Cards Don't Matter - Why You're Testing Wrong

This article is also available in video form at https://www.youtube.com/watch?v=HFzO0Dx2xTA

“The cards you have in your deck don’t matter” - that’s something that top players sometimes say. But what does it really mean? When it comes to answering that question, context is important. When you hear a top player say that, it’s usually when they just lost to a deck which they regard as worse than their own. And they might actually be right: they might have indeed just lost to an inferior deck. Their reaction, of course, is borne from the heat of the moment. But there’s something more to unpack here. Consider the notion that the cards you have in your deck do not matter, but that it is instead solely how you pilot your deck that determines the outcome of matches. Of course, on a basic level this is wrong – cards do matter. But the reason why it feels that they sometimes do not is that you’re probably testing your decks in the wrong way. This article will focus on deckbuilding – specifically, it will focus on how you can make sure that your deck is good after you’ve built it. Playtesting is not just important, it is essential – it’s how most top players prepare for big tournaments. Yet even top players – the same ones who may get heated in the face of an unhappy defeat – are usually doing it wrong when it comes to this key part of preparation. In this article, I will try and explain why.

Playtesting is easy. You throw together a bunch of cards, jump into GWENT, and play 200 games with your deck. As you play, you adjust things depending on what you’re playing against, how the deck feels, and what you feel is missing. Then you play 200 more games with it. Then 200 more. And hopefully, at some point, the deck will be refined and you won’t be able to think of anything better. Either that, or you run out of time and a tournament comes around, forcing you to play with an unrefined deck. The best deckbuilders, then, are the ones that can quickly spot what’s wrong with a deck and refine it in a lower number of games – playing out hundreds of matches to test your deck is ideal in a vacuum, but in practice, you need to be more efficient than that to keep up with the twists and turns of the meta.

So, you build a deck, test it, work out the creases over time – sounds easy enough, right? But how do you become a great deckbuilder? How do you manage to figure out what the optimal card for that gold slot is in 20 games instead of 50? You play the game. A lot. For 200 games, maybe more. And as you grow more experienced, you adapt faster.

There are, of course, many issues with this approach to playtesting. The first of them is that it requires you to invest a lot of time into it. Not everyone can play 10 hours a day and work endlessly on refining a deck. Scratch that – almost no one can realistically put in this much time. Even if you could, tournaments often require you to prepare multiple decks in short time frames, and doing this degree of playtesting for 2 or 3 or even 4 decks in this manner is impossible. Maybe you could have a friend try to prepare a Nilfgaard deck while you work on Scoia’tael you’d make it in time – but you don’t really trust anyone as much as you trust yourself, and even the best guide to piloting any deck cannot match the experience of actual playtesting. Deckbuilding with trial and error as the method for refinement dictates that you will end up putting hundreds of hours into the game and playing thousands of games – you’ll gain experience, but it’s in no way efficient.

Even worse, it’s unreliable. If you’re testing your decks by playing a lot of games with them and refining as you go, then you open yourself to wrong conclusions. What if you play against 30 Radovid decks in a row? Surely a lot of your control tools will feel redundant, and you’ll adjust your deck accordingly. But what if Radovid isn’t a good tournament deck? In that case, you’ll have a deck that’s teched against Radovid when you actually aren’t facing Radovid. What if Radovid falls off in popularity in the next few days? If that happens, since you’ll still be playing games, you’ll adjust your deck again and remove the anti-Radovid techs: but then you’re back to square one, and you just keep constantly adjusting your deck to what you’re encountering in your specific MMR ecosystem. There’s no guarantee that you’re actually making it better or approaching an optimal version – you’re merely treading water in a vast sea of players mostly doing the same thing.

One way to struggle against the tides of the meta and the inefficiency of deckbuilding-by-the-grind is to join forces with your fellow players. By playing within a testing group, in a controlled environment, you get a group of very good players working together to test specific matchups and strategies against each other as much as they can. But this brings up a whole host of new problems: if you only play within a controlled environment, you risk falling behind on what’s being played in the meta. Worse still, you risk creating a bubble in which you think something is doing well in a certain matchup, but you’re not playing against up-to-date techs or the latest trends that people on the ladder are favoring – you’re just playing against the version of that deck your friend plays. So, again, you reach an optimal version of your deck that isn’t actually optimal – it’s just good against a specific thing. This particular method of preparation can – and has – led to significant surprises and perceived upsets in tournament scenes, particularly Gwent Challenger, where certain players banded together in small scrim groups to prepare, and many ended up getting knocked out much earlier than they expected in the Swiss marathon of Day 1.

What, then, does it mean to speak of an optimal version of a deck? The answer, unfortunately, comes to us from math – unpleasant for some, but ultimately essential to actual success.

Gwent is, as with all card games, a game of numbers. We don’t often think very hard or very long on the mathematical aspects of the game, because we’re humans and not computers. And yet, the fact is that there is, theoretically, an optimal way to play the game, a combination of cards that absolutely maximizes the value you can get in the greatest number of situations possible. And though even the best of players can be too dumb to see it, the way that the math of Gwent works guarantees that there is a combination of cards that has the maximum probability of winning you games.

“I had three but it tested bad so I play two.”

“It was good in theory but bad in testing so I took it out.”

“It was great in testing so I played it.”

These are all very commonly heard arguments for why you play a certain card, and I’m here to tell you that they’re horrible arguments. But why, you ask? Surely you shouldn’t play things that don’t do well in your testing. And surely if you tried a card and it played great, you should play it in your deck.

The underlying fallacy to this attitude is that results are directly tied to effectiveness. Unfortunately, this isn’t true. While good cards get you better chances of having good results, the fact that you get good results doesn’t guarantee that a card is good. The truth of the matter is this – and only this: “Results don’t make cards good, they don’t make players good, and they don’t make decks good.” If you’re preparing for a tournament, looking at what decks have been successful in the last one should be purely a meta call – decks that did well will be played again. But the fact that the decks did well in no way guarantees that they are good decks. Thinking that a deck doing well is a guarantee of the quality of the deck implies that results are fair – but card games have variance, and the truth is that the best deck in a tournament can lose all the games it plays. Just as well, the absolute worst deck in a tournament can win all the games with enough luck. That’s just the nature of card games. All you can do is try to bring the deck that gives you the highest chance of winning. And yet, even if your deck is flawless, you’re still not guaranteed to win. Maybe you play it poorly. Maybe the player playing the best deck in the last tournament played it poorly, and a person playing a worse deck played it very well. It doesn’t matter. What matters is that results don’t guarantee quality.

This is valid for tournaments, so it’s valid for playtesting as well. The fact that a card does well in your testing does not guarantee that the card is good. I’ll say it again, because this is what really matters: results do NOT guarantee quality.

“He won a tournament with it” is the exact same fallible logic as “I liked it when I tested it so I played it.” Both of these notions ignore the truth that the best combination of cards is indifferent to such results, and this is exactly where your “data” gained from playtesting should not hold weight.

How do you deckbuild, then? When you’re in the deckbuilder staring down at a bunch of cards, how do you choose what makes the cut and what doesn’t?

In deckbuilding, theory and math should always be more important than results. How you determine what’s good in theory will be the subject of another article, but let me give a very practical example of this. Imagine you’re choosing a final gold slot for a Nilfgaard deck in which one of your win conditions is to have the absolute highest amount of gold value possible. Consider your options. Should you play Geralt, Regis: Higher Vampire or Tibor? Geralt is 10 points of gold value with possibly 3 more if he triggers Brave. Regis: Higher Vampire is 6 strength of gold value, but he’ll often reach the 10 that Geralt does and denies some value by stealing a bronze card from your opponent. He is, therefore, worth more than 10. Tibor, on the other hand, gives you 23 points of gold value and gives your opponent a bronze card. Imagine that bronze card is worth, in average, 8 points. In that case, Tibor is 15 points of gold power. The decision between him and Regis: Higher Vampire will be made by looking at the meta and deciding if, on average, Regis: Higher Vampire gets more than 15 points – that is, by seeing if he usually gets 9 or more points off of his effect. Of course, a lot of things factor into this decision – meta calls, the composition of your deck, etc. But the important thing is that it’s theory that determines what cards you choose to put on your deck, and not playtesting. Testing is subject to variance – theory isn’t. If your thinking is solid, theory should always possess more substance than whatever results you may glean from testing. And, lucky for us, theory takes a lot less time than playtesting.

One of my future articles will address specifically how to improve your theoretical thinking and how to do math at an advanced level in card games. But for now, we’ll let the point rest: theory is more important than results when choosing cards.

Let us then get back to testing. You might be wondering – if theory is so much more reliable than testing, and if you should consider theory over results when building a deck, why would you even play the game? You could just sit at your desk and theorycraft for days without actually trying anything. Is playtesting completely useless?

Not at all. Playtesting is very useful – people just use it for the wrong things. Playtesting should not determine your card choices, or at least not directly. Instead, playtesting is useful for two things:

First, playtesting is useful for identifying problems. Even if you’re an amazing player, you’re still not a computer. Sometimes, things get past you in theory. Maybe Tibor sounds like the correct choice in theory, but then you test it and you realize that the bronze cards you’re giving your opponent are worth much more than 8 points of value. Maybe Triss: Butterfly Spell sounds like a fantastic card and you just forget that you don’t run that many bodies. Either way, things get past you in theory, no matter how good you are. By testing the things that you’ve decided through theory, you get the chance to notice things that might have slipped your mind before.

2 – Second, playtesting is important for perfecting your execution. While theory should be at the forefront of deckbuilding, actually playing the decks that you build properly is also important. And you can’t actually play a deck properly if you haven’t played with it. Playtesting, then, is a great way for you to get familiar with a deck and to guarantee that your play with that deck is optimized. While it shouldn’t determine what you put into your deck, it should certainly give you a hand in figuring out how you should play that deck. It helps build good habits – decks will have an optimal order to play cards in many matchups, and you want to get used to doing things perfectly so that when an irregular situation or matchup arises, you know what the norm is and can consider how much you may need to deviate from it.

Ultimately, though, math is king, simply because math doesn’t change. Let’s do an experiment: flip a coin. Flip it again. If you flip a coin two times, it might not land once on heads and once on tails. That’s variance for you. But the math is sound. You still have a 50% chance each time of it hitting either side of the coin. The variance doesn’t matter – you can’t control variance as a player. All you can do is guarantee that the math is in your favor. If you make sure that you have the highest possible probability of winning, it doesn’t matter that you lose the 75% draw in that one game – what matters is that, over the long run, you’ll win a lot more games than if you had gone with something else. All we can do as players is maximize our percentages: and for that, variance has to be minimized. As such, theory and hard math should always rule over practice and anecdotal evidence.

So, to sum up:

  1. In deckbuilding, theory in crafting should be valued over results in testing.

  2. Testing should not be used to directly dictate card choice or substitution.

  3. Testing should be used to correct technical play and to make sure you didn’t miss something in your theorycrafting.

  4. Math doesn’t change. The path to victory is paved with math.

I hope this will help you get consistently better results at GWENT – and next time you get the urge to say that cards don’t matter, remember that what doesn’t matter is variance. Yes, you lost that game because you had bad luck, even though your deck was better: but in the long run, you’ll win a lot more games by having better cards.

Thank you for watching, and tune in next week for some more advanced GWENT content!

Gwentlemen

Gwentlemen
Supporting the growth of the competitive scene of Gwent