Guide · the mechanic, and its arithmetic
A bid board’s race ends. Here is exactly when.
Hundreds of pay-to-rank boards launched in a fortnight and nearly all of them rank by money. That choice has a consequence nobody writes down: the price of first place can only go up, so the race stops the moment one person is willing to outspend everyone else. This page works out what happens instead when rank is a score.
- Rank by money alone
- 90 of 100the rest sell a slot, or make you earn it
- Chance the k-th entry leads
- 1 / kfor any continuous score, any distribution
- Leaders in a million entries
- ~14H(1,000,000) ≈ ln(n) + 0.577
- Median top bid in the wave
- $5measured 2026-09-09
1 · What a pay-to-rank board actually is
One public list, no algorithm, no queue. Position is the product.
You submit a URL, you pay, and the amount decides where you sit on one page everybody sees. There is no ranking algorithm to reverse-engineer, no moderation backlog, and no eligibility rules — which is the whole appeal, and the reason the format spread to somewhere between 355 and 464 sites in a fortnight depending on who is counting.
Underneath the categories there are only 5 mechanics, and two of them account for nearly everything: pay to climb (58 of 100 boards), where your payments accumulate into a running total, and highest bid wins (32), where one number decides it. The other 10 sell a fixed slot, make you earn the rank, or run an open auction after paying to enter. The full count, board by board
2 · Why the race ends
The incumbent sets the bar. That is the whole mechanism.
On a bid board the price of first place only ever goes up.
to take #1 you must pay more than the current #1 paid so the next bid is strictly higher than the last and the race ends when one person's ceiling is above everyone else's
Nothing is broken when a bid board stops moving — it has finished. The leader is whoever had the highest ceiling, and every challenger now faces a bar that leader set. This is why a board can look busy, carry a hundred rows, and have had the same name at the top since the week it launched.
It is also why a time-bound listing is a different product from a permanent one. 75 of the 100 boards sampled sell a rank that expires, so the price on the board is a rate, not a price — and when it lapses you are bidding against the ceiling all over again.
3 · What happens if rank is a score instead
The arithmetic of records, which is the same for every continuous distribution.
Suppose rank is not a price but a score drawn fresh each time — and suppose the score is continuous, so no two entries tie. Then the k-th entry is the best so far with probability exactly 1/k. Not approximately, and not dependent on the shape of the distribution: it is simply the chance that, of k things in a random order, the newest one happens to be the largest.
Add those up and the expected number of record entries over n of them is the harmonic number H(n) = 1 + 1/2 + 1/3 + … + 1/n, which is close to ln(n) + 0.577. It grows logarithmically, which means it barely grows at all:
| Entries | Expected record entries | Of which are lead changes | Chance the newest one leads |
|---|---|---|---|
| 100 | 5.19 | 4.19 | 1 in 100 |
| 1,000 | 7.49 | 6.49 | 1 in 1,000 |
| 10,000 | 9.79 | 8.79 | 1 in 10,000 |
| 100,000 | 12.09 | 11.09 | 1 in 100,000 |
| 1,000,000 | 14.39 | 13.39 | 1 in 1,000,000 |
The middle column is the one people quote; the third is the honest one, because the very first entry sets the record without taking it off anybody. A hundred entries produce about four lead changes. A million produce about thirteen. Frequent enough that the board is a live race, rare enough that every takeover is an event worth posting — and, unlike a bid board, the difficulty curve arrives on its own. Nobody has to throttle anything, decay anything or write a rule.
This is the arithmetic OutRip runs on. A card’s rating is 1 / √p over its six trait probabilities, so it is continuous and unbounded — there is no top card that ends the game, and first place is always beatable, just harder each time. Every frequency table and the worked example
4 · The one number to check before you pay
Not the rank price. What the board has already taken.
Every board on earth will tell you what #1 costs. Almost none will tell you whether anyone has ever bought it. Across the 94 boards that publish an observed top, measured 2026-09-09: the median is $5, 81 of 94 are under $25, and 4 sit at exactly $0 — nobody has ever paid them anything. Median entry price across the 87 boards quoting one is $1.
One board is the exception that created the genre: seethebids.lol publishes $238,741 of revenue and 1,411,967 visitors on its row for outbid.lol, while outbid.fyi puts $363K across all 355 boards it lists — so a single site is most of the money in the category. Everything downstream of it is much smaller than its rank card suggests. Rank #1 in front of nobody is worth nobody, whatever it costs.
Three of the free directories publish per-board spend, so you can check any board in about a minute before you pay it anything — the six of them, and what each one publishes.
Questions
What people ask before spending on one of these.
How do pay-to-rank leaderboards work?
You submit a URL and pay, and the amount you pay sets your position on one public list. Of the 100 boards rankmarkets.lol renders on its index, 90 rank by money alone: 58 add up everything you have paid, and 32 take the single highest bid. The remaining 10 sell a fixed slot, make you earn the rank after paying to enter, or run an open auction. There is no algorithm and no moderation queue, which is the entire appeal. Measured 2026-09-09.
Why does the same person stay at #1 on a bid board?
Because the price of first place only goes up, and it goes up by whatever the last person paid. Every challenger has to clear a bar that the incumbent set, so the race stops as soon as one person is willing to spend more than anyone else finds it worth. The board has not broken; it has finished. That is why a board can look busy and have had a single leader for its entire life.
How often does first place change hands on a leaderboard?
If rank is decided by a continuous score rather than by a price, the k-th entry takes the lead with probability exactly 1/k, whatever the distribution is. So the expected number of record entries over n of them is the harmonic number H(n), which is about ln(n) + 0.577. Ten thousand entries produce about ten leaders; a million produce about fourteen. The lead changes often enough to be a live race and rarely enough that each one is an event.
Are pay-to-rank boards worth paying for?
Check one number first: what the top spot on that board has actually sold for. Across the 94 boards that publish it, the median is $5, 81 are under $25 and 4 have never taken a dollar. A board with no spend has no visitors, and rank #1 in front of nobody is worth nobody. outbid.lol is the exception that made the genre — seethebids.lol puts it at $238,741 of revenue across 1,411,967 visitors. Measured 2026-09-09.
Is a pay-to-rank leaderboard gambling?
A bid board is not: you pay a known amount for a known position, and nothing is left to chance. Boards that add a random element are the ones to read carefully, and the question to ask is whether there is any outcome where you receive less than you paid for. On OutRip there is not — every pack returns at least what you paid in card value, and the floor is asserted by the mint before any card is written, not promised in marketing copy.
Sources. The board counts, ranking models, entry prices and observed tops were read off the directories’ own pages on 2026-09-09 — rankmarkets.lol for the per-board model and price sample (100 rows rendered of 369 it reports indexed), seethebids.lol, outbid.fyi and leadingboards.lol for the counts, and outbid.fyi for the genre-level revenue figure, and seethebids.lol’s own row for the outbid.lol revenue and visitor counts. This page previously quoted outbidstory.lol at “roughly $200,000 in about 100 hours”; that site was re-fetched on 2026-09-02 and no longer publishes the figure, so it has been replaced rather than carried. outbid.lol itself answered 429 on 2026-09-09, as it did on the 27th, the 31st and the 2nd, and is not quoted. The records result is standard: for exchangeable continuous draws, P(the k-th is a record) = 1/k and the expected count is the harmonic number H(n) — the table above computes H(n) at build rather than restating it. Alternatives · vs outbid.lol · How rating works