November 2, 2015
Table of Contents
The Grind, Part III: Deck Selection
Hi guys, welcome to ‘The Grind’. The Grind is a four-part series aimed helping you to achieve the dream of ranking up to Legend (for the first time). As always guys, comments, likes, questions, etc are welcome.
The four parts are:
- Part I: Introduction & General Tips
- Part II: Research & Self-Mastery
- Part III : Deck Selection
- Part IV: Psychology
Right now you are reading Part III. This article is cram-packed with general advice about what decks to play. Okay, let’s begin!
Choosing Your Deck: The Basics
***To understand who ‘Johnny’, ‘Timmy’ and ‘Spike’ are you will need to read the personality section in Part IV***
Okay so let’s start talking about one of the most important considerations you will make all season: What deck(s) should you play?
Here’s a quick test, fill in the blank in the following sentence:
“I Should select the ____ deck in the format”
If you answer was “[the] best fucking [ deck]” then congratulations, you are correct. If you answered something like “[the] most fun [ deck]” then I would recommend that you read the personality section (In Part IV). Yes, I’m looking at you Timmy!
In short, we need to be Spikes, and the Spike thing to do is pick the ‘best fucking deck’ in the format. No questions asked. And guess what Johnny? The ‘best deck’ isn’t that project you have been working on for the past week!
TIP #1: Net-deck; You must play the ‘best deck’.
How do we find out what the ‘best decks’ are? Well any good meta report should do that (e.g. Tempostorm’s Snap-shot).
Once you have looked at all the ‘Tier One’ (on HSP the top decks are listed under “Decks to Beat”) decks, you need to check your collection and see what you can build. If you don’t have the cards, consider blowing some cash on packs to get the cards you need.
TIP #2: Budget decks = Budget results.
It might be tempting to try to take a good deck and then replace X Legendary that you don’t own with something you do have. I would recommend against this practice however, and that is simply because budget decks usually get budget results. If you don’t own Dr. Boom you could try replacing it with something you do have (such as Kel'Thuzad). But here’s the problem: If Kel’Thuzad was better in the deck than Dr. Boom then it would be in the deck. In other words, Kel’thuzad isn’t as good as Dr. Boom and your results will probably end up reflecting that.
And let me remind you once again; The Legend grind is hard, even a 1% difference in win-rate can make huge difference in whether you succeed or fail. Don’t make things harder for yourself by playing weaker versions of Tier One decks. With that said, some decks are a lot more flexible than others. Playing ‘Freeze Mage’ without Alexstrasza is basically impossible, but if you don’t own Tirion Fordring you can still play some versions of ‘Secrets Paladin’ (e.g. the aggressive Secretkeeper & Divine Favor builds).
In short: search the internet and your collection to see what you can build. If you lack the cards spend some cash on a few packs, you tight-arse.
TIP #3: Don’t pick something too hard to play.
As of typing, Grim Patron Warrior is generally considered the ‘best fucking deck’ (
it won’t be the moment the Warsong Commander nerf goes live, however Update: Patch is now live). However, as a general rule of thumb I would advise against picking such a deck. The reason for that advice is that this deck requires considerable skill to pilot well. And frankly, you probably do not have those skills (if you did, you would have hit Legend already and therefore wouldn’t be reading this guide!). I recommend trying to find the strongest deck you can that is as simple as possible to play.
Currently Secrets Paladin is a good candidate for this; it’s a strong deck without significant room for misplays.
During your Legend run you will probably use 2-3 decks. All of them should be “the best decks”. But what the ‘best deck is’ depends on where you are on the Ladder. This is the focus of the next few sections…
Your Deck: Ranks 25-to-10
***In this section I am going to start by stating what the ‘conventional wisdom’ is. A few sections later I will question these ideas and offer my own. ***
Okay, let’s begin!
The conventional wisdom states that at they very start of the season speed is more important than win-rate. Thus, the ‘best deck’ at this rank is something that wins and loses quickly.
A deck like Warlock Zoo and/or Face Hunter are usually good choices. I know, I know, I can hear all screaming at me: “Face Hunter is skill-less Cancer!!”. For those of you that do have that attitude, please read this and see if I can convince you to give Face Hunter a shot.
Other players will be doing this as well, so you might be tempted to try to counter the meta and play Control Warrior. Doing this may increase your win-rate by a few percentage points, but on the downside you are also probably playing 6-7 fewer games and hour. To progress at these ranks all you need is to maintain a 50-ish% win-rate and win-streaks will do the rest!
In short, the conventional wisdom says you ought to play fast decks, since the extra win-rate a slower deck might give you is not worth the time it takes to complete a game.
Now, at these ranks I am ready to bet that there are plenty of Johnny’s and Timmy’s reading this thinking that they can just play a ‘fun deck’ at these ranks because winning and losing isn’t all too important. Well, I’d urge you to reconsider, for two basic reasons:
- Remember in Part II I said you want to be practising against Ex-Legend players? Well, you are not going to learn much from that experience if you are playing some ‘goofy’ deck.
- The faster you hit Rank 5, the more time you give yourself to make that final push from 5-to-Legend.
I’ve seen it happen before: Some bloke plays causally most of the season, but three-quarters through the Season they start to take things seriously and cruise all the to Rank 3. And then it hits them like a tonne of bricks:
“Ah,*if only* I had more time! With 3 more days of playing I could have hit Legend. But alas, there are only three hours left in the Season. Sigh. Oh well, I guess I’ll just have to try again next time.”
Don’t be that guy that almost made it; get yourself into ‘try-hard mode’ within the first minutes of the new season.
Your Deck: Ranks 10-to-5
At this phase in your climb, you are probably getting to the point where your very fast ‘cancer’ deck is starting to become ineffective (e.g. maybe the meta is slowly moving toward Control and Mid-range decks). Thus, its time to change decks. Your new goal is pick a deck that:
- Is (still) reasonably quick.
- Good win-rate (55+%)
- Might be good for the Rank 5-to-Legend push.
Of those three points, I’d like to draw your attention to the third point; If you start playing a deck that could be good at Ranks 5+ now then by the time you reach Rank 5 you already have a decent amount of experience with the deck. And that experience should translate into better play, and better play equals a higher win-rate.
Also, once you hit Rank 7 or so I’d recommend that you start paying close attention to the meta. This is because when you hit Rank 5 you are going to want to have some meta-data to inform deck choice.
TIP #4: As you get closer to Rank 5, start keeping track of the Meta.
The data you want is not your win-rate versus various classes, rather, you want to record the popularity of the various decks in the meta. If you downloaded a deck tracker like I told you to do (in Part I) this data collection would be automatic.
Are Fast Decks Better?
I want to revisit something I said just a moment ago:
“At the start of the Season: Speed > Win-rate”
Earlier above I regurgitated this advice because this is what basically everyone says and does. But is this advice actually correct? In what follows I’m going to study the problem via computer simulation.
So, the conventional wisdom says that (up to Rank 5) it is better to climb with fast decks because a low(ish) win-rate with fast games climbs the ladder faster than a slow deck with a high(ish) win-rate. Well, this seems like an empirical question we can actually test. While writing the Python code (click here to run it in your browser) I also had a bit of brain wave and wondered whether deck-switching could be an effective ladder strategy:
DECK SWITCHING STRATEGY:
The idea of this strategy is to mix speed with win-rate. First we play a really fast deck (e.g. Face Hunter) for a bunch of games. The moment you win two games in a row you swap to a slower, higher win-rate deck. Once you lose a game you go back to Fast Face Hunter. Rinse and repeat.
The Theory is that by switching to a higher win-rate deck we maximise our chances of getting bonus stars from win-streaks. Meanwhile however, the use of the fast deck at all other times should enable us to play a decent number of games per hour.
So what’s the optimal way to climb up to Rank 5?
- Fast decks?
- Slow Decks?
- Win-streak Switching?
As I ran a bunch of tests with my code I found something interesting: there is no ‘general’ answer. With the right set-up all strategies have a chance to shine. As a simple illustration:
Deck (A) has a 60% win-rate and takes on average 15 minutes per game.
Deck (B) has a win-rate of 50% and takes 5 minutes on average per game.
What deck hits Rank 5 faster? Well, Deck (A) is faster by about an hour.
BUT here is where things get a little interesting, suppose we add 3% win-rate to both decks (but keep time per game the same).
In this case Deck (B) is a about 6 hours faster!
Okay what about keeping the original win-rates (50%, 60%) but add 3 minutes average time to each game.
In this case Deck (A) is once again faster, by about 15 hours
What about deck switching? Well in the first case Deck switching is the best strategy (it’s about 3 1/2 hours faster than just using Deck A exclusively). In the second case it is the second best strategy (about an hour slower than Deck B). In the third scenario deck switching is once again the best strategy (by about an hour). Sample Size of all tests above = 5,000.
I think it is fairly easy to make sense of these numbers, adding X to each side obviously improves one deck more than the other. For example, adding 1 to 100 is a bigger percentage increase than adding 1 to 200, so I think that is what we are seeing here.
Anyway, from my tests it would seem that every strategy seems to be viable, given the right initial conditions. In the spoiler below I have included three more test cases (60k trials), each test case shows a situation where one of the three respective strategies is optimal.
**ALL TESTS RANK 20-to-RANK 5**
SAMPLE SIZE: 60,000
FACE HUNTER: 53% win-rate, 5 min per game.
HANDLOCK: 59% win-rate, 9 min per game.
- Hunter & Handlock: 21 hour(s) and 35 mins
- Hunter Only: 22 hour(s) and 54 mins),
- Handlock Only: 23 hour(s) and 12 mins
SAMPLE SIZE: 60,000
FACE HUNTER: 50% win-rate, 5 min per game.
HANDLOCK: 53% win-rate, 7 min per game
- Hunter & Handlock: 33 hour(s) and 2 mins
- Hunter Only: 36 hour(s) and 27 mins
- Handlock Only: 32 hour(s) and 5 mins
SAMPLE SIZE: 60,000
FACE HUNTER: 57% win-rate, 6 min per game.
HANDLOCK: 65% win-rate, 12 min per game.
- Hunter & Handlock: 18 hour(s) and 26 mins
- Hunter Only: 18 hour(s) and 9 mins
- Handlock Only: 21 hour(s) and 2 mins
If I had the time & inclination, maybe I could plot a 3-d graph in order to figure out under what circumstances each strategy is best, but I am far too lazy to fuck-around with that. And besides, this article is supposed to be about providing practical tips to get you to Legend. Which begs the following question:
“What practical use is this information?”
Well, I think there are two issues that we must address before I can offer any practical advice:
- Is the code correct?
The first problem is to suggest that the results could be the product of faulty reasoning/faulty code. The main reason for providing a link to the code is to enable others to check what the simulation is doing. If there is an error or two, hopefully they can be corrected and we can re-run the experiments. But even if we assume that the code and the inferences are good we still have one big glaring problem: Uncertainty.
In my experiments I took for granted some very important information; I assumed that we knew (before playing) what the win-rate and what the average game length is for the fast deck and the slow deck. In the real world, these numbers are not known; we have to guess at them.
Basically, for this sort of analysis to be useful we need to make good starting assumptions about game length and win-rate. And that usually requires a large data set. On the bright side, lots of people post data analysis on Reddit and so sometimes the data you need might just fall onto your lap. As an example, ‘Staluxa’ recently studied game length. Here are his/her results (sample size: 600 games):
|Druid (mid-range)||Warrior (ctrl)||Hunter (face)|
|Average time (m)||6.12||7.36||4.96|
As with all data sets there are always problems; in this case the sample size is 600 games (problem: too small) all played by a single individual (problem: some players a naturally slower than others (think Lifecoach) and so these numbers might not accurately represent the time it takes *you* to finish the average game) at Ranks 5+ (problem: maybe games are faster on average at lower ranks?). But despite these issues the data does give us some insight as to how long games take on average. It is interesting to note that the time difference between Face Hunter and Control Warrior is about 2.5 minutes, in most of the tests above I assumed that the difference between the ‘slow deck’ and the ‘fast deck’ would be a lot more than that. So okay, let’s quickly re-run a few tests, this time we will use real game length numbers and then assume a win-rate difference of 5%.
|GAME LENGTH [Mins] (Ctrl Warrior):||7.4||7.4||7.4|
|GAME LENGTH [Mins] (Face Hunter):||5||5||5|
|WIN-RATE [%] (Ctrl Warrior):||55||57||60|
|WIN-RATE [%] (Face Hunter):||50||52||55|
| TIME TO RANK 5 [hh:mm]
| TIME TO RANK 5 [hh:mm]
| TIME TO RANK 5 [hh:mm]
Alright so what’s the conclusion? Well it seems that perhaps the classic advice of playing fast decks to climb is incorrect; spending 2-3 additional minutes per game is worth it if you get and extra +5% in win-rate.
Okay, so let’s try to wrap this section up with some practical laddering tips. For the tips that follow I am going to assume that the following two premises are true:
- The difference in time between a ‘slow deck’ and a ‘fast deck’ is assumed to be about 2-5 minutes.
- The difference in win-rate between a ‘slow deck’ and a ‘fast deck’ is assumed to be about +/- 5%
When these two assumptions are true I would recommend that you climb with the slower deck (or play the fast deck with win-switching). I say this primarily for two reasons:
- Even in the cases where the ‘fast deck’ hits rank 5 first the difference is often not that much. In short, the slow deck doesn’t actually suffer a significant time penalty, imo.
- If the “slower deck” is what you plan on playing Rank 5 to Legend then it could be a good idea to play the deck at the outset: The extra experience with the deck and its match-ups should help improve your post-rank 5 win-rate with it.
In conclusion; so long as the ‘slow deck’ in question isn’t a total snail (20 minutes a game!? fuck that) and wins about 5% more often the extra experience you get with the deck probably trumps any small gains in time you would get by a playing a fast deck (and this conclusion generously assumes the fast deck is actually faster, which in many cases it isn’t).
The other option is to ‘play it safe’: In every single test I have run thus far the switching strategy is always the best or the second best strategy. Often the time this strategy takes to hit Rank 5 is usually very close (i.e. within 1-2 hours) of the best strategy. Ergo, this is strategy offers a good balance of speed and experience (with your Rank 5+ deck).
TIP #5: From Rank 20-to-5 play a fast deck and then swap to a higher win-rate deck (ideally the deck you want to play to Legend) in order to maximise the chances of getting (and maintaining) win-streaks.
With this said however, I would like to stress that due to the fact the real numbers are usually unknown this tip might be a bit tricky to pull off in practice; in many cases you simply have to guess that Control Warrior has a higher win-rate than your Face Hunter deck. And rather unfortunately these guesses may turn out to be wrong.
Your Deck: Ranks 5-to-Legend
Okay, so at Rank 5+ you don’t really want fast games, you want to win…a lot. You want a strong and consistent deck that ‘fits’ the current meta.
TIP #6: At Ranks 5+; Win-rate > Speed.
Do demonstrate that, let’s again return to Python. I ran a few experiments (starting from Rank 5). It turns out that with win-rate of 50% and games played at blistering speed of 2 minutes and 30 seconds each takes about 40 hours to hit Legend. Which is the same amount of time it takes to hit Legend with a win-rate of 55% and 10 minutes per game. Thus, it would from this test that getting an extra 5 percent in win-rate seems to be worth playing 4x longer games for. In another experiment I doubled the time difference (5 versus 10 min respectively) and had the win-rates differ by 3 (52 versus 55), in this case the faster deck hit legend an hour quicker. Basically what I think this two experiments show is that increases to win-rate at Ranks 5+ is a lot more valuable than decreasing the time it takes to complete a game.
Okay so if win-rate is more important than speed it makes sense to swap to a high win-rate deck.
“BUT how do we know what that deck might be?”
Do you remember I told you to collect data a little bit earlier? Well, let’s make use of it now, open it up and have a look.
As you Analyse your data the first question you need to ask is:
“What are the three most common match-ups you face?”
Now that you have a list of those three decks, you want to try to select a deck that best counters the three most popular match-ups. Now, I’m sure some of you are now wondering…
“Why only try to counter the top 3 most popular match-ups? Why not the Top 4? The Top 5? “
Well, we could try to use game theory to ‘solve’ for the whole meta-game, but the problem with this approach is essentially to do with sample size: its unlikely that you will ever get hold of enough data to reliably claim something like “Shaman represents 10% of the meta”.
The second reason is simply going to be an assumption; the ‘Pareto principle‘ is an idea that applies to a wide variety of phenomena, basically it concerns how resources get distributed across systems. Well, I’m just going to assume that something like this is true of Hearthstone and boldly claim that “In any given meta, the top-3 decks probably account for about half of all your games”. For those interested, in the spoiler below you can see a quick test of this assumption using real data (TL;DR… it seems legit, lolz).
Games played = 104 (Ranks 5+)
Top 3 Most-Played Against Decks:
Deck A = 22 Games
Deck B = 16 Games
Deck C = 14 Games
The Math: (22+16+14)/104 * 100 = 50%
Games Played: 198 (All Ranks)
Top 3 Most-Played Against Decks:
Deck A = 36 Games
Deck B = 31 Games
Deck C = 28 Games
The Math: (36+31+28)/198 *100 = 47%
Okay sure, this is only two tests. But nonetheless, it seems likes my hypothesis has some substance to it.
In a nutshell my tip is:
TIP #7: If you can, Pick a deck good against the 3 most popular decks.
For example, suppose you are playing a meta where the main three decks are:
Freeze Mage, Control Warrior, Face Hunter
In such a Meta picking Control Warrior is a good choice; You are favoured against Hunter and Mage and are equal against Control Warrior. In such a scenario your deck selection gives you an edge.
Suppose, for example, Mage and Hunter are both 60% match-ups and the mirror is (obviously) 50%. If you play 90 games (30 versus each deck) your record would be 51-39, which equates to a win-rate of 56.6%.
If we further assume all your opponents are of equal skill (i.e. win-expectancy = 50%), that would imply that good deck selection added 6% to your win-rate.
In short, finding a deck that beats the current meta makes the Legend Grind way easier than it otherwise would be. But this is a hypothetical example, in real meta’s there are 9 classes and a great deal of uncertainty as to what the meta *actually is*. Okay, let’s look at some real stats shall we:
In true Johnny Fashion I built my own Control Hunter deck (you can read about the build process here). These are my stats from the September Season. As you can clearly see, at ranks 5+ I played 104 games and averaged a 57% win-rate. I think sometimes people confuse what the numbers mean; in this case the 57% is not a measure of how strong the deck is, rather, it’s a measure of how well the deck fits into a particular meta.
If you look at the match-ups, the most commonly faced opponent is Secrets Paladin (22 games), followed by Hunter (16 games), and then finally Mage (14 games). Of these three match-ups only one is good, thus I think we can say that this deck was not particularly suited to the meta I was playing it in.
What happens to that 57% win-rate if we pretend I played 10 more games against Druid and 10 fewer games against Paladin? Well, if we assume the 80% match-up figure is accurate those extra 10 Druid games are likely to be 8 wins and 2 loses. So my Druid score is now 19-4. If we subtract 10 Paladin games (40% Match-up) then we are likely to expunge about 6 losses and 4 wins from my record. Thus my Paladin score now stands at 6-6. My stats would then become 63-41 (60%). Notice also that the *true* win-rate against Paladin is 40%, but the 6-6 stats suggest it is 50%. This serves as a good demonstration as to why we need to be cautious when interpreting small data sets.
But anyway, the main point I’m illustrating is that win-rate is not necessarily an indicator of skill or deck strength; In a different meta (e.g. one with more Druid and less Paladin) my win-rate would increase significantly, and every percentage increase can drastically reduce the time to Legend (+3% could easily be about 5-10 hours of play).
In short, picking a good deck for the meta can significantly impact your Rank 5+ Win-rate. If you know what the meta is (which is why that data you collected earlier is really important) you can counter it. Countering the meta could add maybe +5% to your win-rate (if you are lucky), and thats a HUGE BOOST to get. Please remember however, that Meta’s can change rapidly, thus today’s solution might not work tomorrow. Thus every time you find a ‘window of opportunity’ I recommend cramming in as many games as possible.
Okay, now let’s look at something at the concept of Tech cards…
“Why do we ‘Tech in’ Cards?”
So you have ‘the core’ of your deck and then there are probably a handful of flexible slots where you can add cards according to your own taste; You could, for example, replace 1x King's Elekk with a Flare in your Hunter deck. And most likely this change will improve some match-ups at the cost of some others. At I guess I would imagine such a change would give you a big Boost versus Freeze Mage and Secrets Paladin, a moderate boost in the mirror and against Tempo/mech Mage, and a tiny boost versus Druid (due to your ability to unstealth Shade of Naxxramas). Against everything else, this change has probably reduced your win-rate.
For arguments sake, lets suppose this change increased one of your match-ups by 5% but another match-up has got 5% worse as a result of this change. If we assume a meta with all decks equally represented, this change ought to have no bearing on your win-rate (the terms cancel each other out). But in a meta where X deck is more popular than Y deck this tech change can have a meaningful impact on your over-all win-rate (if you add 5% to 100 and subtract 5% from 50 the result is net positive). If you are interested in this sort of stuff here’s an article on it.
Okay so I think we can all agree that with a big enough sample of games this sort of deck-tweaking really can improve your win-rate. Nonetheless, I would advise you against this practice. If you want to counter the meta, I’d advise you to pick a different deck instead.
TIP #8: Don’t add Tech-cards to your deck UNLESS you see the Pro’s doing it.
The reason I make this suggestion is because it is really easy to mess it up. For a somewhat humorous example check out this vod where Pro-player Cifka still manages to lose against Freeze Mage despite running (and drawing into) double Kezan Mystic and a lot of heal cards — It’s a classic story of ‘the counter-deck that failed to counter’.
You might think it is easy to add tech cards to your deck and ‘just win’ but often it isn’t that simple. For example, did adding Flare really improve your Secrets Paladin match-up? Could it not be possible that with one less 2-drop in the deck you fall behind on board faster as a result of this change. Might falling behind on board cost you more games than destroying secrets wins you? And that my friends, is the crux of the problem, sometimes changes can have unexpected consequences. If you look at the Cifka game for example, adding in all those heals and Mystics meant he had to cut something else, maybe the stuff he cut would have been able to deal with the Archmage Antonidas + Frost Nova combo.
Basically, you don’t want to get all the way to rank 3 and then lose a bunch of games because your tech cards don’t pan out quite as planned. My advice is really simple: Let the pro’s experiment with tech cards. If said experiments appear successful copy them.
In short, let someone else make the mistakes so you don’t have to!
Time to Switch Decks?
Okay, so let’s say you have been doing reasonably well, you’ve just hit Rank 3 and have started to notice that the meta is starting to change. You have a choice:
“Do I carry on playing the same deck or do we swap and (once again) try to counter the meta with our deck choice?”
When it comes to Legend climbs, you will see plenty of proponents on either side of this debate: I’m not going to pick a side.
- Pro’s of swapping decks: Maintain a competitive edge over the competition
- Cons of swapping decks: Cost (i.e. do you have the cards?), How well can you play this new deck?
Pro players can just swap decks almost willy-nilly because they can play all decks competently. This might not be the case for you; if you have spent the last 100 games playing Deck Y there is a decent chance that you can play it really well. If you then swap to a deck that you have never played before then the lack of experience might be more significant than the extra edge you get from countering the meta. At the end of the day, you will have to make that call.
And that concludes Part III of my Grind to Legend Series. Make sure you check out the final instalment!
I sincerely hope this guide helps you hit Legend this season. If you have questions/queries please use the comments section below.
Summary of Tips
Here’s a quick list of all the tips in this article:
- Net-deck; You must play the ‘best deck’.
- Budget decks = Budget results.
- Don’t pick something too hard to play.
- As you get closer to Rank 5, start keeping track of the Meta.
- From Rank 20-to-5 play a fast deck and then swap to a higher win-rate deck (ideally the deck you want to play to Legend) in order to maximise the chances of getting (and maintaining) win-streaks.
- At Ranks 5+; Win-rate > Speed.
- If you can, Pick a deck good against the 3 most popular decks.
- Don’t add Tech-cards to your deck UNLESS you see the Pro’s doing it.
References and Further Reading
‘THE GRIND’ SERIES:
- Smashthings, The Grind, Part I: Introduction
- Smashthings, The Grind, Part II: Research & Self Mastery
- Smashthings, The Grind, Part III : Deck Selection
- Smashthings, The Grind, Part IV: Psychology
ON PICKING DECKS:
- Modded, Metagaming Demystified: Breaking Down Your Deck
- Smashthings, Slow Decks, Fast Decks, Deck-Switching [Python Code]
- Smashthings, Playing to Win Part One: Defeating the Scrub Mentality
- Staluxa, Game Duration Analysis for 600 Games [Reddit Post]