TL;DR: Deadlock's matchmaking actively produces the smurfing it claims to fight. Rank freezes at Initiate 2 are documented and reproducible. Hero pool changes penalize honest players the same way account rentals do. The MVP system rewards personal stats over winning. New accounts outrank mains with ten times the playtime.
I have been playing Valve games since childhood. Ten years across the Counter-Strike series from CS:S to CS2, five years in Dota 2. When I moved to Deadlock, I was hoping to see Valve apply the lessons of over a decade. They have not. The planned matchmaking system carries the same structural problems from both games, now combined and amplified. Multiplayer Valve titles have suffered from three interconnected issues for 14 years: smurf accounts, account boosting, and account rentals. Rather than solving these before Deadlock’s launch, Valve layered a “performance-based” system on top of the same broken foundation; and whether by design or by negligence, the outcome is entirely predictable: poor match quality, rank volatility that shrinks with account age, a saturated black market for smurf accounts, systemically manufactured demand for those very accounts, and third-party service advertisements embedded in leaderboard nicknames.
What does a working matchmaking system look like?
Look up where donk, s1mple, and m0NESY play today. Under a binary Elo system, the only path to success is winning more than you lose against players of your own rating, over hundreds of matches.Keep that in mind as we examine what Valve is actually building.
Dota 2: Thirteen Years of Uncontrolled Precedent
Dota 2 is the most thoroughly documented case of what happens when a matchmaking system cannot close its own exploits. A shadow economy has formed around it with significant turnover. At high MMR, account rentals are so normalized that players in the top 1,000 know every rental account by name and leave lobbies at the hero draft stage when they spot one. Valve is fully aware of this and has done nothing about it in Dota 2. Their proposed solution in Deadlock is binding a hero pool to each account so the system detects when an account changes hands, penalizes honest players: anyone testing a new hero sees their rank drop immediately, which pushes them into lower lobbies and makes creating a smurf the rational choice. The mechanism meant to deter account sharing actively incentivizes it.
*Obscurus is Valve's name for the rank below the bottom of the ladder
Support players are the primary victims of any performance-based system. Their contribution cannot be captured in raw numbers; their role exists to enable others. This creates pressure to purchase boosted accounts, since climbing independently requires playing as a semi-carry rather than as a support. In Deadlock, where the system rewards individual statistics, this problem will be felt in its worst possible form. A player documented on the forum, a support main who deliberately keeps a low souls-per-minute to stay active across the map, lost rank after wins precisely because the algorithm read his numbers as underperformance.
The MVP system makes this structurally worse. The performance layer does not evaluate whether you won - it evaluates whether you individually maximized the right numbers on the right hero. This creates a perverse incentive: playing to win and playing to rank up are not the same objective. The system has been observed distributing post-game awards in ways that confirm this: a team that spent most of the match losing but secured a final teamfight victory can receive one MVP designation, while two players from the losing team are judged to have produced better individual statistics and receive key-player recognition.
*Match 78596946 pulled from the top of the EU leaderboard documents exactly this outcome
Valve appears to have introduced this asymmetry as an attempted fix for a real problem: you play well, your teammates feed, you lose, and the system punishes you for it. Their solution was to decouple individual recognition from match outcome by awarding key-player status based on personal statistics regardless of which team won. The intention is understandable. The execution is not.
The problem is that some losses cannot be prevented by individual play, no matter how well you perform. If an enemy carry goes 20/0 and has outfarmed your entire team for twenty minutes, the game is structurally over. No amount of personal skill on your part changes the math of a fed hero with a massive net worth advantage walking through your base. You cannot stat-sheet your way out of a structural deficit, and this is because the system has no way to distinguish between a loss caused by your underperformance and a loss caused by a situation that was already decided before you could do anything about it.
So the patch addresses the symptom which is a good player losing rank they didn't deserve to lose without addressing the cause, which is that feeders and smurfs create unwinnable games in the first place. And in doing so, it introduces a new problem: the correct way to play for rank is no longer the same as the correct way to play to win. A player who sacrifices their own statistics to enable a teammate by dying to buy time, rotating instead of farming, or playing the support role in a critical moment, will find the algorithm reading those decisions as underperformance, even if they directly contributed to a victory. The player who farms safely and posts clean numbers while their team loses will sometimes be rewarded more than the player who tried to win and came up short. Rank and victory have become two different objectives, and the system has quietly chosen which one it actually measures.
Counter-Strike 2: The Precedent Already Set
CS2’s Casual map selection allows a Global Elite player to enter an uncalibrated Office lobby and dismantle beginners without any consequence to their rank. Deadlock’s Hero MMR system creates an equivalent mechanic: a player learning a new hero de facto lowers their rank and enters lobbies below their real skill level.The HvH (hacker-versus-hacker) pool, a segregated lobby for detected cheaters, reveals Valve’s priorities clearly. Instead of banning cheaters, the system isolates them to preserve active-player metrics. Given that Deadlock is partly a shooter, an equivalent solution there is more than likely.
CS2 leaderboards contain accounts with third-party service advertisements embedded directly in their nicknames, which are visible evidence that reaching top rank through cheating and boosting is operationally easier than getting banned.
Under a performance-based system in Deadlock, the entry threshold for the leaderboard is even lower, which means this advertising will appear even faster.
Prime matchmaking completes the picture. Valve monetized a problem they created. Fair matchmaking became a paid feature rather than a baseline condition. The precedent exists, and there is no reason to believe Deadlock will not follow the same model.
Deadlock: All of the Above, Simultaneously
Every problem described above is present in Deadlock at once, and the matchmaking system actively amplifies them.Hero MMR working against honest players. A player ranked in the top 70 of the European leaderboard lost more than 200 positions after several wins. The only change: shifting his primary hero from Dynamo to Paige. Deadlock's Hero MMR is not a hidden inference because it is a visible in-game system, displayed directly in the UI as downward indicators on heroes outside a player's established pool, and documented to place those heroes into easier lobbies. What remains opaque is how far the system extends: the forum case above suggests that a sustained shift in hero pool composition affects not just lobby difficulty on that specific hero, but overall leaderboard position that affects even when the account's win balance is positive. An honest player rotating his main is penalized through the same mechanism as an account changing hands.
Rank freezing is documented and systemic. Forum records show players with 150+ hours and a positive win rate whose rank does not move up or down; it is literally frozen at Initiate 2. Their friends with half the games played and a negative win rate freely climb to Ritualist and Arcanist. One forum user described creating a new account to test the theory: he reached Ascendant 4 in roughly 100 games, while his main account remained frozen at Initiate 2 across hundreds of matches.
The result is reproducible. Another user’s first account holds 600 matches and barely crosses 50% win rate after hundreds of hero experiments. His new account reached Emissary 3 in 16 games, 12 of which were wins. This is not an outlier; it is a pattern reproduced independently by multiple forum users.
The conclusion is already underway. Players are not complaining about smurfing as a moral choice; they describe it as the only rational exit from a system that punishes loyalty to a main account.
The system does not merely fail to fight smurfing. It produces smurfing.
The Initiate Paradox: Empty and Overcrowded at Once
Rank distribution data confirms this statistically. According to deadlocklab.com (last 30 days, 314,937 tracked players), Initiate accounts for just 3.68% of active players - despite being the starting rank for every new account. For context, the Faceit CS2 EU ladder (5,178,584 players, faceitanalyser.com) is the closest available reference point for a matchmaking system that functions predictably: Level 1 holds 2.56% of players, rising steadily to a peak at Level 4 (26.04%), then tapering off toward Level 10 (6.58%). Faceit is not a perfect parallel - it requires identity verification when smurf activity is suspected, and its player base skews toward competitive-minded users rather than the general population. But those same properties make its distribution more meaningful, not less: the bell curve holds even without casual players to inflate the lower ranks, which means the shape reflects genuine skill spread rather than an artifact of player demographics. Low ranks are organically populated because new players arrive, calibrate gradually, and the system does not eject them. Crucially, the system is predictable: a win moves you up, a loss moves you down, and the relationship between performance and rank position is legible to the player. In Deadlock, the bottom of the ladder is statistically near-empty for three simultaneous reasons: smurf accounts sprint through Initiate in 10–20 matches and vanish from the statistics; newcomers either drop the game or get stuck; and those who do get stuck generate a disproportionate number of matches (6.92% of all ranked matches at just 3.68% of players, per the same data).
The most telling figure: Initiate 2 is flagged by deadlocklab.com as the highest-grind subrank on the entire ladder, with an activity index of x2.42, implying that players there generate matches at more than twice the site-wide average pace. That subrank maps precisely onto the forum reports of players frozen at exactly that level.
Sources: deadlocklab.com/ranks (last 30 days, updated May 3, 2026) | faceitanalyser.com/rank-distribution/cs2/EU/
This is why Initiate is statistically empty and simultaneously full of complaints. It is not a contradiction, but two different phenomena of the same system. Most accounts sprint through Initiate in 20 matches via fast calibration and are statistically invisible. Those who get stuck are stuck permanently, because the algorithm is already confident in their assessed level and stops moving their rating. There is no middle ground.
Players frozen at Initiate face a compounding problem beyond the rank ceiling itself. Because the rank is used as a calibration floor, their lobbies are disproportionately filled with two types of opponents: fresh smurf accounts completing calibration, and deliberate feeders who are players who intentionally throw games, knowing they can register a new account at any moment if their loss streak goes too far. For an honest player trying to climb, this means winning a fair match is structurally improbable: the smurfs on the opposing team outclass the lobby by design, and the feeders on their own team actively work against them.
The system places its most vulnerable honest players into its most adversarial environment.
This creates a structural trap for any honest player who learns on their main account. Learning inevitably means losses in the early stages - each of which is cemented into the rating as the system builds confidence in the player’s level. By the time skill has genuinely improved, the algorithm has already classified the account with hundreds of matches as evidence. Escaping without an improbably long unbeaten streak is mathematically close to impossible.
A new account solves this instantly. The same skills, maximum volatility, fast upward calibration in 20 games. The system has not become fairer; only the account changed. Deadlock does not merely tolerate smurfing as a workaround. It makes smurfing the only rational answer to its own mechanics for any player who dared to learn the game honestly.
The Fix Does Not Require a New Algorithm
1. Binary Elo, one queue, no black box.
Win: plus. Loss: minus. No individual statistics, no Hero MMR, no hidden performance layer. The support player who gave their life to enable a teammate objective gets the same plus as the carry with a triple kill - and crucially, can take risks without fearing the algorithm will mark them as a liability based on damage and farm numbers.2. Elo substantially raises the cost of boosting and that is the key distinction.
In Dota 2 under the current system, a booster running a handful of cheese heroes can lift an account to 8,000–9,000 MMR in a reasonable timeframe and sell it for around €100 at low cost of production is a high-volume market with strong margins. Under Elo, a booster must consistently win across hundreds of matches. The system does not eliminate the market, but it makes production expensive, which naturally limits supply and the scale of the problem. Under Deadlock’s performance-based system, the cost of boosting falls even lower than in Dota 2: it is enough to show the right numbers on the right hero across 20 matches. The smurf account market gains its lowest barrier to entry yet across all three Valve titles.3. Limit MMR spread within a party instead of penalizing individual performance.
The one genuine problem that a performance-based layer addresses is boosting within a premade group, since individual performance is harder to inflate when playing alongside a booster. The identical problem can be solved with a simple MMR spread restriction within parties; without any of the side effects described in this post.4. New accounts should start at median MMR with symmetric volatility; not at the floor with downward-sticky volatility.
The current design assigns new accounts the minimum possible MMR and then applies high volatility for the first matches, creating a period where losses are catastrophically sticky. The floor placement is the root cause of the Initiate environment being uniquely hostile: smurf accounts calibrating downward and deliberate feeders both concentrate at the bottom of the ladder, and the system deposits every new player directly into that environment. The correct approach is the inverse: start every new account at median MMR with a moderate volatility window of roughly 20 matches, then let the rating settle naturally from a position that reflects the realistic skill distribution of someone new to ranked play. A newcomer who genuinely belongs at Initiate will calibrate there within 20 games, but they will do so from above, passing through a fair skill band on the way down rather than being trapped in the most adversarial lobby in the game from match one. This single change would eliminate the structural smurf advantage at calibration, reduce the concentration of feeders at the floor, and give honest learners a recovery path without requiring a new account.5. Volatility must remain meaningful regardless of account age.
This is the only way to give an honest player a path out of a rank acquired during their learning period. If the system needs to be rebuilt incrementally, the minimum viable first step is decoupling Hero MMR from rank volatility, and stopping the reduction of that volatility after N matches.Valve has already seen this system in action. It is the one the best players in the world use in CS2 on a third-party platform instead of the official ranked mode. That is not coincidence and not advertising. These are players who earn their living from this game and they have collectively chosen to do so somewhere else.
Last edited:
