The arena format is the joke. The method underneath it is not. Here is exactly how a token gets scored, why paid promotions are quarantined, and why we publish on a fixed clock.
Every figure on a fight card is pulled from public APIs before publication. We do not take numbers from project Telegrams, influencer threads, or anyone whose job depends on the number being big. Four feeds, each with a specific job.
GeckoTerminal supplies pool-level data: price, OHLCV, liquidity in USD, 24-hour volume, buy and sell transaction counts, and unique buyer and seller counts across 200-plus networks and 1,500-plus DEXes. The unique-buyer count matters more than it sounds. Volume can be manufactured. Distinct wallets are harder to fake at scale.
DEX Screener supplies pair data and, more importantly, its paid-products feed. It openly sells Token Profiles, Boosts, and ads, and exposes endpoints listing the latest boosted tokens and the tokens with the most active boosts. We poll those endpoints. This is how we know who paid for visibility.
Moralis covers the pump.fun lifecycle on Solana: newly created tokens, tokens mid-bonding-curve with exact graduation progress percentages, and graduated tokens with liquidity and graduation timestamps. A token at 40 percent bonding progress and a token that graduated six hours ago are different animals, and the card says which one you are looking at.
CoinGecko provides the reference layer: broad market data, sector aggregates, and the denominators we use to say whether a move is a move or just the whole market breathing.
Five inputs, scored mechanically. 24-hour volume (is anyone actually trading this). Liquidity (can anyone actually exit). Volume-to-liquidity ratio, which is the one that catches liars: a token doing 30x its pool depth in daily volume, especially on a low transaction count with repetitive trade sizes, is exhibiting the classic wash-trading signature documented in academic work on DEX manipulation. We score healthy turnover up and cartoonish turnover down. 24-hour price move, capped in the scoring so that one 900 percent candle does not outrank everything with a pulse. And red flags, which subtract: unlocked or unlockable liquidity, mint authority not revoked, top-holder concentration where deployer-linked wallets sit on 40 percent or more of supply, and buy/sell asymmetry consistent with a honeypot (hundreds of buys, single-digit sells is not enthusiasm, it is a wall).
The score decides card position. Main event, undercard, or a line in the obituaries. No editorial override. If a token we find funny scores badly, it scores badly.
Matchmaking is mechanical. The day's field is ranked by composite score and paired off the board: number one fights number two in the main event, three fights four and five fights six on the undercard. Nobody ducks anybody, because nobody gets asked. An odd field leaves the lowest seed without a dance partner and we say so.
Each bout is scored over five judged rounds, each round tied to one metric family. Round one is volume: half again the opponent's turnover takes the round 10-9, ten times takes it 10-8, anything closer is even. Round two is liquidity, the depth of the exit door, judged on the same bar. Round three is momentum: the 24-hour move, capped in judging exactly as it is capped in the composite score, so one 900 percent candle cannot steal a fight. Round four is turnover, volume divided by liquidity, and it is judged on health rather than size: churn up to about 30x pool depth scores as activity, churn past 50x scores as the wash-trading signature and loses the round no matter how large the raw number is. Round five is discipline: fewest red flags on file wins.
Rounds are scored 10-9, or 10-8 for a blowout, or 10-10 when the judges genuinely cannot separate the two. Totals decide the bout: unanimous decision when the loser takes no rounds, split decision otherwise, majority draw when the totals tie. The desk line quoted on each bout converts the composite-score gap into American odds through a fixed logistic curve. It is deterministic, it is not a prediction, and it is certainly not a price. Same inputs, same line, every time.
None of this changes what gets measured. The bout is a presentation layer over the same five scored inputs described above. It exists because a side-by-side comparison under adversarial framing is a genuinely better way to read two tokens than two separate table rows. The joke is load-bearing.
Every card is archived, and the archive is what makes the records real. A fighter's W-L-D line is the sum of its judged bouts across every card it has appeared on, recomputed deterministically from the archived data. Nothing is stored that cannot be re-derived.
The championship is lineal, boxing rules. The first main-event winner took the vacant belt. The belt changes hands in exactly one way: the reigning champion loses a judged bout. Winning a main event while holding it counts as a title defense. Not appearing on a card is not a loss; the champion keeps the belt in absentia and the desk notes nothing, because there is nothing to note.
The Obituaries are checked daily. Any fighter featured on a past card with at least $10k of liquidity whose pool now holds under $1k, or whose pair has vanished from the index entirely, is pronounced. Cause of death is recorded mechanically. The median token covered here goes to zero; the obituary section is that sentence, enforced.
The SHIT Index is a volume-weighted composite of the tokens currently on the card, rebased at each publication. It tracks one thing: whether the memecoin casino in aggregate is getting hotter or colder, independent of any single token's fate. It is not investable, which given the constituents is probably a feature. When the Index and majors diverge, the divergence gets a line in the write-up. When they do not, it does not.
Tokens whose teams purchase promotion, which we detect primarily through DEX Screener's boost and ad endpoints plus disclosed marketing spend where we can find it, are quarantined to a separate section and are never eligible for the main event, regardless of score. The logic is not moral. It is statistical. A purchased trending multiplier contaminates exactly the attention metrics our scoring depends on. We cannot cleanly score a token that is paying to distort its own inputs, so we do not pretend to. The Desk reports them, labels them, and keeps them away from the organically scored card. If a token stops boosting, it becomes eligible again after a cooling-off period, because the data does.
Cards go out on a fixed calendar, full stop. We do not publish early because something is pumping and we do not hold publication because something is dumping. A publication that times its output to market events is, functionally, participating in them. Publishing on schedule means nobody, including us, can use our release timing as a trading signal, and nobody can pay to accelerate coverage. It is the cheapest compliance control available and it costs us nothing except the occasional stale-looking headline, which we accept.
This is entertainment and information. It is not financial advice, and we are not being coy: we affirmatively announce that we do not advise. We do not tell you to buy anything, we do not tell you to sell anything, and if you derive a trading strategy from a publication formatted as a fight card, that decision and its consequences are yours. The median token covered here will go to zero. That is not a warning. That is the base rate.
Play-by-play of an absurd market, published on a fixed schedule. Subscribe to the newsletter; it costs nothing, unlike the tokens it covers.