Dota 2 Ranked Season History: Every Reset Since 2017

Before the Season System: The Old MMR Era (2013-2017)

Dota 2 introduced ranked matchmaking in November 2015, giving players a visible MMR number that went up and down based on wins and losses. Before that, matchmaking existed but player ratings were hidden. The raw MMR era — a single number on your profile with no medals, no seasons, and no resets — ran from 2015 until late 2017.

The system was brutally simple during this period. Win a ranked game and gain approximately 25 MMR. Lose one and drop 25. Solo MMR and party MMR were tracked separately, and some players gamed this by heavily grinding party games with friends to inflate their party rating while keeping their solo number lower and more accurate.

The system had no floor protection, no ceiling guards, and no seasonal pressure. If you hit 6,000 solo MMR in 2016, you kept that number indefinitely unless you played ranked games and lost. Many players stopped playing ranked once they reached a peak they were happy with, effectively locking in their number forever. This created a significant stagnant leaderboard problem: inactive accounts clogged the top of the MMR charts for months or years.

Valve’s answer to this problem arrived with one of the most significant patch updates in Dota 2 history.

Season 1: November 2017 — The Medal System Arrives

Patch 7.07, released on November 1, 2017, fundamentally changed how Dota 2 tracked and displayed player skill. The introduction of seasonal rank medals replaced the bare MMR number on player profiles with a visible badge system. This was the first Dota 2 ranked season, and it reset every active ranked player back to a calibration state for the first time in the game’s ranked history.

What changed in Season 1

• Seven medal tiers introduced: Herald, Guardian, Crusader, Archon, Legend, Ancient, and Divine. Each tier had five stars, creating 35 possible visible rank levels plus an unranked state below them all.
• 10 calibration games required: Players had to complete 10 placement matches before receiving their new season medal. The system used prior MMR history to anchor calibration, so you were not starting from zero — but the outcome could still surprise players whose performance had shifted significantly.
• Solo and party MMR merged in display: Prior to Season 1, solo and party MMR were tracked and shown separately. Season 1 introduced a unified medal display, though the underlying rating system continued to track performance in both contexts.
• Phone number verification required: Valve introduced a requirement that a unique phone number be linked to each account before it could participate in ranked matchmaking. This was designed to reduce smurf account creation and improve queue quality across all brackets.

The MMR thresholds for Season 1 medals were never officially confirmed by Valve, but community consensus at the time placed Herald at the lowest bracket (approximately 0-700 MMR) and Divine at the top tier above roughly 4,000-4,500 MMR. These thresholds shifted in later seasons as player distribution changed and new tiers were added. Valve has never published exact bracket boundaries, which remains a persistent source of community debate. For current tier breakdowns, see our Dota 2 boosting services guide.

Season 2: Approximately May 2018 — First Full Recalibration

The second ranked season began approximately in May 2018, though Valve did not always announce season transitions with precise dates in advance. Season 2 was the first time the community experienced the now-familiar ritual: the countdown to reset, the 10 calibration games, and the uncertainty about where you would land versus where you were before.

What changed in Season 2

• Recalibration adjustment system: Valve introduced adjustments that could shift a player’s MMR by more than the standard per-game amount during calibration games. The goal was to correct placements that Season 1 had gotten wrong — players who had been placed too high or too low due to the MMR merge could now shift more quickly to an accurate bracket.
• Behavior score priority queue: The connection between behavior score and ranked queue priority became more explicit. Players with lower behavior scores faced meaningfully longer wait times, particularly during peak hours, making the behavior system feel more consequential for day-to-day play.
• Post-season audit warning: Valve published a community blog post approximately in late 2018 noting that their systems had detected widespread manipulation of calibration matches and warned that accounts found to have exploited the calibration period would face penalties. This set a precedent for post-season enforcement.

Season 3: Approximately November 2018 — Immortal Rank Added

Patch 7.20, released approximately November 19, 2018, introduced the Immortal rank, which became arguably the most structurally significant addition to the medal system since Season 1. Where previous seasons had Divine as the ceiling, Season 3 added a new pinnacle tier reserved for the top players in each region.

What changed in Season 3

• Immortal rank introduced: The top 5,000 players by solo MMR in each region received an Immortal medal. Unlike other medals, Immortal displayed a leaderboard position number rather than a star count, meaning rank 1 Immortal and rank 4,999 Immortal were clearly distinguished in a way that the old Divine star system could not accommodate.
• Divine restructured: The introduction of Immortal pushed some players who had previously been considered the peak of competitive Dota 2 into a tier that was now explicitly second-best. Divine became the rank just below Immortal, and the community quickly understood that Divine 5 and Immortal 1 were separated by a leaderboard threshold rather than a fixed MMR number.
• Regional leaderboards made public in-client: Valve introduced leaderboards visible inside the Dota 2 client showing the top Immortal players sorted by MMR within each region. This created a new public competitive layer and gave high-MMR players an additional reason to grind beyond just hitting a medal level.
• Calibration games remained at 10: The recalibration mechanic did not change structurally for Season 3, but the placement algorithm continued to evolve behind the scenes. Players at the high end of Divine entering Season 3 calibration were the first to discover whether they would place as Immortal or just miss the leaderboard threshold.

Season 4: Approximately May 2019 — Role Queue Experiments Begin

Season 4, beginning approximately May 2019, coincided with Valve’s first serious experiments with a role selection system for ranked matchmaking. For the first time, players in certain regions were asked to select a position (1 through 5) before entering a ranked game, and their performance in that role would influence their rank separately from games where they picked any position freely.

What changed in Season 4

• Role queue beta testing: Valve rolled out an optional role selection system in select regions as a structured test. Not all players had immediate access, and the behavior of MMR changes in role queue versus standard ranked varied by region and phase of the beta. Players who opted in often found queue times longer but game quality higher within their selected position.
• Behavior score became a hard gate: Players who fell below a certain behavior score threshold were removed from ranked matchmaking entirely until they raised their score through unranked play. Valve publicized this as an anti-toxicity measure, and it was the first time behavior score functioned as a complete blocker rather than just affecting queue priority.
• 10 calibration games continued: Season 4 recalibration kept the 10-game format. The algorithm placed some additional weight on role performance for players who had participated in the role queue experiments, giving early adopters of role queue a slight calibration advantage if they had performed consistently in their stated positions.

Season 5: Approximately November 2019 — Role Queue Goes Global

Season 5, launching approximately November 2019, marked the point at which role queue stopped being an experiment and became the default ranked experience across all regions. This was the most significant change to how Dota 2 ranked games actually played out on a game-by-game basis since Season 1 introduced medals.

What changed in Season 5

• Role queue mandatory for ranked: All ranked matches now required players to select their intended position before queuing. The matchmaking system used these selections to build teams where each position had a declared player, dramatically reducing the in-game argument about who plays what role that had plagued unstructured ranked queues for years.
• Role performance badges: Players who consistently performed well in their selected role received performance-based badges visible to teammates before the game started. This affected the social dynamics of ranked games and gave high-performance players a visible credential beyond just their medal.
• Recalibration influenced by role performance: Season 5 calibration games were played in specific roles, meaning your placement was influenced by how well you had performed in the roles you selected during calibration. This was the first explicit link between role-specific performance and calibration outcome, a precursor to the fully split core/support MMR system that followed in Season 6.

Season 6: Approximately May 2020 — Core/Support MMR Split

One of the most consequential structural changes in Dota 2 ranked history arrived approximately in mid-2020: the introduction of separate Core and Support MMR. Instead of a single MMR number covering all five positions, players now had two distinct ratings — one for playing as core heroes in positions 1, 2, and 3, and one for playing support in positions 4 and 5.

What changed in Season 6

• Two MMR ratings per account: Core MMR tracked performance as a position 1/2/3 player. Support MMR tracked performance as a position 4/5 player. Both numbers were visible on your profile, and your displayed medal reflected whichever was higher.
• Separate calibration per role type: Players who wanted both ratings properly set needed to calibrate separately for core and support. A player with 4,000 core MMR might have 3,200 support MMR, and the two no longer pulled each other up or down as they had under the unified system.
• 10 calibration games per role type: The core/support split effectively doubled the calibration workload for players who cared about both ratings. Most players chose to focus on calibrating their primary role type and let the secondary role sit at whatever it settled at through regular play.
• Medal reflects the higher of the two: Your visible badge showed based on your higher MMR between core and support. This created some visible discrepancies — a player could display an Ancient medal while having Guardian-level support MMR. This was by design: Valve wanted to reward specialization rather than force players to maintain high ratings in roles they rarely play.

The core/support split was broadly welcomed by the community as a long-overdue recognition that carrying and supporting are genuinely different skill sets with different expertise requirements. A player who had reached 4,000 MMR as a carry main but had only 2,500 MMR worth of support knowledge was no longer in a position to accidentally harm support-specific matchmaking quality when filling.

This change also created the calibration strategy that persists to this day: many players deliberately enter recalibration windows focused on their stronger role type to maximize the quality of their placement. If your support MMR was lower than your core after Season 5, Season 6’s recalibration was an opportunity to establish a more honest support baseline. The connection between calibration windows and rank climbing strategy is why many players consult our guide to safe MMR boosting before major calibration periods.

Season 7: Approximately November 2020 — Pandemic-Era Adjustments

Season 7, beginning approximately November 2020, took place during a period of unusually high player counts driven partly by pandemic-related lockdowns in many regions. Valve’s matchmaking system was handling significantly more concurrent players than typical for that time of year, which affected queue dynamics, pool quality, and MMR distribution across brackets.

What changed in Season 7

• High player volume effects: With more players in the pool, queue times shortened but bracket distributions temporarily shifted. The influx of returning and new players created imbalances in lower brackets, particularly Herald and Guardian, where matchmaking quality fluctuated more than in previous seasons.
• Anti-smurf detection improvements: Valve indicated in community communications that their systems had improved at identifying accounts playing significantly below their true MMR level. Some accounts began receiving automatic rank corrections mid-season rather than waiting for the calibration window, which was a new behavior from Valve’s enforcement side.
• Behavior score floors reinforced: Players with very low behavior scores continued to face ranked queue restrictions. Valve’s in-client communications about this became more explicit during Season 7, with clearer messaging about what behavior score threshold was required for ranked access and how to raise it.

Season 8: Approximately May 2021 — Calibration Accuracy Refined

Season 8, arriving approximately May 2021, represented a period of relative structural stability for the ranked system. The core/support split from Season 6 had settled into place, role queue was standard practice, and Valve’s development focus shifted toward refining calibration accuracy rather than introducing new structural overhauls.

What changed in Season 8

• Calibration variance adjustments: Valve refined how much a player’s MMR could move during calibration games, trying to reduce the frustrating experience of very wide swings that landed players far from their expected bracket after a streak of bad or good calibration luck. The goal was calibration that was both quick and accurate rather than just quick.
• Leaderboard threshold adjustments: The MMR required to reach the Immortal leaderboard ranking changed approximately with patch 7.29 (April 2021) as player distribution in the upper brackets had shifted significantly since Season 3 first introduced the Immortal tier. Valve made these threshold changes in response to community feedback that the top-5000 cutoff was no longer representing the same relative skill level it had in 2018.
• Boosting detection improvements: Valve introduced detection improvements aimed at identifying accounts that were being played by someone other than the account owner. This affected recalibration patterns for some accounts and prompted discussions in the boosting community about operational security practices. TeamSmurf’s approach — exclusive use of VPNs matching the client’s region and strict role-limited play — reflects these requirements.

Season 9: Approximately November 2021 — Behavior Integration Tightened

Season 9, approximately November 2021, continued the stable structure established in recent seasons. By late 2021, the core/support split was well established, role queue was standard practice, and the community had largely adapted to the approximate six-month rhythm of recalibration. The main community complaints centered on calibration accuracy — a perennial concern that Valve continued to address with ongoing algorithm adjustments.

What changed in Season 9

• Calibration game weighting adjustments: Valve made internal adjustments to how heavily individual calibration games weighted the final placement calculation. Win streaks and loss streaks during calibration had more predictable outcomes, reducing some of the variance that had left players feeling that calibration results were arbitrary.
• Behavior score integration tightened further: Players with accumulated low behavior scores found it harder to access calibration games or found their placement outcomes penalized in matchmaking quality. The signal that behavior score sent to the ranking system became more influential over time, not less.
• Community pressure on MMR transparency: Season 9 saw renewed community debate about why Valve did not publish exact MMR thresholds for each medal tier. Players attempting to gauge how close they were to a star promotion had to rely on community data from sites like Dotabuff and Stratz rather than official information. Valve’s silence on this topic continues through 2026.

Season 10: Approximately May 2022 — The Transition Season

Season 10, approximately mid-2022, was notable as a transition point for the broader ranked system structure. Valve had begun signaling in community communications that the strict bi-annual season format might evolve into something more dynamic, and the Season 10 reset was accompanied by quieter changes to how the system functioned under the surface.

What changed in Season 10

• Soft optional recalibration for stable accounts: Valve began allowing accounts with highly stable recent performance history to opt into a lighter recalibration experience rather than the mandatory 10-game calibration window. This was a small change in practice but a significant philosophical shift — the first acknowledgment that the hard calibration mechanic was not serving all players equally well.
• MMR decay discussion: Valve publicly discussed (without implementing at this stage) an MMR decay system that would gradually reduce inactive accounts’ visible MMR over time, addressing the stale leaderboard problem that had existed since ranked matchmaking launched in 2015. The community reacted with mixed feelings: decay would make leaderboards more accurate but would penalize returning players.
• Regional Immortal threshold variation: By Season 10, the MMR required to reach Immortal leaderboard ranking varied significantly by region. Immortal in some lower-population regions required less MMR than in Southeast Asia or EU servers where competition at the top was denser. Valve acknowledged this asymmetry without committing to regional standardization.

2023-2026: The Shift to Continuous Ranking

Beginning approximately in 2023, Valve moved away from the strict six-month season structure. Instead of hard resets every approximately May and November, the ranked system transitioned toward a more continuous model where medals updated more gradually based on rolling recent performance rather than resetting to a calibration state at fixed intervals.

This change reflected several years of accumulated community feedback that hard resets were both exploited (creating windows where skilled players could deliberately deflate to recalibrate higher) and frustrating (players who had ground out a rank over months could see it destabilized by a single bad 10-game calibration run).

What the continuous model changed

• No hard reset dates: Players no longer need to play 10 mandatory calibration games at the start of a new season. Instead, rank adjusts continuously, with larger adjustments possible after periods of inactivity, significant recent performance changes, or during designated soft recalibration windows.
• Medals update more frequently: Medals can update on a weekly cadence or after defined game-count thresholds rather than only at season boundaries. This makes rank feel more responsive to actual recent performance rather than the single snapshot of a calibration period.
• Soft recalibration windows retained: Valve kept the concept of heightened calibration periods tied roughly to annual schedule points, including the September window that follows The International. These are now softer gates — players can participate in the heightened-adjustment window or opt to let their rank update normally. The choice matters: entering a soft window with strong recent momentum produces better outcomes than entering it cold.
• Behavior score fully integrated: By 2024-2025, behavior score had become deeply integrated with rank stability. Players with high behavior scores experience smaller downward rank adjustments on losses and faster recovery after inactive periods. Players with degraded behavior scores see rank instability even during winning streaks.

For players looking to maximize their rank during the September 2026 soft recalibration window — historically one of the most active periods for rank movement in the Dota 2 calendar — entering the window at a strong recent MMR position is more important than ever under the continuous model. Working with a professional service to climb before the window opens remains effective. Our review of the best Dota 2 boosting services covers every reputable option in detail.

Full Dota 2 Ranked Season History Table

The table below summarizes every major ranked season in Dota 2 history. Dates marked “approximately” reflect community documentation and patch release windows rather than official Valve announcements, which were rarely given with specific calendar dates.

Patch numbers marked with ~ are approximate, based on the patch release dates that most closely match the described season transitions. Valve did not always announce season changes in the same update as the most associated patch.

How Each Reset Actually Works

Understanding the mechanics of a recalibration period — whether a hard season reset under the old model or a soft window under the current continuous model — helps you approach it as a strategic opportunity rather than an anxiety event.

Your MMR does not go to zero

The biggest persistent misconception about ranked resets in Dota 2 is that you start fresh with no history. You never did. Your previous MMR has always anchored your calibration. The system uses your calibration games to confirm your bracket, adjust you slightly if recent performance has shifted, or correct significantly if you clearly belong somewhere different from where you finished last season.

Calibration games carry more MMR weight

During active calibration periods — both under the old season model and during the continuous model’s soft windows — wins and losses carry more MMR weight than standard ranked games. The exact multiplier varies and Valve has never published the formula, but community analysis consistently shows calibration games producing larger positive or negative MMR swings than regular ranked games. This cuts both ways: a strong calibration run can lift you faster than normal grinding, while a bad run can drop you further than typical loss streaks.

Behavior score matters more during calibration

Valve has indicated clearly that behavior score factors into calibration outcomes and matchmaking quality during calibration windows. A player entering calibration with a high behavior score (8,000-10,000) will consistently calibrate closer to or above their previous rank, while someone with a degraded behavior score may find their placement penalized even at a 50/50 win rate. Behavior score is not a minor factor — it is a meaningful input into where you land.

Role-specific calibration matters post-Season 6

Since the core/support split arrived in Season 6, calibration is meaningful per role type. Playing calibration games as a position 1 carry sets your core MMR. Playing in position 4 or 5 sets your support MMR. Most players with a clear primary role focus their calibration games entirely on that role type to produce the most consistent and positive placement outcome. Mixing roles during calibration typically results in a blended adjustment that does not fully reflect your best performance context.

Preparing for the September 2026 Calibration Window

September has historically been one of the most strategically important months in Dota 2 ranked matchmaking. It falls after The International — when the professional Dota 2 season winds down and pub player activity surges as players who took a break during TI return to ranked. Many Immortal and Divine players use September as their primary climbing window of the year.

Here is how to position yourself for the September 2026 calibration window:

• Raise your behavior score before the window opens. Play support, avoid intentional feeding, use the commend system actively. Entering any calibration window with 9,000+ behavior score gives you the best possible placement mechanics. Raising behavior score takes time — start at least two weeks before the window.
• Enter at your highest recent MMR. Under the continuous model, calibration anchors to your current rating. Entering the September window after climbing to a new high rather than after a recent losing streak is the single biggest controllable variable in your calibration outcome. This is why many players use boosting services in the weeks before a recalibration window — the entry point determines the outcome range.
• Focus exclusively on your strongest role type. Do not split calibration games across core and support if you have a clear primary role. Concentrated calibration games in your best role produces the most reliable positive outcome. Diluting across roles produces an averaged result that undersells your actual ceiling.
• Win your first two calibration games. Early momentum in a calibration run has an outsized effect on the system’s confidence about your rating, particularly under the rolling-adjustment model. Treating game one and two as the most important games of the month is not superstition — it reflects how the placement algorithm weights early signals.
• Avoid tilting off the calibration window. If you lose two in a row during calibration, the worst response is to keep playing in the same emotional state. Take a break, reset, and come back fresh. The window is not a deadline that closes in hours — use it deliberately rather than reactively.

If you want to enter the September 2026 window at a stronger MMR position than you have earned through regular play, TeamSmurf’s professional boosters work exclusively in Dota 2. No outsourcing, no account sharing, no hero restrictions. See our Dota 2 MMR boost page for current pricing and availability.

Frequently Asked Questions