K9999-ex wins 92.7% of fights. Captain Commando wins 10.4%.
Same engine. Same platform. Same 24/7 stream. An 82-point gap in win rate between two characters running on the same MUGEN build. MUGEN wasn’t built for balance — and over half a million matches of SaltyBet data prove it.
Here’s the thing about MUGEN that everyone who’s used it already knows: it’s an open-source fighting game engine where anyone can create a character. There’s no balance team. No patch notes. No competitive testing committee reviewing submissions. The 10,000+ fighters on SaltyBet were made by thousands of different creators over the span of decades, each with their own ideas about how strong a character should be.
The data shows what happens when you throw all of them into an arena together.
We looked at every fighter with at least 50 matches in our database — 6,997 fighters total. Here’s how their win rates distribute:
| Win Rate | Fighters |
|---|---|
| 0-19% | 86 |
| 20-29% | 652 |
| 30-39% | 1,304 |
| 40-49% | 1,557 |
| 50-59% | 1,547 |
| 60-69% | 1,241 |
| 70-79% | 509 |
| 80%+ | 101 |
The good news: most fighters cluster in the 40-59% range — 3,104 of them, 44% of the roster. The system roughly works for the majority. But the tails are extreme, 86 fighters win less than 1 in 5 (insert leonardo-di-caprio-gif). 101 fighters win more than 4 in 5. And the gap between the very best and very worst is 82 percentage points, a sizeable delta.
Among fighters with 50+ matches, these are the most dominant:
| Fighter | Win Rate | Record | Tier |
|---|---|---|---|
| K9999-ex | 92.7% | 51-4 | X |
| Shiva-hell | 89.1% | 204-25 | S |
| Primeus | 89.1% | 171-21 | S |
| Element -a- | 88.8% | 174-22 | S |
| Flawless specter | 88.4% | 191-25 | S |
| Mega weapon | 88.0% | 162-22 | S |
| Natural kain | 87.6% | 198-28 | S |
| Bloodbane | 87.4% | 160-23 | S |
| Vergil umvc3 | 86.9% | 226-34 | S |
| Android 16 ub22 | 86.9% | 232-35 | S |
K9999-ex sits at the top with 92.7% — this fighter has lost only four times. Ever. Across 55 matches, it’s nearly perfect.
I found this to be pretty surprising, so I dug in: all to X-tier opponents: R-akatsuki, Shelby cobra, Gustab m. type-y, and Vatista. Against everyone else? Undefeated, mic drop.
K9999-ex has never fought outside of X-tier — which makes the 92.7% win rate even more impressive. It’s not beating up on weaker opponents. It’s dominating the tier that’s supposed to be reserved for the most overpowered characters.
But here’s the pattern that jumps out: 9 of the top 10 are S-tier. Zoom out further and it gets even more stark — 89 of the 101 fighters with 80%+ win rates are S-tier. Not X-tier, not A-tier, but S-tier.
You’d expect X-tier — the tier reserved for the most overpowered characters — to dominate these lists. It doesn’t. Only 4 of the 101 are X-tier. The reason? X-tier fighters mostly face each other. When everyone in your tier is overpowered, win rates compress toward 50%. Dominance requires weaker opponents, and S-tier gets plenty of those.
Now let’s look at the other end:
| Fighter | Win Rate | Record | Tier |
|---|---|---|---|
| Captain commando | 10.4% | 7-60 | P |
| Sub-zero umk3 | 10.8% | 8-66 | P |
| Snake man | 11.9% | 7-52 | P |
| Tiblekun | 12.4% | 11-78 | B |
| Goku ssj5 EX | 12.8% | 10-68 | P |
| Piccolo mvc | 12.8% | 12-82 | B |
| Deltarune lancer | 13.0% | 7-47 | P |
| Mono freaks | 13.0% | 7-47 | P |
| Illusion no.17 | 13.2% | 12-79 | P |
| Shang tsung mk2 | 13.9% | 10-62 | P |
Captain Commando loses 9 out of 10 fights. Goku SSJ5 EX — a name that sounds like it should destroy everything — wins 12.8% of the time. The name on the select screen doesn’t mean what you think it means in MUGEN. (Subtle reminder that when faced with a choice between Goku and Vegeta, Vegeta is your better bet).
The tier pattern mirrors the top end perfectly: 8 of the bottom 10 are P-tier (the lowest tier), 2 are B-tier. S-tier owns the winners. P-tier owns the losers. The tier system is doing what it’s supposed to do — sorting fighters by strength — but within tiers, the spread is still enormous.
These fighters keep getting scheduled because SaltyBet cycles its full 10,000+ roster. There’s no retirement. Captain Commando will fight again tomorrow, and he’ll probably lose again. That’s just how it is in the salt mines.
What makes a MUGEN character dominant? We can’t see inside the code, but after analyzing over half a million matches, the data points to a few mechanical factors. If you want the full technical deep dive on how MUGEN characters work, we wrote a separate piece on that. Here’s the short version:
AI aggression. Every MUGEN character has an AI file that controls how it behaves — when to attack, when to block, when to use specials. Some characters are coded to attack relentlessly, chaining moves together with minimal downtime. Others idle, walk into corners, or fail to block incoming hits. The AI file is often the single biggest factor in whether a character wins or loses, regardless of how impressive its sprites look.
Infinite combos. Some movesets allow inescapable combo loops. Once the AI lands the first hit, the opponent never gets back up. Characters with these combos can end fights in seconds, regardless of the opponent’s stats.
Projectile zoning. Characters that spam projectiles from full screen can win without ever taking a hit. If the opponent’s AI doesn’t know how to close distance or jump over fireballs, the fight is effectively over at round start.
Hitbox and hurtbox mismatches. Some characters have tiny hurtboxes (hard to hit) or massive hitboxes on their attacks (easy to land). When a character with both advantages fights one with neither, the outcome is predetermined before the first frame.
If you’ve built MUGEN characters or studied the AI, you probably know more about the specifics than we do — we’re reading the outcomes, not the code. The data tells us that these fighters dominate. The MUGEN community knows why.
After looking at 6,997 fighters across over half a million matches, a few things stand out:
Most fighters are balanced enough. The 40-59% cluster — 3,104 fighters, 44% of the roster — suggests that for nearly half the characters, the system works reasonably well. They win some, they lose some, and the outcomes aren’t predetermined.
The extremes are extreme. MUGEN’s open nature means there’s no floor or ceiling on character quality. A creator can make a fighter that’s virtually unbeatable or one that can barely function. Both end up in the same arena.
S-tier is where dominance lives, not X-tier. This was the most counterintuitive finding. The tier you’d expect to house the dominant fighters (X) actually compresses win rates because those fighters mainly face each other. S-tier is where fighters get to feast on weaker opponents — and the data shows it.
The tier system helps, but it can’t fix everything. SaltyBet sorts fighters into tiers based on performance, which is the closest thing MUGEN has to a balance mechanism. But within each tier, the spread is still massive. Two S-tier fighters can have win rates 20+ points apart.
This is exactly what our prediction model learns from — over half a million matches of which characters beat which, and why some fights are more predictable than others. The 80%+ fighters? Those are the easy calls. The messy middle? That’s where the real challenge is.
See these stats live during matches. The SaltyTrack Chrome Extension shows you win rates, head-to-head records, and AI predictions right on saltybet.com. Free, no account needed.
SaltyBet uses virtual currency only. No real money is wagered or exchanged. SaltyTrack is not affiliated with SaltyBet.