Handtracking and Set Prediction for Story Deck runs
Przewodniki
/
Handtracking and Set Prediction for Story Deck runs
Zaktualizowany 11 months ago przez AShinyBlackPhoenix

Alright people, now that I am finally done with 5D‘s, I have the time to talk about some of the stuff I noticed in regards to AI behaviour while playing through 5D‘s a lot, which can be pretty important for Story Deck runs.

I am talking about a neat little technique known as handtracking.

Now you might be wondering: Bird, what are you talking about? Handtracking is just keeping track of cards the AI added to their hand and hasn‘t played, yet. Right?

That is a form of handtracking indeed. However, there is a second form of handtracking, which will come in pretty handy. It is a technique, that allows you to use the position a card is played from hand to predict face down cards, that the AI set during their last turn, at least in certain situations.


As a disclaimer: This technique doesn‘t work 100% of the time, but it is consistant enough to be relevant for any duel, that has the AI have mutiple copies of a card they can set face down in their backrow. Sometimes the AI just doesn‘t set all their cards at once for example. Generally speaking there are some duels, where predicting set cards is way more consistant than in others.

This is also something I would consider an advanced strat, so if you are just starting out learning the game, I recommend focusing on a couple of other things first like AI manipulations, getting familiar with the deck lists and how to win the duels fast in general.

Anyway, for this form of handtracking you need to have 3 types of information, which are the following:

  1. The deck lists:

The deck lists are important information, since this technique only works on cards the AI has multiple copies of, so knowing in which duels that is the case is really important.

  1. The hand doesn‘t get shuffled:

As just stated, the hands don‘t get shuffled unless a card got added to the hand by an effect or a card was stacked to the top of the deck. This means, that the card the AI's topdecked will almost always be the left-most card in their hand.

There are some cases, where that still doesn‘t matter. For example if the AI topdecks a Terraforming, they will add a field spell, which shuffles the hand, but anything that isn‘t that exact field spell was in their hand before.

Another case would be, if you stack a face-up monster to the opponents deck. In this case you know, every card, that isn‘t that exact monster was in their hand already.

  1. There are 2 types of set cards:

So you might be wondering what I mean by this? It‘s very simple: there are cards the AI can set multiple copies of and there are cards the AI can only set 1 copy at a time and keeps their duplicate copies in hand.

As a general rule you can say any Continuous Card is a card the AI can set multiple copies with the same name at the same time, any other type of card the AI can only set 1 copy at a time.

There are some exceptions to that rule: in 5D‘s I found "Mektimed Blast", "Golden Apple" and also sadly "Scrap Iron Scarecrow". These are cards the AI can set multiples of, regardless of them being Normal Traps (sorry guys, I wish I had good news for 5D‘s reverse, but that run will forever be ass).

There are probably a couple of more exceptions, these are just ones I can name.


So what does knowing all of this help me with?

Knowing these 3 things means, we can use the AI‘s behaviour of setting only 1 copy of anything that isn‘t a Continuous Card to our advantage by predicting their face down cards.

Let‘s take a really good example for this: The Pegasus VS Kaiba duel from the DM campaign.

In this duel you are playing as Pegasus against Kaiba, who has some pretty annoying trap cards to deal with. Kaibas deck contains 3x "Shadow Spell" (Continuous Trap), 3x "Negate Attack" (Counter Trap) and 1x "Crush Card Virus" (Normal Trap). Lets pretend Kaiba started the duel with 2 set card. We also assume that he didn't set a spell, which he sometimes does, if he draws multiple copies of one without a target for it.

This means, that there can be 3 combination of cards set: 2x "Shadowspell", 1x "Crush Card Virus"+1x "Shadow Spell" or 1x "Crush Card Virus"+1x "Negate Attack". With the rules we have established the AI cannot set 2x Negate Attack, because it is a Counter Trap. That also means, that if you see him flip a "Negate Attack", the other set card cannot be another copy of it.

Let‘s say you declare an attack and Kaiba flipped a "Negate Attack". Now on his turn he sets a new card, that is not his left-most card, which means it is a card he was already holding in hand. This new set has a very high chance to be another copy of "Negate Attack", because he was holding the card until a set copy of "Negate Attack" was activated, a card he can only set 1 copy of at a time.

If the card he played was the left-most card instead, there is no way to predict what he just set.


Another thing, that can be done with what has been described, is to detect bluff sets.

Let's assume the AI didn't set anything for a couple of turns, and gets into a losing position. Suddenly it starts setting a card, that it has been holding for a couple of turns at this point, which you can tell by the position it has been played from.

This is pretty much a guaranteed call on a bluff in this situation, because the AI isn't holding potential backrow, that it can respond with for that long. If it is past the second turn, the AI will usually have set all their backrow at this point, unless it is a card, which it will only set one of at time as I described above or it is a card they try to bluff with, which the AI usually does, if it gets into a losing position.

Same rule also applies here: If the card it played was the left-most card instead, there is no way to predict what it just set.


But, predicting face-down cards can also be applied to cards the AI can set multiple copies of, just not through the means of Handtracking. This one is a fairly simple right to left rule (left to right from the AI's perspective). This can be summarized very simple: If you see the AI flip a card, that it can set multiple copies of, all the cards to the right of the activated card cannot be another copy of that card. The AI will activate the cards from right to left, if it has multiple copies of the same card set and is trying to activate a copy of it.

Let's go back to our example from earlier with the Pegasus VS Kaiba Duel. If he set 2 cards the positions of those card will be the middle column and the second from the left column. Let's pretend Kaiba did have a set "Shadowspell". Based on the position you see the card get activated, you can get a potential read on the other set card.

If it is activated in the middle column, the other set card can still be another "Shadowspell". It can, however, still be any of the other backrow cards as well. But, if you see them activate it in the second from the left column, the card in the middle column cannot be another copy of "Shadowspell" due to the card being to the right of it.


This pretty much sums up how that method of handtracking I discovered during my 5D‘s journey works. It is a really helpful thing to consider for anybody, who is familiar with the deck lists of the campaign they are running.

I hope you find this information useful. I will leave a couple of clips here of me performing handtracking for anyone, who is curious seeing it in action. I was also able to pull it off in other peoples streams, but don‘t have clips from it, so the clips are only from my attempts in this case. Otherwise, all I can say is try it out for yourself.

Regardless, have a nice day.

Statystyki gry
Obserwujący
112
Przebiegi
260
Gracze
51
Najnowsze wątki
Opublikowano 3 years ago
games:thread_reply_count
Opublikowano 5 years ago
games:thread_reply_count
Opublikowano 4 months ago
games:thread_reply_count
Opublikowano 3 years ago
games:thread_reply_count
Opublikowano 4 years ago
games:thread_reply_count