Here are the recommended PW to play with Arena Rector. This is just my opinion. Fill free to discuss any changes.
Below is my list of priority PW to use (from top to bottom). Of course it always depends on your metagame. A field with a predominant Eldrazi might favour more Elspeth, sun champion for example.
1. Nissa / Angrath
2. Vraska /God Bolas / Ugin
3. Karn / Grim Sorin
4. Ajani / Elspeth / Flip Garruk / Predator Garruk
5. Liliana / Ob Nix /Gideon of the trials
6. Bolas PW / Sorin Lod
I hope this will explain how I came with those numbers and help you choose the best PW for your NicFit deack with Arena Rector.
Draw:
0 – no card draw
1 – draw with drawback
2 – straight draw a card
Discard:
0 – no discard
1 – discard with drawback
2 – straight discard a card
CREAT CREATURES:
0 – does not create a creature
1 – create some sort of board presence
2 – create a creature
Kill creatures:
0 – does not kill creatures
1 – kill creatures with drawback
2 – kill creatures
Kill Jace:
0 – does not kill Jace
1 – kill Jace indirectly (create board presence, deal DD, become a creature)
2 – Nuke the blue mage
Kill TNN
0 – does not kill TNN
1 – can kill it with proper setup
2 – Nuke this bastard
Cost:
0 – expensive CMC>=6 (require Arena Rector)
1 – medium CMC = 4 or 5
2 – Cheap CMC =<3
Color WBG:
0 – require at least two color splash
1 – require one color splash
2 – in color
Other:
0 – no additional bonus
1 – have some sort of interaction
2 – win more interaction
Ult:
0 – useless ult
1 – game changing ult
2 – game ending ult
Regardless of how I feel about the list your scoring is off here and there. How can Liliana and Ugin score a 1 on colour using the criteria you define? Ugin only requires generic mana to cast so it's in no way relevant what colours of mana you're able to produce and Liliana is in the same colours as Sorin, Vraska and Big Garruk, which you score at 2. Also, you fully omit the unfair MUs in your scores which I think is very unwise. They're typically AnyFit's biggest weakness and as such should absolutely be addressed/acknowledged. And then there's the lack of weight for various metrics. Not all are equally important. It's also kind of odd to weigh "kills a single creature" as equal to "wipes the entire friggin' board, planeswalkers included". One is vastly more impactful and likely to turn the game around in your favor than the other yet under the current criteria they score equally.
On the list itself - you're failing to accomodate for cases covered. You can run all the high priority targets you want but if they all just cover the same bases you're worse off than running a combination of possibly lower priority targets that do round all the bases.
I understand what you're trying to do and it's something I really do applaud but it's still a ways off from actually being as helpful as you intend it to be.
Originally Posted by cavalrywolfpack
This kind of weighted matrix for deckbuilding will always be shit
1. Inevitable oversimplification of factors leads to obvious nonsense like 'Angrath is the best planeswalker'
2. People will then try using some kind of qualitative argument to re-assign the weights such that the hierarchy of card choices actually makes sense (or at least agrees with ideas that they already hold)
3. If people are agreeable to step 2 it means that they can be persuaded by a qualitative argument
4. If people are able to be persuaded by qualitative arguments then the grid of numbers is pointless
If you do want to take some sort of spreadsheet approach you're probably best off making a grid where you tick off which PW "beats" which MU. That way you can get to a configuration that ticks off all boxes.
Making those spreadsheet is fun, time consuming and always a great puzzle. The biggest difficulties are properly defining your criteria.
Like Kombatkiwi said in his post.
At the beginning I always start with something big. My first grid had around 25 different options to prove a PW is useful.
But this gave strange results and did not provide a straight answer, so I binned it.
I have decided to define criteria the way I struggle in the current metagame. From my long experience in card games I have selected these categories. Maybe I should add this to my post to clarify more and avoid confusion.
1. Discard – mostly helps against combo and control.
2. Draw – card advantage is always good
3. Creating creatures – helps against PW, control. Provide some board presence after Deed/Deluge reset or top deck fight.
4. Kill creatures – I think this is obvious. I mostly have problems with Zombie fish (poor rhino) and at the moment delver/strix
5. Kill Jace – so I killed a lot of PW over the past years. Jace with 5 loyalty on an empty board was the most difficult piece of crap to ever remove if you do not draw your Vindicate/pulse.
6. Kill TNN – I think this is also obvious, we don’t run counterspells so Deeds + Deluge main should make it easier. Some extra stuff will be great. I hate this card.
7. Cost – we are able to play high CMC but not to high I think 6 is really the max. Big boys like Tyrants/Dromoka are amazing but expensive. Same situation for PW.
8. Color – Here the biggest difficulties are splashes and Double W. Ugin has a typo probably.
9. Other – this is what you are looking for Echelon. We cannot describe every PW. I thought even to remove this category. But how describe the usefulness of Nissa –1, Ajani lock/life gain, etc. This is simply impossible.
10. Ult - this is obvious you want to win on the spot
These are my criteria, the one I struggle the most. Of course I can easily adjust my deck to kill Jace and Garruk Predator will be less useful. You just play 2 Thragtusk, Titania and a couple of pulses.
You want to kill TNN play 3 deeds, 2 Deluge and plague main.
I know the list is not perfect, but this list will give you an overview about different possibilities and how you can play your new teacher in NicFit.
You can always argue. Ugin is great against belcher because you can wipe the board after he makes a tons of goblins. Vraska is amazing against storm because you can kill their LED. Ajani can lock there black source, etc. Every PW is useful depends on the situation. The more criteria you add the more vague the answer you will have.
Coming back to “nonsense” like Angrath. Yes he is amazing. Did you try him?
This is one card for 5CMC provide edge against combo + clock, steal Zombie fish+swing+sac to Tower/therapy/intent, try him with Meren, his -3 will deal at least 1DMG (Steal DRS/strix) and open the doors. He is an amazing grinder and I strongly recommend trying him. It is sweet when your opponent has to block his own Leo with TNN to not die. Tears and Pain this is what NicFit does the best.
Yes, Bolas is better in a vaccum against combo and you can fetch him in turn 4 /5 if you are lucky. That is great, you will probably win. Ugin owns the entire board, etc. but those big boys are difficult to put in to play before turn 5.
We have to remember this is not a combo deck, PW are here to turn the table and put the last nail in the coffin before you die to TNN, DRS and other crap and for combo match up I already play additional discard in form of Brutality/though seize in the main.
Turn 6 Vraska into turn 7 Bolas is amazing btw.
I can easily produce a grid with all the possible matchup available in MTGtop8, but really it will not say me much more. If I evaluate that Liliana is amazing against BUG TNN how she will help me in against a board of TNN+DRS. When at this moment Ajani -2 will be amazing to remove DRS and give me an extra turn to live, probably two. I am open to discuss better criteria. I can add more PW if need looks like I forgot Kaya.
I tried to binary describe the usefulness, but it did not work. So I am stuck with:
1. useless,
2. might help,
3. extremely useful.
Splitting into percentage is extremely difficult. How useful is Liliana +1 to Angrath +1 to Kaya -2. This strongly depends on the matchup.
Sorry for the wall of text, I hope this will help
This is exactly what I mean.Coming back to “nonsense” like Angrath. Yes he is amazing. Did you try him?
This is one card for 5CMC provide edge against combo + clock, steal Zombie fish+swing+sac to Tower/therapy/intent, try him with Meren, his -3 will deal at least 1DMG (Steal DRS/strix) and open the doors. He is an amazing grinder and I strongly recommend trying him. It is sweet when your opponent has to block his own Leo with TNN to not die. Tears and Pain this is what NicFit does the best.
Yes, Bolas is better in a vaccum against combo and you can fetch him in turn 4 /5 if you are lucky. That is great, you will probably win. Ugin owns the entire board, etc. but those big boys are difficult to put in to play before turn 5.
If you want to convince people that Angrath is good, just make your points like this. Nobody cares about a table of numbers because they don't know your thought process that went into it.
Once you have explained the strengths/weaknesses of the card and we understand situations or matchups when it's good or bad, then defining its 'creature kill' power as 0 or 1 or 2 doesn't provide any additional information.
I'm not saying it's a 'big difficulty' to properly define the criteria, just that it's a total waste of time.
This
I disagree. I think there's a proper way to do it but the algorithm is a bit more complex (and the data entry is a bitch). What you have to do is make a series of evaluation criteria for card vs card. Don't rate a cards abilities on a point by point basis. Insteads of draw, discard, kill etc for Elspeth you measure Elspeth vs Tombstalker, Elspeth vs Young Pyromancer, and so on. Do this for the important cards in the meta (mtgtop8 provides the most popular cards), and then weight those cards by how common they are. Then further weight by the next metagame or the current winners metagame if you would prefer.
As a system, that works quite well however it scales horribly and seems to be impossible for any one person to implement... at least until "someone" finishes writing a program to parse oracle text and build the table automatically.
I've done some work on that front some years ago, but it's still far from advanced enough to be able to figure out if Elspeth/Ugin is a proper answer to Ugin or not. I'm at the point where I can say with some certainty that they can count as removal (in the most generic sense) and that's about it. This is accomplished by an overlong, ugly ass if-statement with a loooooot of ORs that check if the oracle text contains a certain string. There has to be a more elegant way to do this.
1st Problem - You have to account for way too many things:
- 'Raw' power of different cards (whatever this means)
- How a card fits with the mana curve of every other card in the deck, and consider the % of the metagame playing wasteland, stifle etc
- How demanding the colors of different cards are, and considering the % of the metagame playing blood moons etc
- Power and toughness of different creatures you can play vs other commonly played meta creatures
- Some way to quantify how good unique sb cards like Lost Legacy are (or aren't) vs various strategies
- Have some penalty for 1 drops because of chalice decks
- etc
I'm sure you can come up with a million other variables, it's so difficult to claim that your system successfully captures every single possible attribute of different cards and their interactions between each other
2nd Problem - Assuming that you have correctly identified every possible factor involved in card selection, the weighting for each has to be 100% data driven.
For example, you can't just say "1 mana cards get a -10%*D multiplier (where D is the meta% of chalice decks) because I think that this reflects the drawback that these cards represent vs Chalice of the Void". How do you know it's 10%? Maybe Chalice actually isn't that important and it's only -2%. Maybe this is the wrong approach and rather than give individual cards a penalty vs Chalice you need to give the entire deck a 'Chalice penalty' that is a function of the number of 1-mana cards squared. The point is that whatever rule you choose has to come from empirical evidence of how 1 mana cards actually affect your win percentage vs chalice decks. If your justification is that "10% feels about right" then the numbers are just a proxy for your intuition, at which point you should abandon the numbers and admit you are just building your deck based on how good you think certain cards are.
3rd Problem - You can't disagree with the results
Say you think you have accounted for every element of the strength of different cards and the web of how they all interact with each other and you have empirical data that shows how strongly each different factor affects deck strength and you plug it all into an algorithm which spits out a decklist.
You look at the decklist and it seems shit. But you can't make changes without testing it because that is just an admission that intuition is king and the numbers are pointless.
So you test it in some games and it still seems shit. So you go back and tweak your parameters. But this means that you failed at either problem 1 or problem 2. It is possible that by some iterative testing process you can eventually arrive at some ideal list, but I seriously question the efficiency and effectiveness of doing this by numerically updating rules and weights as opposed to the 'normal' heuristic way.
If you really believe this approach to be viable, why would you limit it to specific archetypes? Why not apply it to the whole of legacy? If you can really build such a system it would necessarily have the capability to solve the entire format. I suspect that to end up with a Nic Fit deck you would have to include axioms that aren't empirically justifiable (like "DECK must contain 4 (FOUR) Veteran Explorer and 4 (FOUR) Cabal Therapy") otherwise your program would probably just give you a list with 4 Delver, 4 Brainstorm, 4 Ponder...
You're missing a few points here and there, but that's ok.
But your problem with the approach is that, if done properly, it's a giant and complex undertaking..? I don't see the problem.
I mean, the fact that people tend to disagree with whatever does not conform to their opinions is nothing new. Especially if it's a piece of software that reaches that conclusion. That doesn't mean the software is wrong though. We're just unwilling to believe/trust it is.
Echelon's 'good in matchup' is a nice semi-qualitative way to make the Matrix. That way most planeswalkers wont overlap.
Kind of similar to the pay-off suite in nyxfit. I have yet to meet a deck where no enchantment are game-changing.
Thank you. It's probably the easiest way to get to a configuration where you can handle all the things you need to be able to handle.
Edit: I've done this analysis, but it does point out that enchantments generally have a much broader application than planeswalkers.
My argument is that unless you do it "properly", the software is just a filter that converts an opinion into a number. Therefore, the only function of the software/matrix is to obfuscate the fact that its recommendations ARE opinions. (By employing some illusion of numerical objectivity). A discussion of how to best implement the matrix is therefore a pointless distraction. We are all literate humans capable of discussing the merits of something in qualitative terms, there is no benefit to having every thought distilled into a numeral:
"Ugin is a 0 in terms of CMC but in terms of power it's a 10, so it scores higher than other PWs and is a must play"
"Actually it's not significantly stronger than Sorin, which is only a 7 in terms of power level. So Ugin is really more like power level 8. However Sorin's CMC score is 2, which means overall Sorin is 9 and Ugin is 8, so Sorin is better"
"Ah, but you're neglecting Ugin's ability to eliminate noncreature permanents and TNN, which is easily worth 3 extra points"
Why do you want to have that kind of discussion? You can remove all references to numbers and the message is exactly the same:
"Ugin has extremely powerful abilities and comes down with a lot of loyalty so I think it's essential even though it's hard to cast"
"Is it really much stronger than other planeswalkers which are much more easily castable? (Sorin, Vraska etc)"
"Yes, no other PW can wipe multiple permanents at once including TNNs"
The counter argument is that I'm strawmanning you and it only needs to be done "properly" so that there is no longer any dependence on opinion, but I think you seriously underestimate the difficulty involved. If it was even _possible_ to achieve, it would require so much playtesting to get the data that most reasonably good players with that data could reach a similar conclusion without needing the magic formula. The matrix scheme is so difficult to do properly and the marginal benefit is so small that suggesting it as a viable way to build a deck is plainly wrong
Yeah, no, you don't have a clue and are missing what the conversation was on in the first place. This conversation is pointless.
On an unrelated topic, I'm taking 5-Siege on Saturday (4x Rhino + Doran), hoping to catch some folks napping. Mana-base = stabilized. Play-testing = 20+ hours. Fun-factor = 101%.
The cards I'm hoping to reach conclusive opinion on: Doran, the Siege Tower, Liliana, the Last Hope, Painful Truths. Everything else is stock from a typical Junk Fit. What I hope to achieve with Doran is to neuter TNN and shave about 1/2 turn off the clock. Liliana is to have another source of recurring card advantage by pinging off small dudes and getting creatures back from the graveyard. Painful Truths is to have a maindeck way to grind back into games against the control decks in the format (Miracles, Czech Pile.) I'll have a full report by Monday, hopefully.
Brainstorm Realist
I close my eyes and sink within myself, relive the gift of precious memories, in need of a fix called innocence. - Chuck Shuldiner
Good luck! Looking forward to the report!
It's actually not that big a list of things to account for. I've done some work on the conditions, if you check my post history you can find them. You don't actually have to weight the power of cards, beat/tie/lose is enough. Raw power naturally appears from that in a card that beats everything.
I think you're misunderstanding what the results of such a system do. A matchup matrix is only the first step. From that, you then use some social network analysis to create groups of cards that naturally work well together in clusters. Once you have that, you can pick and choose your cards based on more manual criteria. All in all it's a system for reducing a card pool to a handful of cards. From there you need to apply some deck building logic (or do what I did and write a program that can build decks for you).- How a card fits with the mana curve of every other card in the deck, and consider the % of the metagame playing wasteland, stifle etc
- How demanding the colors of different cards are, and considering the % of the metagame playing blood moons etc
That one is extremely simple, it's just a simple x vs y calculation. Beat if x wins, tie if they bounce or trade, lose if y wins.- Power and toughness of different creatures you can play vs other commonly played meta creatures
This one is easy.- Some way to quantify how good unique sb cards like Lost Legacy are (or aren't) vs various strategies
No penalty is needed. In the matrix, chalice beats 1 drops. In the weighted results the value of that shifts according to the number of decks with chalice you put in the metagame.- Have some penalty for 1 drops because of chalice decks
This is what you said above. It's a trivial problem to solve.2nd Problem - Assuming that you have correctly identified every possible factor involved in card selection, the weighting for each has to be 100% data driven.
For example, you can't just say "1 mana cards get a -10%*D multiplier (where D is the meta% of chalice decks) because I think that this reflects the drawback that these cards represent vs Chalice of the Void". How do you know it's 10%? Maybe Chalice actually isn't that important and it's only -2%. Maybe this is the wrong approach and rather than give individual cards a penalty vs Chalice you need to give the entire deck a 'Chalice penalty' that is a function of the number of 1-mana cards squared. The point is that whatever rule you choose has to come from empirical evidence of how 1 mana cards actually affect your win percentage vs chalice decks. If your justification is that "10% feels about right" then the numbers are just a proxy for your intuition, at which point you should abandon the numbers and admit you are just building your deck based on how good you think certain cards are.
That's why I built an AI that's capable of taking decklists and playing them against each other. Variance in Magic is too high, a handful of games doesn't give anywhere near reliable information. And yes, I've actually taken decks my program has built for me to tournaments. They did well.3rd Problem - You can't disagree with the results
Say you think you have accounted for every element of the strength of different cards and the web of how they all interact with each other and you have empirical data that shows how strongly each different factor affects deck strength and you plug it all into an algorithm which spits out a decklist.
You look at the decklist and it seems shit. But you can't make changes without testing it because that is just an admission that intuition is king and the numbers are pointless.
So you test it in some games and it still seems shit. So you go back and tweak your parameters. But this means that you failed at either problem 1 or problem 2. It is possible that by some iterative testing process you can eventually arrive at some ideal list, but I seriously question the efficiency and effectiveness of doing this by numerically updating rules and weights as opposed to the 'normal' heuristic way.
There's two problems at work here, one is of building a card pool and the other is tuning a deck. The cardpool problem is the one you were referring to and it will work for any deck to find piles of threats+removal, it's not capable of finding combos. The other problem is one of playing the decks against each other in order to tune them. That problem becomes archetype and in some cases deck specific because you need to approach cards with different views on their use depending on the decks strategy. For example, the criteria for what card you want to name with Cabal Therapy is much different for Nic Fit as you're usually looking to take your opponents enablers than for Storm where you want to take their answers.If you really believe this approach to be viable, why would you limit it to specific archetypes? Why not apply it to the whole of legacy? If you can really build such a system it would necessarily have the capability to solve the entire format. I suspect that to end up with a Nic Fit deck you would have to include axioms that aren't empirically justifiable (like "DECK must contain 4 (FOUR) Veteran Explorer and 4 (FOUR) Cabal Therapy") otherwise your program would probably just give you a list with 4 Delver, 4 Brainstorm, 4 Ponder...
That's not the discussion at all.
@Brael: Don't feed the troll... You know better than that.
There are currently 1 users browsing this thread. (0 members and 1 guests)
Reply With Quote