I saw some discussion about high-variance decks in the Titan Hyper Fast Advance thread and I’m here to put a stop to it. (No disrespect to any posters there, this is a subtle error.)
The classic discussion of variance goes something like this. I can bring a very consistent deck to a tournament, say Kate. All my opening hands give me a 60% chance to win. Alternatively, I could bring a swingy deck like Siphon Maxx, where half my opening hands are nuts with Knight, Siphons, etc, and have 90% chance to win; and the other half are garbage and have 30% chance to win. For a 9 round tournament, it’s unlikely that Kate will carry me to the top via variance, but Maxx might. So I should bring Maxx and hope for good hands.
The problem with this is that the win rate of each deck is the same! They both have a 60% chance of winning any given round. Suppose the prestige works out so that you need 8 of 9 wins to take down the tournament. Then both decks have exactly the same percentage to win: P(Binomial(.6, 9) >= 8).
For more discussion, e.g. (a) you are uncertain about your deck’s win percentage, and (b) you are uncertain about the metagame, check out this article by Frank Karsten.
Fallacy: Assuming we’re talking about decks with 60% win rates in all cases.
My prime example of a high variance deck that was worth playing was NEH (circa 4 months ago); where you might be pushing a 70% winrate or higher. You still might have games where you lose hard, but your ability to just win makes that fine most of the time.
My second example of a high variance deck was Noise. Where sometimes again you can just win, but you have play options outside of that strategy that occasionally “just wins” you the game.
I believe I even argued that the deck that was posted probably didn’t have the 60% win rate at the top and probably wasn’t a good choice because of that.
High variance can cause you to win games you shouldn’t. And can get you to places you didn’t think you’d normally belong in a tournament. But unless the deck has a consistant side, its probably a bad tournament deck if your point is to win.
Not to step on your crusade or anything against “high variance” decks or people who want to play them. They are probably not a good a way to win a tournament; so, context matters.
Back to your fallacy, if they’re all 60% win rates, play the high variance deck if you want. It’s got the same win rate.
Not sure how NEH can be considered high variance. Before clot it had an absurdly high win-rate at all levels of play, due to basically all the ice being good at every stage of the game, all the operations being playable and all the assets being instantly playable as well as really strong card draw and a reward for doing what you should be doing anyway (scoring agendas). Nothing about that deck seems high-variance to me.
I think higher variance decks are best to leverage a lack of playskill relative to others in the field. I have always felt comfortable playing a deck like RP, that’s very consistent despite a lot of players claiming greater than 50% winrate against it, because as long as you can leverage playskill, you’ll be able to get a lot higher than the aggregate matchup percentage against most of the field.
As a very strong player, the things I want to play against least early in tournaments are stuff like kill decks, (punitive takeover especially, but also anything with Mushin in it) NEH, (high variance not because your deck is inconsistent but because so much hinges on how good the runner’s random accesses are), Leela in some cases, (bounce your HQ ICE siphon turn 1 bullshit), Noise, (did you find Pawnshop quickly?), and standard breaker Anarchs like Whizzard who you usually have to beat by pushing gear checks.
High variance corp decks are usually like that, where your deck isn’t necessarily inconsistent, but you force the runner to play on central accesses or make high variance guesses like against Mushin. You can beat players a lot better than you if they steal Government Takeover on turn 1. Higher variance runners are usually more of the inconsistent but powerful type, (Whizzard can be great if you have a good start and the right breakers, Leela’s ability can be next to worthless or totally gamebreaking based on 1 or 2 random accesses during the course of the game).
The core of the deck is a high variance move: the astro chain. Which is a combo where you must assemble the right cards together, typically in the right order without the corp interrupting you (previously: biotic/astro/sansan & cash vs legwork/imp; currently: biotic/astro/sansan/shipment/cvs vs legwork/imp/clot/clone chip/smc)
The rest of the deck took this highly variant combo and wrapped it up in a very tight and consistent package, but the core of the deck is this high variant monster we all have named and choo-choo’d numerous times. How do we know this is a deck which utilized a high variance tool and not just a really good consistant deck? When a tool came out that targetted the high variant portion of the deck, the deck required serious changes and its win rate dropped severely. The entire deck became high variance where it started waffling between hard wins and hard losses.
High variant combos and moves are very powerful, and when you can wrap them in a consistant framework they become THE DECKS to beet. Other decks that fit this model are: noiseShop, 7PointShutdown, and siphonSpam. These decks were nigh unstoppable until the meta answered them. Granted, in some cases there were additional strong consistent components to these decks, but they were mostly one-trick ponies and those consistent pieces were placed to push the high variance move to happen more frequently.
All the agendas got milled
I got all the agendas scored in a single turn
My opponent can’t ever have cash and hence defenses
It would be wrong to describe these decks at the point that they were peek performant as just “high-variance” decks, but ultimately that’s what they were. Variant combo decks that were given enough support to become temporary monsters until answers appeared in the meta.
If you are deck building and you find a component that is high variance and helps you win games, it may be the first component of a deck like these. You catelogue that information and return to it, comparing it to newly released cards until you think the scales have tipped enough for you to take advantage of it.
The win rate is what ultimately matters. The article isn’t wrong about that. But recognizing raw power and how to harness it is an engineering skill you want to have.
Hmm I think I get what you’re saying but the astro-train is the icing on the cake that takes a good but not amazing deck and adds a swathe of “oh fuck he just won” to the possible outcomes. I won a ton of games fast advancing beales off a sansan or just scoring out behind a lotus field depending on the opposition. If the deck only won by comboing out with astro I’d agree, but that was just one super-strong facet of it’s overall strength.
I’m sure @calimsha would agree that the way to win with NEH wasn’t to dick around waiting for astros all game.
Removing those victories from the possible outcomes against opponents that utilise clot correctly is what’s beasted the win-rate of the deck. I just don’t see how that makes it high variance, when considering the deck before clot existed. It was ultra-consistent.
If astro had been a blank 3/2; you’re right that you could still win with that deck, but I think it wouldn’t have had nearly the power that it did; it’d be a lot more like where HBFA, a strong consistent deck with a pretty even power level between all of its parts. And you’re right that the deck was playable without scoring an astro chain; and yet, that astro chain made things so much easier didn’t it? Sometimes it just ran away with the game and there was nothing the runner could have done differently.
That’s what we’re talking about. A portion of the deck that guarantees the win. And the way you recognize a deck is worth playing is that, as you’re saying, the rest of the deck is still very good.
I really think I do agree with you. The deck as a whole was way too consistant to be considered a deck with variance-win-rates. But, it was a high variance style of play, where you could rely on a portion of your games being really fracking easy.
The classic example in my history is Crazy Train. A deck which looked really bad on paper, but could harness that train (plus turtlebacks) for an almost instant win when it came time. When I went up against a particularly skilled opponent on the edge of making top 4 in a store champ (@spags) the deck totally curved out in a completely unninteractive manner such that the better player couldn’t win. The rest of the deck was… fine. But that core, when it went off meant I didn’t need pure skill to win.
High Variance are usually these all in strategies. Cambridge PE or Anatomy of Anarch Reina are for me high variance decks. But don’t forget that you as a player mitigate this by your gameplay decisions as well. Basically you can always decide between high variance or low variance action to take in your turns.
@mediohxcore hit it on the head (and this was also in Dave’s original article) - deck choice needs to consider the average skill level of other players relative to you.
It’s way oversimplified to say a deck has a 60% winrate.
Rather, put player skill with the deck on the X axis and opponent skill against the deck on the Y axis and then you have a chart.
A reasonably strong player with Reina Headlock or Cambridge PE can win for example 90% against Joe Scrub, but only 30% against someone who knows the matchup inside out.
On the other hand a complex deck like Shutdown combo may be completely unusable at low levels of player skill, but at high level of player skill you can calculate the path to victory in many scenarios.
One argument Sutcliffe’s article makes which hasn’t been mentioned, is that if enough players play the high variance option, you have to play it to win.
The example is from Magic: You consistently see the same people in the top 10 in the big tournament, but the winner is almost always somebody from the tier below. That’s because the best of the best choose consistent decks which maximize their chances at getting top 10 finishes (and pro tour points and cash prizes). But if enough “lesser” players bring the high-variance decks, one of them will win - because somebody will luck out. Most of them won’t, of course. But if 50 players pick the gambling option, there’s almost no chance that none of them will hit the jackpot.
I don’t think this is necessarily true, largely because of elimination rounds. If you have 6/8 of the top 8 play some consistent deck, (midrange?) and 2/8 of them are tier 2 players with some inconsistent deck, (burn that really wants turn 1 goblin guide or something?) the players playing the inconsistent deck, while maybe having a better shot than if they were playing something else because their deck is high variance and they would have been outplayed with something else, still don’t necessarily have as good of a shot as the rest of the field, as they have to get through 3 elim rounds without taking a bad beat to win.
I think the big misunderstanding here is what “high variance” actually means. Deck performance is not summed up only by “win rate”. The better metric that would really illustrate deck variance is score differential. Score will always range from -7 to +7. Let’s say that we have two decks, both of which, on average, win 50% of the time. The high variance deck has score differentials of -7 50% of the time and +7 50% of the time. The low variance deck scores +1 and -1 respectively (these are the most extreme examples but serve as good illustrations). In practice, that high variance deck can almost autowin, but when the win fails, it fails big. The low variance deck will perform very consistently, but is highly sensitive to play errors. If you make a mistake that is worth a point (or less!) to the opponent, chances are you’ve given away the game.
The analogy in the video game world is “burst damage”. Often burst damage is preferred simply because it gives the opponent less of a chance to react, but the tradeoff is that if the damage can be mitigated there’s usually no follow up.
So why does this come in to play when talking about high skill vs. low skill players? In order to win a tournament, a player needs to be able to string wins together one after the other, making the swinginess of a high variance deck unacceptable for someone who can keep up the consistent play needed for good results with a low variance deck. Low skill players though who can’t successfully pilot low variance decks to the same degree may be better served by a high variance deck that can sometimes pilot itself to a win with the right combination of circumstances.
Being highly sensitive to play errors is a really important factor given how long larger tournaments can run.
Personally, I try to avoid playing such decks in a longer tournament because I know that I fatigue easily and will make those mistakes more often when it’s 8 or 12 hours in.
Conversely, if I estimate that I’m in a meta where others will fatigue easily, I’d want a deck that can punish such a mistake quickly and heavily (with a flatline, for example) so I have a larger rest/rehydration break between rounds.
The Titan deck in the OP is really good for this, as you can catch people out when they make a risky run, and worst case you lose quickly and can relax.
Another thought is that variance isn’t fixed - different matchups can have different levels of variance.
Take, for example, a non-MaxX (because she makes this discussion even more complicated) Anarch build running the core set Anarch breakers + Net-Ready Eyes and D4V1D, a few SMCs, and some reasonable card draw.
This deck has some reasonable consistency breaking a lot of ICE, especially with a few Parasites to remove problematic ICE. However, Yog runs into it’s normal nemesis- Lotus Field. In order for this deck to break a Lotus Field, it has to draw a NRE, no ifs ands or buts. Lacking a hardware tutor, we’re relying on our card draw to get us there.
This means that this deck is higher variance against decks packing Lotus Field than against those that are not. NRE is great, but in a not-Lotus matchup its just a support piece the deck can get by without. In a Lotus matchup, if it draws NRE, then we’re off to the races. If not, uh oh.
If it weren’t for these types of interaction factors the game would be much easier to solve (and much less fun to play!) Confounding factors like player skill vs. card value and conditional card value make quantitative analysis difficult if not impossible, enabling the internet slap fights that we all love so much.
I agree with most of the comments. My point was to distinguish between
“Kate is consistent with 50% win rate, but Maxx is swingy and flat out wins half her games, flat out losing the other half. I need to ride a variance spike to win the tournament so I’ll bring Maxx.” (this is nonsense)
and
“Sure, pros have a high win rate with kate, but I am noob and would make play mistakes and lose. But I can bring Maxx and I won’t screw up Maxx’s nut draws, so I can still come close to the 50% win rate.” (this makes sense (but you should probably just get better)).
Edit: if you understand the difference and phrase things correctly, great, I was not correcting you :). But I expect that many people do believe there is such a thing as an inherent swinginess of decks, even in a fixed matchup between two fixed players. I wrote the post to correct that idea.
