It’s 9 PM on a Friday and there are five names in the Discord. Somebody says “queue up,” and everyone quietly does the math on who’s about to spend forty minutes dying in a doorway.
There are about twelve of us on the roster. A normal night is three to six, depending on who’s got kids asleep and who’s still at work. Every few months we get eight of us into a custom lobby and it feels like a holiday. We rotate through Overwatch, Marvel Rivals, and Apex Legends depending on the week.
The problem is that our skill spread is a canyon. Silver at one end. Grandmaster at the other. Everything in between.
For a long time I thought that was a people problem. Turns out it’s a systems problem, and once I understood how the systems work, the whole thing got fixable.
The Night I Stopped Blaming My Friends
There’s a guy in our group, he’s been Silver for two years and he’s fine with that. Great hang. Terrible aim. He shows up every week and he loses every gunfight he takes.
For months I did the thing you probably do. I gave him tips. I sent him VOD timestamps. I told him to lower his sensitivity, then raise it, then lower it again. I put him on characters that “don’t need aim.” He tried all of it. He kept getting deleted three seconds into every fight.
And I kept getting quietly annoyed, which is the part I’m not proud of. (also, this is what shouting pillows are for)
Then one night he said something that reframed the whole thing for me. He’d gone and played a bunch of games solo that week without us. And he said, “It’s weird. When I play alone I actually get kills.”
That’s when it clicked. He wasn’t playing Silver lobbies with us. He was playing my lobbies. I think I intuitively knew this, but it was so easy to forget.
The Matchmaker Is Built to Do This to You
Here’s the part nobody puts on the loading screen.
Respawn has said out loud how Apex Legends builds a match. Their own developer blog says the algorithm “uses HIGHEST for Pubs and Ranked,” and then adds, in the same breath, that this “makes it very difficult for new players to join in with their more skillful friends, as the match is made based on the more skilled player.”
Read that again. The lobby is built around your best player. Not the average of your squad. The ceiling.
And there’s a second layer stacked on top. In unranked battle royale, Respawn says they “increase the effective skill value of the squad, to slightly counter the inherent bonus we believe premade squads have.” So you get penalized for the highest-skilled member, then penalized again for being a premade at all.
Blizzard doesn’t publish a formula for Overwatch, and the matchmaker’s own listed inputs include group size, group wideness, and something called grouping advantage. They’ve conceded the effect directly, saying that grouping across a big rating gap “is a major source of wide skill disparities in matches.”
NetEase publishes nothing at all about how party rating works in Marvel Rivals. They just cap your party size instead, and the cap gets smaller the higher you climb.
The 50% That Isn’t Yours
Riot has said publicly that their matchmaker targets an expected win rate of 50 plus or minus 1 percent. Blizzard targets roughly the same, with most matches landing between 45 and 55 percent.
Sounds fair. Sounds like exactly what you’d want.
But that 50 percent belongs to the team. Not to you. Not to your Silver.
So think about what has to happen for a squad with a Grandmaster and a Silver in it to sit at a coin flip. The Grandmaster has to overperform against opponents calibrated to their level. And the Silver has to underperform against those exact same opponents, every single round.
The group wins half its games. Your worst player has a miserable night in all of them. The wins included.
That’s the whole thing. That’s the entire problem in one paragraph, and it isn’t a character flaw. It’s arithmetic.
Stacking Is a Buff, and You’re Paying Tax on It
There’s a Microsoft Research paper on skill rating that measured something wild. Solo players won about 47 percent of their matches. Parties of ten or more won 89 percent. A basic sum-of-ratings model predicted those big parties should only win 67 percent.
So playing as a coordinated group is worth roughly 22 points of win rate that nobody’s individual rank captures. Voice comms are cheating, basically, and we all know it.
Modern matchmakers know it too, and they correct for it. Which means your group is getting a difficulty bump for being organized on top of the difficulty bump for having a Grandmaster in it.
Guess who absorbs that bump. It isn’t the Grandmaster.
Three Games, Three Different Walls
Before you plan anything, you need to know what the games will even let you do. I went through the current rules for all three of ours as of August 2026.
Marvel Rivals caps your ranked party by tier. Bronze through Gold can queue in any size except five, so a full six-stack is legal. Platinum, Diamond and Grandmaster get capped at three. Celestial and above get capped at two.
So a Silver and a Grandmaster in the same ranked Rivals match is off the table the second anyone in the group hits Platinum. But if your whole crew is Gold or below? Six-stack ranked, all night. That’s genuinely the best mixed-skill competitive setup in any of the three games.
Apex Legends requires everyone in a ranked squad to be within two ranks of each other. EA’s own help page says it plainly. Outside that window, matchmaking just fails. Respawn ran two experiments during Season 29, tightening premades to one tier and banning premade squads entirely at Diamond and above. Both tests ended in July and neither one carried into Season 30. They’ve said a results post is coming, so this could move again.
Overwatch will technically let you do it. Blizzard removed hard rank caps back in 2024 and replaced them with the Wide Group system. Any group containing a Grandmaster is flagged wide, no matter the spread. Wide groups only match against other wide groups, they queue longer, and your rank barely moves. One of the designers put it directly: “the wider your group is, the less your rank will change.”
There’s also a gotcha that will bite you specifically, and it’s my favorite piece of trivia in this whole thing. You cannot queue Competitive as a wide group of exactly four. Two, three, or five. Never four. Super cool.
Four is a really common night for us. So that one rule quietly kills about a third of our Fridays if we don’t plan around it. Grab a fifth or drop to three.
In practice the queue time kills it before the rules do. The pool of other wide groups at that spread is so thin that matches sometimes never form at all. There’s a widely-shared story of a Grandmaster who queued with a Bronze friend and never found a single game in 24 hours.
One caveat on all of this: Overwatch just added a new rank called Emerald between Platinum and Diamond in the August 11 patch, which shifts every tier above it. I somehow landed here on support after not having played ranked in about a month. Really wish they had put me in Gold.
Blizzard hasn’t said whether the wide-group thresholds got retuned to match. Worth checking in your own client before you plan a season around it.
So Stop Trying to Fix Your Silver
Here’s the reframe that changed our Friday nights.
You cannot coach your way out of this. Not in a reasonable timeframe. Your Silver friend would need to climb four full tiers to make ranked queueing viable, and that’s a year of dedicated practice for someone who plays six hours a week for fun.
Meanwhile, the actual cause of everyone’s bad time is a matchmaking rule you have zero control over.
So don’t fix the player. Fix the structure around the player.
Everything below is stuff we’ve either run ourselves or lifted from research and adapted. Some of it is going to sound like overkill for a group of friends playing video games. Do it anyway. It works.
Rule One: Two Losses and You Switch
This is the highest-value thing on this list and it costs nothing.
EA published research on a dataset of 36.9 million matches. Players who lost three in a row disengaged at a rate of 5.1 percent. Players with a mixed or winning recent record sat at 2.6 to 2.7 percent.
Three losses roughly doubles the chance someone quietly stops showing up.
There’s a second finding that matters even more for a mixed group. A separate study of about six million matches found that consecutive losses hit players differently depending on their level. A three-game skid is a rough evening for your Grandmaster. For your Silver it’s a referendum on whether they belong in the group at all.
So we made a rule. Two losses in a row and we change something. New mode, new teams, or we stop and play something dumb for twenty minutes. Nobody argues, nobody has to be the one who asks. The rule asks for us.
Valve figured this out from a completely different direction, by the way. The AI Director in Left 4 Dead forces a mandatory 30 to 45 second calm stretch after every intensity spike. It doesn’t buff or nerf anyone. It manages the rhythm of pressure. Same idea, smaller timescale.
Rule Two: The Best Player Shuts Up
Our Grandmaster used to call everything. Rotations, cooldowns, enemy positions, ult economy, all of it, in real time, in complete sentences.
He thought he was helping. He was actively making people worse.
A Grandmaster’s comms are tuned for listeners who can process four pieces of information and act on all of them inside a second and a half. A Silver can act on one. Feeding four to a player who can process one doesn’t give them three spare. It jams the whole intake.
The CS2 coaching material has the cleanest test I’ve found for this. Before you say anything, ask: what should my teammate do differently because I said this? If the answer is nothing, don’t say it. Ten words maximum during a fight. One caller, and everybody else stays quiet unless they have direct information.
We flipped it so the Grandmaster talks least. Sounds backwards. Our win rate went up.
Notes and criticism go in a VOD after the session, and only when the person asks for them. We made “do you want notes on that?” a literal thing people say out loud. Nobody gets coached who didn’t ask for it that night.
Rule Three: Give Everyone a Job They Cannot Fail At
Your weakest players need a role where their contribution shows up on the scoreboard without them ever winning a duel.
That means characters whose value comes from placement, cooldowns, or auto-targeting instead of crosshair control. In Marvel Rivals, Cloak and Dagger’s healing locks on, so there’s no aim requirement at all. Scarlet Witch’s primary auto-locks in range. Rivals even ships a star difficulty rating on every hero, which makes it easy to hand out assignments without anyone feeling singled out. In Overwatch it’s Moira, Mercy, Torbjörn, Orisa. In Apex it’s Gibraltar, Lifeline, and Bloodhound, whose scan turns bad game sense into shared information.
One warning that goes against the usual advice: do not put your weakest player on solo tank. Tank in modern Overwatch is the highest-blame seat in the game. One bad engage loses the fight and nine other people watch it happen in real time. Second support is much kinder.
Then pair them. This is the single best in-match trick I know and it comes from tactical shooters, where it’s called trading.
The weak player’s entire job becomes one sentence: stay close to your buddy, and shoot whoever kills them. They don’t have to win the fight. They have to turn their buddy’s death into a one-for-one. That’s a positioning task, not an aim task, and positioning is learnable in a week.
In Overwatch we do it as a hard pocket. You heal John. John doesn’t die. That’s the whole job. It’s a lot easier than free-roaming and trying to triage five people at once.
Rule Four: Put the Handicap on the Scoreboard
Every real sport solved mixed-ability play a century ago, and video games somehow forgot.
Golf, bowling, chess, Go, sailing. All of them handicap the same way: the adjustment is set before play, it’s completely visible, it’s based on tracked results, and it deliberately does not close the gap all the way. Golf’s handicap index averages your best 8 rounds out of the last 20, which means it measures your potential rather than your average, which means the better player still wins more than half the time on purpose.
Video game difficulty adjustment does the exact opposite. Hidden, mid-match, deniable. Designers treat players noticing it as a failure.
You don’t have that problem. Your friends know each other. You get to use the sports version.
Overwatch custom games hand you the sliders for free. Damage dealt, damage received, healing, health, movement speed, ultimate generation, ability cooldowns, all adjustable per team from 10 percent to 500 percent. There’s a per-hero toggle for all 40-plus heroes, which is a ban system nobody uses. Marvel Rivals gives you the same effect through its free in-client Tournament Room, which does best-of-1 through best-of-9, three different ban and pick systems, host-controlled team shuffling, coach slots, and bot fill.
But here’s the part that makes it fun instead of humiliating, and I got this from a designer writing about Smash Bros.
Don’t frame the handicap as a punishment for the good player. Publish it as a scoreboard. The steepest handicap you can still win at becomes your rank in the group.
That designer’s friends went from refusing to play with him at all to competing over who could win at the lowest handicap. The thing that would have embarrassed the weak player became the thing that entertained the strong one.
We run a version of this now. Our top two players start at 80 percent damage dealt in customs and we ratchet it down until it’s close. There’s a whiteboard channel in Discord. It’s the most competitive anyone in the group has been about anything.
One thing that doesn’t work, and I tried it: scoring multipliers. Making your Silver’s kills count double is patronizing by construction. It announces that their contribution needed inflating. Every serious mixed-ability sport went with composition rules instead, meaning rules about who has to be involved in the play. Make them necessary, don’t make them subsidized.
Rule Five: Never, Ever Run a Captain Draft
We did captain drafts for about three months. It was slowly poisoning the group and I didn’t notice for weeks.
A draft is a public ranking of your friends from best to worst, performed out loud, every single week. Somebody is always last. Everybody watches it happen.
The NBA tried to fix this in 2023 by drafting all the reserves before the starters, purely so nobody would know who went last. It didn’t work. As one writer put it, someone is still the final pick and everyone still knows who that player is. Rearranging a public ranking doesn’t help. Removing it does.
What we use now:
Discord in-house bots already solve this and they’re free. Team Up GG has an Elo-balanced mode that finds the two team compositions with the closest total rating. It also has an anonymous queue where team assignments stay hidden until reveal. In-House Queue does something similar with its own MMR and hidden player names.
Let the bot pick the teams. Blame the bot. Nobody gets picked last if nobody is picking.
A few smaller things that matter more than you’d think:
Name your squads, don’t tier them. Team Nightshift, not B Team. A letter grade carries a verdict.
Rotate one person between squads every week by rule. Then squad membership never reads as permanent.
If you can’t fill two teams or find an opponent, split your own group and scrim yourselves. For a group with one Grandmaster and a few Silvers, this is the only format where they can be in the same match with no matchmaker taxing them for it. No wide-group penalty, no queue, no rank at stake, and you control every variable.
Pickup basketball solved this decades ago with “shooting for teams,” where players shoot threes and the first group to make theirs becomes a team. It’s a randomizer, not a draft. Fast, balanced, nobody’s feelings involved.
Rule Six: Everybody Thinks They Carried
This one is my favorite piece of research in the whole pile.
Psychologists have measured how much credit people claim for a group’s result. The claims consistently add up to more than 100 percent, and the excess grows with the group. Married couples land around 103 percent. Six-person groups hit about 114 percent. When people recall a group they were part of, groups of five to ten average 171 percent, and the worst case in the data hit 236 percent.
The reason is simple. You have vivid memories of everything you did and almost no visibility into what anyone else did.
So “I carried that game” is the statistically expected belief. It is not evidence of a carry. Your Grandmaster genuinely feels like they carried, and their feeling is roughly as reliable as everyone else’s, which is to say not very.
There’s a tested fix, and it’s stupidly cheap. Prompting people to explicitly list what other members contributed cut the over-claiming substantially across every study they ran.
We do a version at the end of the night. One lap around the Discord, everybody names one thing a teammate did well. Takes ninety seconds. It’s not a team-building exercise, it’s a correction for a measured bias, and the vibe difference is not subtle.
The Thing That Makes Weak Players Try Harder
Two pieces of research sit right next to each other here and they point in opposite directions, which is what makes them useful.
The social loafing literature says people slack off more when they expect their teammates to perform well. Your Silver on a team with a known Grandmaster is sitting in the textbook condition for checking out. It’s a predictable response to feeling redundant.
Then there’s the Köhler effect, which comes from experiments with a Berlin rowing club back in 1926 and has been replicated for a hundred years since. Weaker group members increase their effort on conjunctive tasks, meaning tasks where the group’s result is set by its weakest member. They frequently beat their own individual best.
Task structure decides which one you get.
So restructure toward objectives that need everybody. A round that ends if anyone dies. A mission where each person has their own piece to finish. A team score equal to the lowest individual score instead of the sum. Anything that stops your Grandmaster from quietly absorbing your Silver’s shortfall.
There’s a number attached too, and it answers a question everybody guesses at. The motivation gain peaks when the weaker member is at roughly 70 percent of the stronger one’s capability. Bigger gaps kill the effect. Smaller gaps kill it too. So handicap toward 70 percent, not toward equal.
My favorite finding in this whole area: the effect gets stronger when the more capable partner visibly struggles. Effortless dominance kills the weaker player’s motivation. Which is a research-backed excuse to make your best player try something they’re bad at on a regular basis. We rotate in a game nobody’s good at once a month for exactly this reason. Usually something with physics in it.
What to Actually Play, Sorted by How Many Showed Up
Here’s the thing that took us longest to figure out, and it turned out to be good news.
Three to six people is a great number. It lands almost exactly on the party sizes these games are built around, and it opens up the entire co-op catalogue, most of which caps at four or six players and was quietly designed for this exact problem.
Small groups are also easier on the social side. That responsibility over-claiming research? It’s about 104 percent in a group of three and 113.6 percent in a group of six. Social loafing runs lower in groups of three to five too. The night size we already have is the forgiving end of the range.
So now we check the Discord, count heads, and pick from a list instead of arguing for twenty minutes.
Three people. Deep Rock Galactic or Helldivers 2 with the fourth slot open. Or an Apex Mixtape trio, which is a native fit.
Four people. This is the sweet spot and most of our best nights live here. Deep Rock Galactic is the best-designed skill-gap flattener I’ve found in any genre. The mission is resource extraction, not kills, so somebody mining and running resupply is genuinely load-bearing. The four classes hard-partition the work, and the Scout literally can’t reach the gold without the Engineer’s platform. The official difficulty scaling keys off hazard level and player count only, with the wiki spelling out that player experience and individual skill don’t trigger any automatic adjustment. No hidden rating. No scoreboard. Same payout for everyone.
Helldivers 2 does the same job with more explosions. Stratagems flatten aim completely, since a 500kg bomb doesn’t care how good your tracking is. Friendly fire sounds punishing and actually inverts the hierarchy, since your Grandmaster’s aggression is about as likely to kill the squad as your Silver’s confusion, and both deaths are funny instead of blame-worthy. Reinforcements are a shared budget, so dying is a team resource rather than a personal stat.
Also good at four: PEAK, which is eight bucks and hilarious and where the strong player literally ropes and anchors for the weak one. And Left 4 Dead 2, which is still very much alive in 2026 and where getting incapacitated turns you into a puzzle for your friends to solve instead of dead weight.
Just remember: no Overwatch Competitive at four. Quick Play is fine.
Five people. This is your Overwatch night. Quick Play Role Queue has zero restrictions, it’s the healthiest queue in the game at roughly 54 percent of daily players, and the matchmaker tries to pair your five-stack against another five-stack.
Six people. Best all-round number you can get. Marvel Rivals six-stack, which is a full team. If everyone’s Gold or below you can do that in Competitive, which is the single best mixed-skill ranked experience available anywhere right now. Otherwise Quick Match. Overwatch Open Queue 6v6 and Mystery Heroes also fit exactly, and Mystery Heroes flattens the gap on its own since nobody gets to play their one-trick.
For co-op at six, R.E.P.O. is the pick. It’s about ten bucks, it’s the healthiest game in the silly-horror category by a mile, and the entire loop is physically carrying fragile objects with a physics grabber. Two people carrying a piano is a comedy scene, not a skill test. Nobody is good at this game. Killing Floor 2 also seats six, and its Zed Time bullet-time is a shared reward your best player generates for everyone, so your Silver gets slow-motion trick shots they couldn’t otherwise land.
Eight people, the rare night. Plan this one. It comes around every few months and improvising wastes it.
The answer is an Overwatch Custom Game, 4v4, with the per-team sliders turned on. It’s the only setting in existence where your Silver and your Grandmaster play one match on genuinely negotiated terms, and it’s the thing you can’t do on a normal Tuesday. Marvel Rivals Tournament Rooms do the same job with a ban and pick phase and a Random Swap button that reshuffles teams between games.
Then cool down with something dumb. Golf With Your Friends holds twelve, so eight fits with room, everyone putts simultaneously so there’s no waiting, and it’s about a buck fifty on sale until the end of August. Pummel Party and Party Animals both seat exactly eight. Left 4 Dead 2 Versus is 4v4 and you should put your two best players on Special Infected, since that side rewards coordination over aim.
One category to skip entirely, and I wasted money learning this: the big-lobby games. Barotrauma, Among Us, Team Fortress 2, Crab Game, LOCKDOWN Protocol. All genuinely great. All of them need ten to sixteen bodies before they sing. If your group fields four on a Tuesday, they’ll just sit in your library making you feel bad.
Two rules that matter more than which game you pick.
Assign roles out loud instead of letting people self-select. Self-selection is how your most casual player ends up on Sniper and has a terrible night.
When you do split into two groups, don’t spread the skill evenly. Put the strong players together on a hard difficulty and the casual players together on an easy one. Everyone gets a challenge that fits, and nobody has to watch themselves being carried.
When You Still Want a Hero Shooter Night
Marvel Rivals at six, if the whole group is Gold or below: queue Competitive. Genuinely the best mixed-skill ranked experience available anywhere right now. If anyone’s above Gold, run Quick Match as a six-stack or set up a Tournament Room with hand-balanced teams and bot fill. Doom Match is a free-for-all, which is worth knowing about since there’s no team to let down.
Overwatch at five: Quick Play Role Queue. At six: Open Queue or Mystery Heroes. At four: Quick Play only, and remember the wide-group-of-four block. Skip Competitive entirely unless the spread is genuinely narrow that night.
Apex at three: Mixtape. TDM, Control, and Gun Run, rotating every hour. Death costs you seconds instead of a match, which removes the exact thing that makes battle royale brutal for a weaker player. Wildcard is a decent second option with its auto-healing and faster revives. Six people splits cleanly into two trios. Ranked as a mixed squad is a trap and you should leave it alone.
Change the Structure, Keep the Friend
That Silver friend of mine still plays with us every Friday. His aim hasn’t improved much. He’s not going to climb, and honestly he doesn’t want to.
What changed is that he’s now the guy who runs support on a hard pocket, calls one thing per fight, holds the trade angle, and gets named in the end-of-night lap about as often as anyone. He hasn’t gotten better at the game. We got better at running the group.
The reason this took me two years is that I kept looking for a player-shaped problem. There wasn’t one. The matchmaker builds every lobby around the best person in your party, adds a surcharge for being organized, and hands the whole bill to the person least able to pay it. No amount of coaching fixes a rule.
You have a Silver in your group. You probably have a Grandmaster too. Neither of them is the reason your Friday nights feel bad.
Change the structure. Keep the friend.
Got a group with a canyon-sized skill gap? Come tell us what’s worked and what blew up in your face. We’re running mixed-skill nights in the TAG Discord every week and we’d love to steal your house rules.
