PDA

View Full Version : Troop Power by the Numbers



Mokog
09-09-2013, 07:27 PM
Howdy Hexers!

Today I took a look at some data I have been pouring over. Watch and explore card power by the numbers and see what we can learn from the data set.

~Mokog


http://www.youtube.com/watch?v=dp8jzSGLkkk

ramseytheory
09-10-2013, 10:02 AM
I'm not convinced by your methodology here. In general I think trying to assign a numerical power to a card is an incredibly risky game, since if you understand how to quantify how much advantage a troop being a cantrip (say) gives you over a troop having one extra attack, then you've already understood the set. There are also issues with cards which are only powerful in conjunction with other cards, like Replicator's Gambit, and with cards whose power depends entirely on whether a given strategy is viable, like mill or lifegain cards.

When it comes to your specific method, just on the first two pages I can see you're giving Adaptable Infusion Device 4.5 power, Goblin Cooking Pot 3 power, Eye of Creation 0.42 power, and Inner Conflict 0.38 power. That's not a good sign. You also seem to be including tokens in your data set and (going by the last page) taking rarity into account in assessing a cards power. (You're also doing things like subtracting a constant amount for an ability being costed rather than an amount depending on the cost of the ability.)

The other issue is that you're looking at the average power of a card at a given cost, but I'm not sure that's relevant to constructed. Good constructed decks don't play average cards - they play the best cards available. For example, no-one's ever going to play a Mushwocky outside of some weird combo deck when for the same cost they could play a Xarlox or a Vampire Lord. The important question is therefore how the best cards at a given cost compare to each other. That's a question which is much more sensitive to errors in power measurement, so it's probably best answered by comparing cards subjectively. (The easiest way would be to wait until beta hits, then poll a randomly selected group of 50 or so competitive players.)

Sorry if this post came off as abrasive - I do think it's very important to try and get a sense for how much power a high cost card buys you. I just don't think this is the best way of doing it...

One question I do think could be answered with this sort of approach is how much attack or defense we should expect a given cost troop to have, in the absence of other abilities. Ideas like an x/x troop for x being "on curve" work in Magic, but we don't actually know to what extent they carry over to Hex.

Mokog
09-10-2013, 06:30 PM
Thanks for actually looking into the data Ramsey! The file is still incomplete and I wanted to share some of what I was seeing emerge in the date. As for the methodology I will explain it. The vanilla curve is assumed to be 1 resource for a 1/1 creature continuing with single integer increases. Since we have multiple card types what is a 1 cost spell equal? Or a 1 cost constant? There is huge variability. Is an artifact the same as a constant? The system is deep but not impossible.
To begin to dig into the data you have to set up a scaffolding and attempt impartiality to quantify mechanics and then test them against hypothetical and real world scenarios. My current general theory (still in its infancy mind you) is card power = permanent power + cost efficiency + Mechanics power +versatility. The components are still being worked on. As a general theory it is not horrible but needs additional rigor and definition. The data used in the video makes claims about troops because permanent power defines lines the best so far. It is not complete and I disclose at the beginning of the video I still have mistakes I am working on. Good theory is not built in a day.
I do not call this risky but I do call it complex and fascinating. The only risk here is to my emotions and ego. So to mitigate that risk I try to be open, honest and share what I find. To release any of this I have to admit I may be flawed and the process backwards, but when I find a gem in my digging around and data manipulation I want to share it and see if it can help players learn.
So I laid out the raw data in a scatter plot to see what it could tell me about how power changed with the current definitions as the cost increased. There are no averages in those graphs. The lines are to point out visible trends. As for constructed format each player will build a deck to maximize the board power for their given style. This is one reason I canít wait for alpha! First I will have access to the full pvp dataset and with that full access to test out concepts from the data. My comments on constructed are based on the seemingly large divergence from the assumed curve at the different resource costs.
As for your examples, good spotting! That is the hard part about cross comparing card types. In the data set a card like eye of creation has a very nifty power effect that requires additional resources to perform an amazing effect. If you only pay 3 for your eye of creation it is going to give you no effect. It has no real power in that case. Inner conflict pacifies an opponent for two cost but if played on a 1/1 has minimal effect. If you play it turn 4 on a 3/3 crush troop its fantastic. Adaptable infusion device is an artifact permanent that fixes threshold and provides a card draw mechanic. Goblin cooking pot is also an artifact permanent, with a basic ability that deals 1 damage, buries a card, and drops an enemy charge for 1 resource. 2 net resources to act in three zones is nice when you look at it from a different angle. Capturing these aspects is the challenge of the project. So far I have a better grasp on permanent cards like troops and artifacts and soon I will have actions better defined.
If anyone who made it to the bottom of this huge block of text is interested I am always open for some collaboration on the subject.
Thank you for watching Ramsey! Holding people to account and asking questions helps keep us honest. I am sure I will have other controversial conclusions in the future. Keep watching so I can have someone looking over my shoulder keeping me focused.

Norious
09-11-2013, 06:33 AM
I thought I remember someone putting a video together about card value. I would like to this and that video and get great minds of hex together. Collectors like me wanna know!

Mokog
09-11-2013, 07:26 PM
That may have been me Norious :-) Ideally one would be able to take the computed power value of a card and apply a couple modifiers to get its chase or market value for PvP. Btw, I also get to go through this with all the PvE cards and factor in equipment at some point. I am a special crazy so you won't have to be :-)

shocker455
09-12-2013, 01:32 AM
Well there are a few things i disagree with on your methods.

Just glancing over the your power numbers i disagree with a lot of them. Now i know it is a challenging task to give each card a correct power number, that said a person who tries would need to go back and reassess the power values of each card probably 20+ times. Also listening to other skilled view points on why X card should be valued at Y power value. Until the numbers are more precise, any information gained from the data is flawed.

Next cards can't have only one power value. They need a value for sealed, and one for constructed. This is because in a constructed deck, cards can have reliable synergy. For constructed, card that relay on synergy to be effective would have a higher power level. When it comes to sealed the odds of you also drafting the perfect card, then also drawing both cards is low so the cards power level would be much lower.

Why are you comparing power to cost as a ratio? You Hypothetical version would say a 5 cost card, should have a ratio of 5.0 to power. Thus 5 cost cards would have a power level of 25, which would be worth as much as playing 6.25 2 cost cards. I think we easily agree that a 5 cost card is not worth 6.25 2 cost cards. Graphing card cost directly/linearly to card power seems like a better way to me.

Also the cards mana cost to play the card isn't its only cost. It also costs the player a card. In a card game, cards are an extremely valuable resource. This is one of the reasons low cost cards tend to have higher power levels then normally expected.

If i were to predict how the graph would look, id expect an exponential growth graph. Slowly growing from 1-3 cost, med speed 4-6, then fast increase from 7+. If you were to graph the ratio of card cost to power it would probably look like an upside down bell graph, with 4-6 being the middle.

HyenaNipples
09-12-2013, 01:45 AM
This all sounds like an attempt at creating quantitative values for a qualitative subject. Not only is human opinion an inherent part of evaluating cards, but the constantly changing metagame means the entire tapestry of card power is one, giant variable.

Mokog
09-12-2013, 03:42 AM
Yes and no. I agree that in the ideal meta game each card should be qualitative, part of a specific place in the game but ever shifting around in power. What I have seen in general is that there are just better cards than others meaning in some cases for a deck's purpose there is just a better card. That indicates a quantitative measure exists. One of my hunches is that deck archetypes and shard differentiation will end up being qualitative but card choice by deck type is quantitative. I suspect that the game follows many of the concepts in this video (https://www.youtube.com/watch?v=e31OSVZF77w). I want to dig down to that curve so we can have good early structure for looking at PvP and PvE.

HyenaNipples
09-12-2013, 04:15 AM
Hmm, I'll be interested to see the results when you are able to compare your projections with documented PvP data. That will be cool.

Mokog
09-12-2013, 06:12 PM
Have we found someone who has dedicated themselves to recording raw pvp data yet? Maybe someone setting up to do tournament reports and the like? Always looking for good data sources.

Leingod
09-12-2013, 09:27 PM
I think people are waiting to see if CZE has built in tools for that sort of tracking or plans for it. Don't think there is a dedicated league or tourney site yet though.