely explanations... Well put.

[@LimiterRemoval]

Another explanation could be that no deck bricks as much as this one does, I would assume🤷🏻‍♂️

[@redgadgetyugioh]

@@LimiterRemoval 🧱-city

[@ricardomora4531]

Isn't that complex is just a deck that bricks a lot for that is garbage

[@ajippov]

Great video! I'd be curious to learn how you gathered the data. Obviously, you can pull up any replay if you know the replay id, but actually knowing which ones to look for is the key. Now that I think about it, even if you do know the id, I don't know that you can easily extract deck information from it. Would be curious to hear how you went about this. Look forward to more from you in the future!

[@redgadgetyugioh]

Hi, Given a replay data, I wrote a python script that parses the data and converts it to a json object. Given the object, you can do whatever you want and do any analysis you want. You have to infer the deck played given the cards played. That is the tough part and was covered in the first part of the video. Hope it helps 😁👋

[@anaelyahi6095]

Great video, always love to see some data analysis! Is the raw data available somewhere? I did a bunch on Magic: The Gathering, and I would like to check whether my tools would work here too, since I would like to start playing Edison and resume playing YGO. So I must say I am really glad to find such a video published right on time. In particular, I would love to compute some confidence intervals on the win rates, since it is my main metric (especially the more robust formula from Bell and McCaffrey that accounts for player clusters which might skew the results). I can also provide some inputs on your new project ideas since I already implemented some of them in MTG (decklist optimization, tier list theory, MU matrices...).

[@guto5009]

I own machina gadget and quickdraw. I played my friend to introduce him to the format and we played like 6 games, he was on gadget and he won all the games lma9

[@redgadgetyugioh]

Hahaha Gadgets aint that bad 😅

[@iwishuglhf]

It would be amazing if you could make a Video on a Matchup Spreadsheet? Check for 100k Matches wich Decks typically lose to wich Decks

[@redgadgetyugioh]

Hi, Interesting idea. I can definitely make that no problem 😊. Just gotta find time 😂

[@danielvasco6687]

Amazing video

[@redgadgetyugioh]

I appreciate it 🤟

[@OMGVLS]

This is so good. I love this kind of content.

[@redgadgetyugioh]

Wow, this is amazing to hear! I am happy you love it!

[@Dylan-oy6pc]

It would be really interesting to see the data with even more of a time lapse. It would also be interesting to see what the win rates are at only high rating

[@jasonk5335]

absolutely based. This is definitely a tier list for average level players vs a very broad meta. Liked and subscribed, would love a followup video showing the meta percentages of each deck in the data analyzed. Will def be interested in any useful data analysis of edison if you want to go further. Matchup chart, side deck cards, efficacy of controversial card choices. I'll dm you on discord.

[@redgadgetyugioh]

Thanks a lot for the love. Yes that is true, and I think it is needed in the community. To shed some light on the things discussed and the things you mentioned. Will look into your suggestions definitely!

[@loganminto1020]

Wow, awesome video! Would be very interested in the siding video, but also would like to know how diva hero compares to normal hero beat, how ancient gear variants of machina stack up, etc

[@redgadgetyugioh]

Noted! Thanks for the feedback and thanks for watching. I appreciate it 😉😁

[@JORGEMILLAR-kg2fr]

thats great work man, a tier list that takes into consideration the level of the player could be more accurate later

[@redgadgetyugioh]

That is for sure true, I agree as well!

[@TeamFlelf]

Yo awesome content, it's no surprise people want to see more! I'd love to see matchup spreads and individual card winrates when drawn vs played. Possibly a data split between top players? Maybe the top 200 leaderboard vs the rest of the pool and how it affects winrates. Similar to HSReplay for Hearthstone. They would let you sort by rank/deck winrate/decklist and it was a great tool to understand decks/card strength. Data like this for the community is an amazing addition, thank you for sharing your hard work :)

[@redgadgetyugioh]

Hi, Those are some really interesting ideas. I never played heartstone, so I would not know, but I think it seems like a great implementation. I have noted down all your suggestions and will consider them for next videos. Thank you very much for your kind words 😁, and stay tuned for more!

[@MrVovoda]

Thank you for making this ! I have tons of questions. First is what was the deck representation in those 100k+ games ? I feel like it's very important as a first step towards matchup discussion, as 55% winrare is only important in regards to the overall field its matched against. And second is how many decks were not in those listed here ?

[@redgadgetyugioh]

Hi, Thank you for watching. The representation was honestly relatively similar and there was no deck severely underrepresented. Regarding the decks left out, well there is basically many as I just classified deck that do not fit the decks listed as other. Hope it helps and thanks for watching ❤️

[@Railmydad]

Finally finished it but very good video! Please keep them coming, I’ve been asking more people to run data and show the numbers for which cards are better performing etc…

[@redgadgetyugioh]

Exactly, we need such things in the community. Card by card data is something people are interested in so I will add it to my list of ideas. Thanks and cheers!

[@MrBigNoodle]

Please do this for Goat Format!

[@redgadgetyugioh]

Hahahahhaah I am not big into Goat. I will consider it though! Thanks for the feedback ❤️😁

[@AB-ds3zl]

Maybe theres a chance you could help someone do a similar video for goat.would really appreciate it🙏🏻🙏🏻

[@zauls.]

Aweome work! An Edison stockfish would be insane but I imagine would be super hard. I'm not a programmer or anything but yugioh would surely present a lot of challenges compared to chess, which has perfect information, and every game starts with the same pieces laid out on the same squares. Would be really interesting to see how an AI could navigate something with so many more variables.

[@redgadgetyugioh]

Yes exactly, one would have to actually define what yugioh is. This is extremely hard as people who play for 10 years don't know most of the rulings. I think the best thing we have is one of the ygopro simulators or we just wait for master duel code leak 😂.

[@paultissimo9328]

About Transmigration Prophecy, there are some limitations with the card 1. Because Transmigration is a backrow, it can be destroyed with stuffs such as Lyla and MST, forcing you to use your Transmigration at suboptimal time 2. Because Crow is a Dark monster, it is great at manipulating Dark counts for your DAD. Also, you can Allure it and use it for Chaos Sorcerer fodder. You can also use it as a pitch for Dark Grepher when you need to stack your Graveyard. 3. Because Crow is a monster, you can do some interactions that you cannot otherwise do with Transmigration (Pot of Avarice, Sangan, Drill Warrior, retrieving it back from Alchemist's effect for Fairies) Transmigration is better into DDV, something that some Deck struggles vs. However, there are reasons why those Decks utilize D.D. Crow more. For example, Quickdraw struggles against DDV, but being able to reuse those Crows with Avarice and Drill.

[@redgadgetyugioh]

Interesting take. I think those are really valid points, but it being chainable and knowing you can only set it when you know you will get value, seems very strong. Against vayu if you shuffle plague and vayu, suddenly their warriors did nothing. 🤷‍♂️

[@rn5439]

Thank you for doing this! Really sick data

[@redgadgetyugioh]

Thank you for watching. Thanks a lot!!

[@FIIRdesu]

One statistic that would be interesting to see in the future: Divide the sample by player winrate, so we can compare for example Dragon Turbo played by top 10% players with bottom 10% players and with average players.

[@redgadgetyugioh]

Interesting suggestion. I think it totally makes sense. Will look into it 🤟

[@emiliodemarchi9833]

Great content

[@redgadgetyugioh]

@@emiliodemarchi9833 thanks 😊

[@Otzkar]

i would like to see some card statistics. what are the most played cards. what cards ahave the highest win rate, what ratio are cards played in and so forth

[@redgadgetyugioh]

This has been suggested many times and the next video will likely be as such. Not sure, but stay tuned 😉

[@lukasdrobe368]

Missed the Times where people Just building There own cool Decks instead of netdecking the shit Out of the Game and sweating in Casual games😢

[@redgadgetyugioh]

I co-sign this!

[@Brianpeppers15]

For each deck type, try to include the N# matches in your data pull. For example since dragon turbo is so low, maybe the #matches were much less, which might indicate it's a deck played less often or just didn't get played that much in your data pull time frame. Nice video sir

[@redgadgetyugioh]

Hi Brian, That is an interesting take. I already have this data, but I decided to not include it because it is not requested by less technical people unlike yourself. I would still try to make the videos more simplified and abstract some parts, but maybe I could add more info in the description for viewers like yourself. Cool feedback thanks!

[@mryaga3642]

Great video, keep them coming. Looking forward for a video on best side deck cards and even a similar video for tengu format

[@redgadgetyugioh]

Thanks a lot and will do! Side cards is heavily requested, so it seems like it might be next😉.

[@Railmydad]

EEEEEYYYYYYY very cool stuff, love me some data

[@redgadgetyugioh]

Me too haha, I love it a lot. Thanks for the love!!!

[@xTxCxMx]

Love this sorta stuff, and I'd love that matchup chart! I would like to ask, about how big the "other" category was, if there even was one? Maybe you said and I just missed it, but I'm assuming from the spread of winrates that there had to be more than a few people just using random piles of cards they liked. One thing I do think would be really cool to see is outliers. Basically cards/packages that weren't normally played in a specific deck, but decks that did include them performed better than the average deck in the archetype.

[@redgadgetyugioh]

Excellent suggestion. I have not heard this one yet. This seems very interesting to look into. Btw, yes there was an "other" category full of pile decks and the wr was low, thus the high winrates of the mentioned decks. Thanks for the feedback. Stay tuned for more 😉

[@Soulcybering]

Love content like this which is real data based

[@redgadgetyugioh]

Thank you so much for those words. I am glad you like it 😁

[@dvsavocs5290]

nerdy questions: how did you gather all the data? is it a huge json file or tabular? are you planning to release this dataset? i think if this video goes viral and changes the popularity of each decks the validity of this analysis will fall off unfortunately. a deck that was niche but has a good winrate may drop due to newer players trying it, also people may start siding more heavily towards the best deck, which is why the meta is ever evolving. how do you come up with certain solutions to your problems? such as using MLE to estimate which deck is being played. I'm a data science student but I feel like I'm not able to use what I learned theoretically and through exercises in real case scenarios amazing video!

[@redgadgetyugioh]

No worries, keep the nerdy questions coming, I love it! Well actually idk what viral even means in the Edison space, you could fit us all on a single bus 😂. Jokes aside, I think people will still play what they want and the numbers will not change much. As you suggest correctly, the meta changes though so the analysis is really only valid for a few months as you suggest. We can't fight that, only have to cease the moment! Applying the knowledge and coming up with techniques is difficult actually. The problem is also when you implement something you are not sure if it is correct and if there might exist a better technique. But I will just list things I used: Intuition, asking friends, asking teaching assistants, asking people who did something similar (owner of format library), asking on yugioh discord (because of domain knowledge), asking on my university discord (for mathematical help), a lot googling... At some point you get some idea what to do. In fact you get 3 or 4 ideas, and you just kind of do all and take what seems right. There is no really right answer, I just think one should stay humble and know that the knowledge of the community is higher than one individual and exploit that. That is the reason why I ask for feedback now as well 😂. I would never know what you think if I didn't ask. Hope it helps and thanks for the kind words! Cheers!

[@ooAku8]

Love the content, very nice. More of everything please. I would love to have more context to these numbers from 1. A statistical standpoint and 2. From a practical standpoint. Tho mostly 1. Considering the size of the dataset an analysis, like how do Winrates change over DB Ratings, do the decks overall frequency changes when the ratings go up, etc. would all be nice. For example: A deck might have a higher WR by being better in Low Rated play. If you add in more representation in Low Rated play it can start to skew numbers to make comparisions hard. And by adding in standard deviation and such we can easier compare the numbers too. Since some WRs might have higher or lower deviation due to match-count and other factors. Adding such an additional line would make the results from the other projects mentioned more interesting to look at as well.

[@dvsavocs5290]

yeah I agree, especially on how the meta decks change winrate over changes of elo. similarly to how in league of legends yasuo or lee sin have a better winrate in high elo than in bronze

[@redgadgetyugioh]

Thanks for the insights, @ooAku8 and @dvsavocs5290! I love the idea of analyzing win rates over different Elo ratings and this is something I considered, but unfortunately, I don't have the ELO data to do that right now. @ooAku8, adding standard deviation is a great suggestion and could really help clarify the results. It’s definitely something I’ll consider for future updates, but I did not want to make the video too mathy and complicated. It is probably only useful for more advanced viewers like yourself (maybe it could be posted in the description 🤷‍♂). Appreciate your thoughtful input-it’s really helpful!

[@spiritkid9109]

I wanna see loads of analytics - e.g. which traps have the highest winrates when played and so on

[@redgadgetyugioh]

Thanks for the support! Stay tuned because many good things are coming. This is a good suggestion, I will add it to my list of ideas. 🚀❤️

[@J0zeRgu3z]

Brother, you just got a subscriber, sounds to me like you are a great programmer! Accounts with 100k+ followers have never laid out something like this! They got nothing on you! Suggestion, apply this same model on other formats, no only it might be interesting for you it might also reel in more people that can appreciate this !❤

[@redgadgetyugioh]

Heyoo 👋, Thank youbso much for the kind words and the subscribe. I appreciate it. The suggestion makes sense and I will look into it. There is many cool things coming, so stay in tune 😎 😉 Cheers!

[@michelthetrout792]

Great video ! Would be even better if there was the winrate for low/mid/high ladder, anyway keep going (maybe another tierlist about another format)

[@redgadgetyugioh]

Yeah definitely, the two things were already suggested and I will look into it. Stay tuned 😉

[@T-Dawggg8735]

Been playing Blackwings for a while, and plan to take them to my first ever YCS in Indy. I have the most fun when I play with Blackwings. They play around a lot and really only hard lose to Gorz or Colossal Fighter if you don’t have an out to them. Running three Vayu is basically a small Vayu Turbo package anyways and gives an easy choice for an Allure target. Book of moon has continued to become underrated and is an answer to a lot of decks bests monsters. I hate milling Brain Con, Mirror Force, Heavy Storm and all the other strong spells and traps that help secure wins in Edison in Vayu Turbo. People miss on mills a lot more IRL than on DB because it’s a lot more random chance than the DB algorithm. It’ll be interesting to see how many people play Vayu Turbo at Indy now that Sack Phase has won back to back major Edison events in RBET and YCS Raleigh

[@redgadgetyugioh]

I am also on Blackwing myself. It is a very good deck and I would say also fun. I am interested as well on what happens with vayu in the future. Thanks for the feedback :).

[@fidgety9001]

I'm not too surprised by the machine decks doing so well, its main issue isn't the deck itself, never was, its the fact the deck has some pretty nasty blowouts with fortress dragon and if people start committing more side deck space System Down

[@redgadgetyugioh]

Most lists play imperial iron wall in the side because it counters bth, dp and system down as well. Yeah to cydra you lose, but people are not playing it atm so 🤷‍♂️. I agree, the side cards are what hurts this deck mostly.

[@saqlaq96]

Love the idea of the mathematical approach to deck building please do it

[@redgadgetyugioh]

Will consider it for a next video for sure. Thanks for letting me know 😁

[@OutbackYGO]

based, ty for the explanation of what you did at the start!

[@redgadgetyugioh]

Perfect, I am glad you enjoyed it 🤟

[@DoubleBeast]

Gladiator Beast players rise up! 🦎🔱 Been a while since I played, but I had always enjoyed GB even though everybody shits on it nowadays. It would be good to know the total # of Duels per Deck and chart the usage % on a timeline to identify trends & changes in the metagame. Personally, I am curious if GB's win rate has dipped as Edison has grown. I think it was "Crow's Nest" who did some analyses of recent performances, and I recall GB having a very poor win rate lol. Like 48% or something.

[@redgadgetyugioh]

Lets go Gladsss! Very good suggestions actually. I didn't want to make the video too technical so I decided to not show this data, but it makes sense. I think GBs are good on average, but might struggle in bigger tournaments honestly. Hope it helps and stay tuned 😁

[@shadowdraqon2479]

Against spell/trap floodgates like royal oppression and necrovally, i would prefer to have raigeki break to pop them, i dont want to phoenix wing blast them back to the top deck so the opponent can draw them again and i have to deal with it again(same problem that transmigration prophecy has, prolongs the problem instead of exterminating it)

[@shadowdraqon2479]

But against any other card, i would actually prefer phoenix wing blast to make the opponent dead draw a card.

[@redgadgetyugioh]

Sometimes getting an extra turn by topdecking their card might win you the game. The new draw is locked 🤷‍♂️

[@robertholtz769]

For the reason that you touched on, I have found putting Decks into Tiers to be dubious and a bit misleading. What precisely makes a deck one or the other. Some decks it is easy, but others there is a fine line. As the strategy is less revolving around specific cards, like Blackwings, and uses a blend of cards, like Hero Frogs or Quickdraw, then decks can be hard to identify. A solution might be to rank individual Cards as well as Packages of cards. This can be supplemental to the Deck tiers where you recognize "these Decks typically contain these Packages."

[@redgadgetyugioh]

Exactly! I agree it is rather difficult, but because people want to see how the whole deck performs and not just an engine I decided to use the technique 3 shown in the video to solve the problem. You comment is still very valid though, so we will see how this develops! Thanks 😁

[@fongh4108]

Very interesting video!

[@redgadgetyugioh]

Thanks! I appreciate it!!!

[@ThatGuanfranco]

I had this video queued up to watch when it was around 300 views

[@redgadgetyugioh]

😱😱 Since day-one 💪

[@Jameson11ink]

This earned a sub, love data

[@redgadgetyugioh]

Thanks a lot. I appreciate it!

[@01zaborg01]

First of all, this is really cool! Was this done with high rated matches only? I wonder how much the results suffer from diluting the player pool.

[@redgadgetyugioh]

Elo was not included in the analysis. Although debatable, I believe the results would be somewhat similar if the Elo was indeed included. Thanks for the kind words!

[@GRP0]

Great work! Here's an idea, player WR and total matches played are pretty good indicators of how good/experienced the player is + many decks are significantly harder to play (so likely get a lower WR due to that) or to play against (so likely higher WR) -> would be interesting to see how the deck WR changes depending on player strength. I'm obviously assuming most players are rather average, i.e. there's a big strength/xp gap between most and the top class.

[@redgadgetyugioh]

This is a very good suggestion. I think it is totally valid and I will definitely look into what I can do. Thanks a lot for the feedback. I appreciate it 😁

[@Samurai31631]

To the OP. You’d be surprised how small the gap really is. Having said that, the small gap is what makes all the difference. I just had a match vs a top 10 ranked player. He made a small read that got him the W. Most would have walked into it and lost blaming it on luck. “I played around ______ just in case. I had a strong feeling….”. All I could do was tip my hat.

[@GRP0]

@redgadgetyugioh Looking forward to the results! Another interesting thing to look at is what happens after siding...

[@TheSagFro]

Could you make a bulkier tier list with more definition between similar decks?

[@redgadgetyugioh]

Good idea! I have added it to my list of ideas. Thanks!

[@TheSagFro]

@@redgadgetyugioh if you want a tier list with a majority of the variants pre-made I think HydroPump and Masha recently used one

[@redgadgetyugioh]

Yes! Thanks 👍😁

[@crudehuman]

Go red gadget go! Thank you for the analysis

[@redgadgetyugioh]

Thanks a lot for the kind words!

[@aureum79]

you think it would be possible to do something similar in unrated? cause old formats without ladders are just played in unlimited unrated, and having built a website all about hat format data i'd love to have db data

[@redgadgetyugioh]

Hi, Well I think doing something similar is definitely possible. I am only aware of Edison and that is my focus now. Congrats on the website 🤟

[@anthonymooney-gm9dm]

I've been trying to learn how to make ai for a long time now just to make something that can deckbuild and play Yu-Gi-Oh. I believe that we have not even began to touch the surface of what it means to play Yu-Gi-Oh. Especially in the modern format, but I think Edison is an amazing place to start. Firstly, players are limited by the banlist times. There isn't enough time to explore formats. This makes it so that areas aren't explored at all, and only years later they pop up as a huge issue. Vanities emptiness is a good example, but over the years others have popped up. Next, I think that the average Yu-Gi-Oh player's gameplay is garbage. Ive been playing the game since it came out, and at a competitive level. Even we meta sheep, who live and breath Yu-Gi-Oh, can't talk about a game the same way you see a grandmaster talk about their chess game. I've never seen anyone give clear lines in analysis. Not that they can't, I'm very sure they can. However, they don't, and it's not a culture that does thorough game alanysis. So on the technical gameplay side there is also a ton of exploring to do. My dream is a bot that can play Yu-Gi-Oh better than a human, but it also makes its own deck. I've thought about the approach, but don't have the expertise. You'd need 2 different ai. One that builds the deck, and one that evluates positions. In my head, you'd build the one that plays the game first. Simply look at the gamestate, and assign each action a percentage just like stockfish does. The decks (or deck) would be prebuilt, and it'd be put in a pool of all the other decks of the format. Ideally in a huge round Robbin tournent style training simulation. Even if the bot is ok at the game, the next step would be to allow it to build its own deck before the game. To train this, you'd make it face off against pre-established meta decks millions of times. Every time it makes a deck that performs well, add it to the pool of decks it faces against. This is my dream, and I really appreciate the work you've put in. Thanks for the content!

[@anthonymooney-gm9dm]

Adding onto my own comment, I feel like transformers allowing context lends itself nicely to the combo nature of Yu-Gi-Oh. I think if we sicked a model on the modern game, it'd find interactions that broke the game that we never considered. My prediction is that it's be with a 60 card deck, and ice barriers would be involved somehow. Or naturia... Either way, nothing we'd ever touch with a 10 foot pole because we're too 'cool'. What stopped me was the upsurge of attention mechanisms and the math behind it. I barely started learning the math for backpropagation and took a look at what is needed for the attention mechanisms and it was pretty clear I was in way over my head. You need the additional power, too. The game is very comex, but I think it's definitely possible. As an example, the most popular ai video on pokemon I believe completely switched to transformers. His model was able to reach the top percentage of pokemon players. The technical play of pokemon is discussed and studies more than Yu-Gi-Oh, I believe, and I don't think it'd take much to get it better than a human. At the very least, if you needed to scale it down, you could choose a mirror match (this meta, snake eyes would be perfect) and just have it grind it out. I bet it'd give the best players in the world a run for their money. In a best of 50, my money would be on the machine. I'm very interested in your opinion on all this. Where do you forsee the major hangups being? What do you think in general of the capabilities of an ai player? What about a deck building ai? The playing ai would kinda be the foundation of the deck building ai, but what do you think of the potential for a system like this where it builds its own deck before every match?

[@redgadgetyugioh]

Hi, These are some very valid comments you made. I am impressed by the coherence of your thoughts and the level of game knowledge you show. In fact I agree with the majority of the stuff you said, so there is really no point in me discussing the points mentioned. However, since you asked for my thoughts on the possibility of the AI implementation and considering this is my background, I will give my two cents. Well the biggest issue actually is coding the system that interprets yugioh. So since we want to have a model that can play the game, thus create new strategies, the model has to train on a deterministic environment. That would be very difficult to create because yugioh has many ruling issues, that is why I am skeptical of the possibility to have something similar, and I wonder how they did it for pokemon? We con in theory train the model to win, but essentially by simulating human moves (classic ML models on the data) and not playing against itself (GAN). I think this is a task for the guys that run this ygo omega amd nexus simulators. If they have a system that runs a game in one split second without the need to use the GUI, that way you can simulate many games and start from nonsensical play al the way to genius moves. This is the way chess ai is created, but chess is a much simpler games. This is what I think. Thanks for your input though, I enjoyed reading it 🔥

[@redgadgetyugioh]

So, to maybe drive the point home, I am interested very much in the AI part, but to do that someone needs to actually define the game and all the interactions. This is the boring part and I don't have the knowledge nor want to do that. I think our best bet is to wait for a master duel code leak 😂😂😂

[@goldenarmour7975]

Blackwing being number 1 does NOT surprise me. I always thought it was better than Vayu and easier to play.

[@redgadgetyugioh]

Yeah, the deck is cracked. Vayu is a very strong deck and just played by very good players atm I feel like.

[@FlameThe4]

I wonder where lightsworn would rank 🤔

[@redgadgetyugioh]

There is a lightswarn deck included in the video. It is in the middle category!

[@AhidoMikaro]

Cool video!

[@redgadgetyugioh]

Thanks, I appreciate it!

[@matriXmaX99]

Kudos. How did u obtain the data?

[@redgadgetyugioh]

Spent quite some time downloading logs from duelingbook.com

[@shadowdraqon2479]

I dont like transmigration prophecy becuz it puts the cards back in deck, so the opponent can access them again. I much prefer banishing their cards so they cant access them(leviair the sea dragon xyz doesnt exist in edison)

[@shadowdraqon2479]

Specifically vs frogs, if u transmigration prophecy their treeborn frogs back to their main deck, they can just foolish burial them using swap frog, thats terrible imo

[@shadowdraqon2479]

And ofc dd crow is a DARK(allure and chaos sorc and DAD)/low level(junk synchron)/low attack(debri dragon)winged beast(icarus attack)/monster(pot of avarice) So it can bring more synergy to a deck that transmigration prophecy

[@redgadgetyugioh]

Yeah but there is Burrial of DD and Return from the DD, the shuffle makes these two cards less powerful!

[@pakstrax]

Is there a chance to look at the raw data you collected?

[@kurthummel4781]

awesome vid

[@redgadgetyugioh]

Appreciate it 🫡

[@TantanDuelist]

Nice work

[@redgadgetyugioh]

Thanks broski 🫡

[@Jannatul.Firdaus.255]

thank you

[@redgadgetyugioh]

No, thank you, for watching!

[@Luubuu125-1]

what a great piece of work you won't do that for GOAT too, don't you?

[@redgadgetyugioh]

@@Luubuu125-1 Hi, thanks for the kind words. I am overbooked with edison ideas, but I will let you know why I am not so eager to work on goat. Goat is a "good card pile" format, much more than edison, and therefore it is even more difficult to differentiate decks and can increase the complexity of the project. Also, I am not a Goat player, but I am open to all opportunities 🤙, so why not in the future. 🤟

[@bendfut]

Fantastic video. Is your code open source? I've got some pretty powerful hardware and would love to run some more numbers on this! Happy to send all results to you to share (if you think they are worth sharing!)

[@redgadgetyugioh]

Hi, I decided to keep the code closed source for now. Just seems like a reasonable thing to do... In the future I might release it. Either way, I appreciate the offer and your kind words! 😁

[@bendfut]

@@redgadgetyugioh Fair enough! Hit me up on discord if you ever decide to open source it, even if its with a CC BY-ND license! (same discord name as my name here)

[@bendfut]

@@redgadgetyugioh I got bored and started trying to do some data crunching on DB replays. I had a question - how did you get past the captcha when scraping replays?

[@rego1100]

Machina represent!

[@redgadgetyugioh]

Hell yeah! I love Machina too!

[@rego1100]

@@redgadgetyugioh I just got top 8 with it at YCS Raleigh, so I glad to see some hard data that the deck is really good.

[@redgadgetyugioh]

@@rego1100 The deck is fire and I just saw your list on yt. Seems very solid, no wonder u topped!!! Congrats man!

[@codyschnepf3967]

Would u have the deck list for the machina deck?

[@redgadgetyugioh]

I really like this list: kzbin.info/www/bejne/rJnPoYNni52roas

[@luKaz895]

what do u mean with machina control? full of traps?

[@redgadgetyugioh]

Hamster ryko caius lyla engine with traps. If you google recent lists you will find similar builds.

[@wolf1yug]

Did you check the data for the dark gaia deck that plays prisma and miracle fusions? Dark Gaia is played in a couple different ways and the version that topped recently was playing prisma not the rock stuff.

[@redgadgetyugioh]

I have just verified it now, the deck considered in the analysis was the rock version. No info on the prisma version.

[@wolf1yug]

@@redgadgetyugioh I think it’s worth checking since that’s the version that did well lately

[@redgadgetyugioh]

@@wolf1yug Agreed! In fact I like that version a lot personally!

[@shadowdraqon2479]

Imo, the true value of playing dragon turbo is, some of the opponents cards become useless/dead. Their heavy storm hits nothing and MST will only be useful if they draw it at the same turn future fusion is played, if its drawn after FF is already used, the damage has already been done(foolish burial times 5)

[@shadowdraqon2479]

Koaki meiru drago makes the opponents light/dark special summons useless like hamster(ss ryko) and all of the boss monsters, gorz/trag/ redmd/ JD/ kristya /chaos sorc/dark simorgh etc

[@shadowdraqon2479]

Prime material dragon(or stardust if they can make it) makes the opponents destruction cards useless, bottomless/torrential/mirror force, ryko/gemini spark/hero blast/absolute zero/icarus attack/black rose dragon/snow man eater/fossil dyna/etc

[@redgadgetyugioh]

Yeah I agree, the deck can be very powerful. I play it also sometimes and it is fun. Unfortunate that often it loses to bad draws...

[@AB-ds3zl]

VERY NICE VIDEO!! CAN YOU DO IT FOR GOAT FORMAT?

[@redgadgetyugioh]

@@AB-ds3zl Thanks 😁👍. Goat is not my focus unfortunately, but I am open to it.

[@AB-ds3zl]

@@redgadgetyugioh maybe we could donate some money to u as goat comunity, i think we would rly appreciate if you could do this exact same analysis for GOAT :D maybe you d be interested

[@redgadgetyugioh]

@@AB-ds3zl Thanks for the offer, but I could not accept that as I wish to be independent in choosing what to work on. Another issue is the dataset. I do not have the data for Goat. If you know a database of Goat replays do share it :).

[@AB-ds3zl]

@@redgadgetyugioh How did you obtain the dataset for EDISON? maybe we could get you the necessary dataset

[@vokyex]

can u make the analysis open source?

[@RobotBowtie]

How did you collect the data?

[@redgadgetyugioh]

I wrote a program to download and parse json logs that get returned by duelingbook.

[@SteveSorex]

No Synchro Cat? :(

[@redgadgetyugioh]

Unfortunately not my friend. Forgot about it 😂☠️

[@Apropos-YGO]

This is the best edison video I have seen in quite some time I couldn't help but notice Diva Hero was not represented here. Was there not enough data on it? Were the hands difficult to differentiate from other decks?

[@redgadgetyugioh]

Thanks a lot on your feedback! I appreciate it. The diva-hero was an oversight on my end. Completely forgot about that deck!

[@Frogslicer]

CHAIN BURN!!!!!!!!!!!

[@redgadgetyugioh]

😈😈😈

[@MH_Zard]

No Flamvell? oof.

[@redgadgetyugioh]

I know right... 🤭😉

[@DuelChannel000]

I didn’t like the video. Too many decks are missing; the meta changes accordingly to the most played decks so those statistics will be outdated as soon as the rotation changes; displaying winrates over an indefinite number of games per deck is imprecise; matchup chart is missing; I think the main objective of the tier list should be showing the best deck in terms of matchups, over a defined number of games against each other deck. I felt like it’s not a tier list, you just made a winrate leaderboard. That said, good job and nice initiative.

[@redgadgetyugioh]

Hi, Thanks for your feedback, and I am sorry you did not like the video. Let me address some of your concerns, so maybe that clarifies some things. 1. Many top decks are missing: These were the decks I decided to include as I did not want to make a very long video. Some other decks could definitely be added. 2. The meta changed: 100% correct and this video will be outdated soon no doubt about it, but one can not influence that and can only cease the moment. 3. Displaying winrates over an indefinite number of games in imprecise. I am not sure what this means as more games you have the more precise your estimate is. The decks with low game count were excluded by default and are not seen here. 4. Matchup chart: Was not really a part of this tierlist unfortunately, although it could be considered for a different video. 5. I think a leaderboard which ranks decks based on power is essentially a tierlist in my book, so maybe we disagree on the definition of a tierlist. To conclude, a lot of things are relative and I hope this explanation helped. Cheers!

[@IcarusSenpai]

noob called dragon turbo bad 🤡

[@redgadgetyugioh]

Noob doesn't understand unfiltered unopinionated simple data analysis 🤡

[@IcarusSenpai]

@@redgadgetyugioh dont talk trash and play gadgets you noob