It is not a rare sight to see competitive players complain about the ladder system in Hearthstone: it happens at the end of every season as the competition for those coveted top 100 spots is fierce, and perceived to be in in a large part a matter of luck in addition to skill.
In this post, I will examine the current ladder system, showcase why it is fundamentally broken for this particular environment, and propose an alternative model that could be used to fix it.
The basics of the Hearthstone ladder system
The ladder in Hearthstone is divided into 25 ranks above which is the Legend rank. Within the first 25 ranks, a player earns a star each time he wins a game and gains a rank for each two to five stars gained, depending on the rank. Furthermore, a player loses a star for each loss starting from rank 20, and gains a bonus star for each consecutive win starting from his third win of the streak up to rank 5. There are no more bonus stars starting from rank 5, so to proceed from rank 5 to legend, a player needs to win 25 games more than he loses.
On legend rank, the stars are replaced by a rank number that shows how the player’s matchmaking rating (MMR) compares to other legend players in the region. The MMR itself is not displayed.
The matchmaking system on the ladder is based on rank on ranks 25 to 1, and on the player’s hidden MMR in legend.
Based on everything we know so far about Hearthstone’s MMR system, it is most likely the same MMR system Blizzard uses in its other games as well, that is to say an improved Elo-based system akin to Microsoft’s Trueskill. You can find a presentation (sadly, with poor audio) by Josh Menke on the system on Youtube.
One of the major improvements over the standard Elo system is the evaluation of the uncertainty regarding player skill: if you win or lose a lot, the system becomes more uncertain of your skill level, and widens the range of opponents you can meet to help find your correct skill level. This is easy to see on legend ladder when climbing ranks, as the rank of your opponents increases rapidly when on a win streak.
The ladder is reset each month and so is the MMR rating of each player. Depending on the rank a player reaches, he receives bonus stars at the start of a new season. For example, legend players begin the new season from rank 16. On the EU server, around 12,000 people reach legend each season, and from the statistics shown in Blizzard’s competitions we know that top players reach legend in around 130-200 games.
There are season rewards depending on the rank reached. These start at a seasonal card back on rank 20 up to a number of golden cards for a legend rank finish. Furthermore, the top 100 players receive Hearthstone Championship Tour (HCT) points, enough of which will then qualify them for participation in Blizzard’s seasonal championship preliminaries.
The problems of the Hearthstone ladder system
The main problem with the Hearthstone ladder system relates to its use in esports with some issues related to the grind presented by the monthly reset and to the way ranks function.
The demotivating monthly grind
As it takes a lot of games to reach legend, easily at least 200 games for the majority of players capable of reaching that rank, there are lots of Hearthstone players who only embark on the journey once for the Legend card back reward that you receive when reaching that rank for the first time.
Many players simply choose to play up to rank 5 for the sweet spot of monthly rewards (the golden epic card) – and a number of them do so with a good win rate, which means that the system is never able to adjust to their correct level of play.
Thus, while the goal of the ladder is to rank players according to their skill and provide them with evenly matched opponents, the monthly reset prevents it from doing so as players need to play a lot of games before reaching their correct position again.
The win farming on low ranks
As there are cosmetic rewards for winning games in ranked mode – golden hero portraits for 500 wins with the hero – players have an incentive to try to win a lot of games, fast. Because the matchmaking below legend rank is not based on MMR, it is trivial for a player to concede a number of games to reach a lower rank and farm for easy wins there.
Granted, there is no matchmaking system that can completely eliminate this behavior, as you can decrease your MMR by losing as well.
The top 100 issue
The big issue with the Hearthstone ladder is the way top 100 legend players are awarded HCT points at the end of each season. HCT points from ranked play are currently awarded as follows:
- 1st: 15 points
- 2nd – 10th: 12 points
- 11th – 25th: 10 points
- 26th – 50th: 8 points
- 51st – 100th: 5 points
- Any other legend rank: 1 point
What many players have observed is that the high legend ranks are fickle, especially so in the last moments of a season. It is possible to lose tens of ranks in a matter of hours on the last day of the season without playing, as well as lose tens of ranks on the final day from a single loss. For a player around rank 25, for example, playing a single game can result in a top 10 finish or a finish outside the top 100 altogether.
This has led a number of exasperated players to remark that only the final day matters, nothing else. However, this is strictly speaking not correct, as there is no MMR decay during the season, so you only lose ranks when others pass you and not due to inactivity, and there does not seem to be huge MMR inflation either, so the MMR gained on the final day does not reduce the days before it into obsolescence. As an example from the June season, Apxvoid camped #3 legend on NA from 23rd June until the end of the season and (tentatively) finished #7. If there was huge MMR inflation, this should not be possible.
Nonetheless, the way the ladder is built has led to mass behavior where top players grind a ton of ladder on the final days of the season in an attempt to secure a top finish. They also generally stop playing when they reach a relatively high rank in order to attempt to protect it. This leads to a large clump of players sitting on almost the same MMR at the end of the season, so even a short win streak can propel a player from 100 to top 10 – with the risk that those games send him the other way down to 300.
The other rational response that has been caused by the system is that top players grind ladders of multiple regions: if they can sit on a decent top 100 rank on one region, they can then take risks to try to go for a top 10 or top 50 finish on other regions. This is hardly a desirable feature of the ladder system.
Given the randomness of this all, does it really make sense to distribute HCT points the way they are currently distributed for ranked play? Does a top 25 finish really earn twice the points of a top 50 finish with any objective measure? The answer is clearly no, it does not.
So, are the players simply wrong in their behavior? Would it be better to play more evenly throughout the season regardless of rank and see where you end up? No, unfortunately it would not. To see why this is the case, let’s take a look at what an Elo-based system is meant to accomplish. The Elo rating system was originally built to calculate the relative skill levels of chess players, not in brief stints, but over their entire careers. Now, let’s take this system and throw it into a card game where randomness plays a part and then reset the count every month so that we only have a limited sample size available. With the inherent randomness and the limited sample size, the only logical conclusion for a player is to first play enough to get roughly to where his MMR should be (the grind to legend more or less achieves this) and then wait for a lucky win streak to get as high of an MMR as possible, and stop playing for the rest of the season. This does not have to be on the last day of the season!
The problem is further exacerbated by the number of people competing. With hundreds of nearly-equal players competing for top 100 finishes, getting that top finish becomes more and more a matter of luck. We are already seeing things get worse and worse every season as the number of players increases, and it is about to reach a point where the system is simply untenable.
It has been suggested that showing the MMR to players would help, but in fact it would only serve to promote camping earlier in the season as if players know the general MMR level needed for a top 100 finish, they can try to get there earlier in the season and just camp their MMR from there.
Almost the same applies to making a specific MMR the criterion for points: you just strive to reach that, and then camp from there.
What we need is a system that encourages people to play while giving meaningful and fair rewards for ladder performance. It can be done, but only by changing the way championship preliminaries work as well. Next, I will show you how.
Why are the preliminaries so limited, anyway?
The fundamental problem stems in large part from the need to reward ladder finishes with spots in the championship preliminaries – and trying to keep the number of participants in the preliminaries limited at the same time.
In the beginning, this was a necessity. Blizzard had a limited number of partners to provide locations for the preliminaries, and they were not willing to hold them entirely online – a decision I agree with by the way, as the prestige of the Championships deserves an offline qualifier. Furthermore, the more participants, the more difficult and time-consuming it is to run a double elimination bracket through all of that.
However, now that some work has been put into it and more and more locations are available (a work in progress, obviously, as the quality of the locations is still not equal), the preliminaries could take on a vastly different shape. I am obviously talking of Swiss format, which enables increasing the number of players a lot as well as better distinguish the group of top players. So, instead of a 150-player double elimination bracket we could have 11 rounds of Swiss with 450-players into a top 16 cut for the Championships. If 450 players sounds like a lot, remember that Magic Grand Prix events often have more than 2000 participants. Hey, we could even have 1000 players for 12 rounds of Swiss into a top 16 cut, that could still be done over two days. In both of these examples, most players with an x-2 record would advance by the way, something I find highly desirable in a card game.
It is only through changing the way preliminaries work that we can relieve the pressure of rewarding short-term top 100 finishes on the ladder.
Giving a sense of progression – separating MMR and ladder
In order to avoid the upheavals of monthly ladder and MMR resets, and to ensure even matchmaking, ladder and MMR could be separated from each other. Ironically, Blizzard already does this in another game – Starcraft 2.
In Starcraft 2, your ladder points consist of roughly your MMR (once you have played enough games to enable the ladder points that start from zero to catch up to your MMR) and a bonus pool. The bonus pool is accrued continuously, but only becomes part of the player’s ladder points as the player actually plays games. You can also see how much unused bonus pool you still have available. This system incentivizes people to play in order to use up their bonus pool and also indirectly punishes inactivity, not by decaying your points, but by giving everyone else extra points.
The star system on ranks below legend is not necessary. In fact, it only promotes uneven matchmaking, as existing players who start to play mid-season start from ranks far below their skill level. Instead, we could have a ladder point system similar to Starcraft 2 and give players rewards as they accrue ladder points: a card back here, a golden common there, there is a lot of room to work with in this kind of structure.
Thanks to the bonus pool structure, all players can accrue a number of points each season no matter their skill, so a number of rewards can be tuned so that it is enough to be active to get them. Skilled players obviously gain points faster, and can reach a higher point totals over the course of the season.
This would make each new season something to look forward to, as you would again have the opportunity to gain new rewards even if your skill level has not increased so you are no longer gaining many points in the current season.
MMR does not need to be reset with a new season, so that all games from the very beginning can be against evenly matched opponents. The ladder itself can still be reset each month. This way it is easy to distribute certain monthly activity rewards, such as the monthly card back.
What about spots in the preliminaries? If the HCT point structure were to remain, HCT points could be distributed based on ladder points, which would reward both activity and skill. A tough top 100 cut would no longer be necessary, as there would be more room for players in the preliminaries, so the points could be distributed to a larger number of players. This would retain ladder as a path to the preliminaries, but also move the real competition to where it should be, a Swiss format tournament, instead of a grind with a significant luck component.
(Alternatively, we could ask the question could the preliminaries simply be an open event, but I guess having some competitive incentives to play ladder is good for the game as it ensures that even top players spend at least some time playing the ladder instead of just private practice.)
Leagues or ranks could still remain, probably based on MMR, to give another form of progression for the players. There could even be a Legend league at the top, but given the huge number of players Hearthstone has compared to Starcraft 2, a limited 200-player Grandmaster League probably would not work. The Legend card back could be awarded after reaching a set MMR value instead of based on ladder points, and thus be a symbol of skill – if such a change were to take place, the old card back would probably be retired though and a new card back introduced for this purpose.
What about ladder competition?
If we go ahead with changes like this, what does it do to ladder competition? Will we lose all the intense fighting over who is the #1? To an extent, yes. However, there is not that much fighting over the #1 spot even now. In practice, players making a serious push for #1 have already secured a top 100 finish on another region or have enough HCT points regardless of their final rank, as otherwise making that push is too risky. OK, there’s Fibonacci, but that’s about it. The ladder competition is already a mix between grinding and camping, and that is hardly a healthy example to set.
For competition at the top, it is enough to publish the ladder points and the pure MMR list of the top players. What is the highest MMR score ever reached, and by whom? Who are the top 100 players by highest peak MMR score? These would be statistics that give top players bragging rights and a reason to play, even if they could secure a spot in the preliminaries with medium effort by just playing enough to spend all their bonus pool.
What about open cups?
There is yet another aspect of the Hearthstone esports ecosystem that is affected if the ladder top 100 competition is made less relevant: open cups. As it is, many players work for their points in open cups in order to alleviate the unpredictability of points gained from ladder. If there are more spots in the preliminaries and gaining points from the ladder is easier, where does this leave all the open cups? Will these grassroots competitions become irrelevant as players no longer need to play in them to qualify for the preliminaries?
I am not too concerned about open cups for now, as tournaments are always tournaments. Players will still want to practice by playing in actual low-stakes tournaments, and just compete. There are also other ways for Blizzard to support open cups than HCT points, but that is a whole another topic.
By changing the way preliminaries work, Blizzard can open up design space to alleviate the issues with the current Hearthstone ladder system. There are many things in the way the Starcraft 2 ladder works that can help shape the Hearthstone ladder into a more satisfying overall experience that rewards playing instead of camping and is able to distribute rewards fairly.