How to spot a staged winning-bet video: the fee line and the round numbers give it away

A win video is a set of arithmetic claims, and the arithmetic is public. Someone showing $2,000 turn into a $10,000 payout is claiming 10,000 contracts bought at exactly $0.20, because settlement pays $1.00 per contract and nothing else. On Polymarket that fill carries a taker fee of rate x contracts x P x (1 - P), which is $112 in a crypto market, $80 in sports, $64 in politics and $0.00 in geopolitical. A caption reading "$8,000 profit" is therefore describing a geopolitical market, a maker fill, or a number nobody paid.

Then look at the entry price. The five positions our paper trader opened on 28 July 2026 were filled at $0.9296, $0.8784, $0.8573, $0.8693 and $0.2965. Not one is round, because a fill inherits the book's price and not your intention. Five checks follow, each of them arithmetic you can run from the screen, and none of them requiring that you trust us or the person in the video.

Status as of 18 August 2026. Polymarket fee coefficients read from each market's feeSchedule object on the Gamma API on 4 August 2026, schedule version 2026-08-04.a. Referral terms for both venues read at source on 5 August 2026. PredictionEdge currently takes no referral revenue from Polymarket or Kalshi, and whether it ever will is undecided.

The line that got this page commissioned is the one line in it we cannot source

Our research corpus carries a single sentence: "Polymarket paid dozens to post videos of themselves winning with fake bets." We hold no document behind it. No filing, no article, no archived post, no docket. We are not asserting it, and nothing below depends on it being true.

It is printed here rather than quietly dropped because dropping it would be the same move this page is about. Every check that follows rests on fee formulas and referral terms published by the platforms themselves, which anyone can open and read in under a minute. The claim is listed again in what we could not verify.

Every taker fill carries a fee you can compute from the screen

Polymarket charges the taker and never the maker. Makers pay 0.00, which is zero rather than a discount. Takers pay a per-category coefficient against rate x contracts x P x (1 - P): 0.07 on crypto, 0.05 on sports, economics, culture, weather and other, 0.04 on finance, politics, mentions and tech, and 0.00 on geopolitical.

Run the claimed win backwards. Payout in dollars equals contract count, because each contract settles at $1.00. Multiply the count by the entry price on screen, then add the fee.

A "$2,000 into $10,000" screenshot means 10,000 contracts at $0.20. Here is what that fill actually costs, by the category the market page would show:

There are exactly two honest ways for "$8,000" to be right. The market is geopolitical, coefficient 0.00, both sides free. Or the fill was a maker fill: a resting order sitting at $0.20 that somebody else hit, which costs nothing on this venue.

Both are checkable against the video. The category is a label on the market page. And a maker fill is not what these videos describe. They describe taking a price that was already there, usually with the word "aped" or "sent it" attached, which is the taker side of the book by definition.

The cheaper the entry, the wider the gap

Fee as a share of stake is rate x (1 - P), because the contract count cancels. It rises the cheaper the contract gets, so the longshot corner of the board is the most expensive place on the venue to take liquidity, which is exactly the corner a 20x screenshot has to come from.

So the 20x is not a 20x. Put $100 into a crypto market at $0.05 and you buy 2,000 contracts. The fee is 0.07 x 2,000 x 0.05 x 0.95 = $6.65, so $106.65 leaves your account and $2,000 comes back. That is 18.75x on money actually deployed. The bigger the multiple in the thumbnail, the further the screen figure sits from the earned one.

On Kalshi, a taker fill with a zero fee line cannot exist

Kalshi's published taker formula is roundup(0.07 x C x P x (1 - P)), documented on docs.kalshi.com for engineers rather than for traders. It rounds up to the whole cent, and the round-up of any positive number is at least one cent. A Kalshi screenshot showing a taker fill with a fee of $0.00 is showing something the formula cannot produce.

The rounding also falls hardest on the smallest orders, which is the opposite of what most people assume:

One contract quoted at 50 cents costs you 52. That is 14% more per contract than the same trade at 100 contracts, from rounding alone.

The trade in the video has to be a trade that can exist

The risk-free lock, buying YES and NO for a combined price under $1.00, does not occur inside a single Polymarket market. The Gamma API exposes bestBid and bestAsk only, and the NO price is the arithmetic complement of YES, so yes_ask + no_ask = 1 + spread by construction. The sum is at or above $1.00, always.

We did not take this on faith. Our own scanner was built to detect intra-market arbitrage and found none, because there is none, and two independent systems reached the same conclusion from different directions: a session review on 28 July 2026 and our collector on 31 July 2026. What the scanner actually surfaces is momentum, not arbitrage. Any video demonstrating a within-market YES and NO lock on Polymarket is demonstrating an interface that does not behave that way.

Cross-venue arbitrage against Kalshi is a separate question. It is open. We have not tested it, it needs a market-equivalence matcher neither of our systems has built, and our own fee arithmetic suggests the round trip eats most of what it would find.

Real fills land on ugly prices

Our paper trader opened five positions on 28 July 2026. The entry prices are the point:

Five prices, four decimal places each, none of them round. That is five observations and it supports no percentage. We are not going to tell you what share of real fills land on round numbers, because we have not measured it.

What it does show is the mechanism. A round stake and a round entry price cannot both be chosen: you pick the stake, the book picks the price. Work it through on any ordinary unround quote. Take $2,000 of a politics contract at, say, $0.2137 and you get 9,358 whole contracts for $1,999.80, plus a fee of 0.04 x 9,358 x 0.2137 x 0.7863 = $62.90. Total out of pocket $2,062.70, payout $9,358.00, net $7,295.30. Not one figure in that sentence is round, and that is what a real fill reads like.

Those five sit inside a larger log. Our own trade log stands at 10 trades, 6 closed, net -$28.30 on $500 deployed as of 4 August 2026, and the $500 is the figure for the log as a whole, not for the five entries above. We publish the losing ones because they are the ones with the instructive arithmetic. Trade 4ad1b0 above was a guaranteed loss at the moment of entry: at $0.2965 the fee alone is 2.81% of stake and the position was opened on a claimed edge of 1.20%, before the market moved at all.

Whoever paid for the video is paid on your taker fees

The economics of a win video are the economics of a signup. Both Polymarket routes were read at source on 5 August 2026.

"Net trading fees" means what Polymarket keeps after the referred user's own rebate, and makers are never charged anything. So a referred user who posts resting orders generates the referrer exactly $0.00, forever. So does a referred user trading geopolitical markets at coefficient 0.00. The programme pays only on taker fills in fee-charging categories.

Put a number on it. To earn $100 through the in-product route at 10% of net fees, the referred person has to pay $1,000 in taker fees inside the 30-day window. In a crypto market at $0.50 the fee is 0.07 x 0.25 = $0.0175 per contract, so $1,000 of fees is 57,143 contracts, or $28,571 of notional at $0.50, from one person, within 30 days.

That shape explains the genre better than any accusation does. It rewards a high volume of taker fills in fee-charging categories inside a short window. It pays nothing at all on the person who becomes a patient maker, and it stops early on the person who becomes good enough to reach Platinum. Someone earning 10% of what you lose to taker fees is not indifferent to which trades you make.

The Dub route is the other half. It pays a flat $10 per funded first deposit with no gate on the publisher, which rewards the signup itself, once, regardless of what happens after. At that rate 1,000 clicks is $10. The deposits are the business.

The $10,000 gate is an account-age check you can run without seeing the account

The in-product programme requires the referrer to hold $10,000 of lifetime trading volume before it pays anything. A creator with a new or small account cannot be earning through it. If their link pays, they are on Dub, on a direct arrangement, or not being paid at all. Asking which one is a fair question, and "the official referral programme" is itself a claim about their own trading history.

Kalshi does not have the programme people imply it has

Read at help.kalshi.com on 5 August 2026, Kalshi's referral scheme is an ordinary user refer-a-friend programme, in the page's own framing. Credits are not cash: only profits made trading with them become withdrawable. Credits expire 7 days from issuance. The code must be entered within 72 hours of account creation and before the first deposit. The referred person must be a US user, and international users are not eligible. There is a lifetime cap whose amount is not published, and a payout amount that is not published either.

A non-US creator posting a Kalshi referral code is therefore not earning cash from that programme, because they are not eligible for it.

kalshi.com/affiliates has returned HTTP 429 to us on four separate attempts and we have never read it. What is documented elsewhere is that Kalshi runs named partnerships with CNN, CNBC (multi-year, exclusive), FOX (News, Business, Weather, One), Madison Square Garden, the Chicago Blackhawks and Men in Blazers. Negotiated brand and data-integration deals. We have found no application form, no rate card and no self-serve tier. That partner list sits in our research corpus with no source URL attached to it, so read it as our record rather than as a citation.

This matters because of how it gets described secondhand. On 5 August 2026 a Google AI Overview called Kalshi's arrangements "custom promotional partnerships with select media publishers", which reads like something a creator could join. The documented membership is CNN, CNBC, FOX, Madison Square Garden, an ice hockey team and a football podcast. In the same hour, a search summary stated that Polymarket pays "a 30% share" on direct referrals; the documentation says 10% direct and 5% indirect. Both were caught by opening the primary source.

The link under the video is part of the evidence

From our competitor teardowns on 29 July 2026, two documented patterns:

• A referral parameter appended to the platform URL, one operator using ?r= with their own site name and UTM tags per placement, so they can A/B test where on the page the button sits.
• A cloaked redirect, another routing through their own /link/polymarket/ path before the destination, which keeps the code out of the status bar on hover.

Neither is illegitimate by itself. Both mean the link is monetised. Hover it on a desktop or long-press it on a phone before deciding the video was disinterested.

The win screen is not the money

Two things sit between the payout screen and a bank balance, and videos end before both of them.

The first is settlement. A position that has not resolved is not a win, it is a mark. Our corpus records a market where the settlement allegedly turned on a death carveout that was not disclosed until after airstrike reports were circulating, resolving at 39.5% rather than YES, and a Spotify-related market that paid out minutes before Spotify deleted 523,000 fraudulent streams. We hold both as corpus summaries, with no resolution page or docket behind either, so read them as reported cases rather than as verified records.

The second is withdrawal, which is the dominant one-star theme in the complaint record on both Trustpilot (1.9 out of 5) and the BBB. Documented cases in our corpus include an $89.27 hold on a winning bet, identity documents submitted on 10 January and still unverified on 23 January, a $25 referral bonus never credited despite repeated assurances from support, a Bitcoin market marked as a loss when the price data supported a win, a market left open after the event had already resolved, and an app failure during live trading with a claimed loss of around $200,000.

These come from BBB and Trustpilot complaint records collected into our research corpus. We hold the summaries, not per-case URLs, so read each as a reported complaint rather than an adjudicated finding. The 1.9 out of 5 is recorded the same way: a Trustpilot rating in this niche, with no platform profile named against it in our record and no capture date. The same corpus holds a support-responsiveness count of 174 of 214 complaints unanswered, 81%, on a BBB profile in this niche. Our record does not name which platform's profile that count belongs to, so we are not attaching it to either one.

The checklist, in order

1. Read the payout as a contract count. $1.00 per contract at settlement. A $10,000 payout is 10,000 contracts and nothing else.
2. Multiply by the entry price, then add rate x contracts x P x (1 - P). If the claimed profit matches the no-fee figure exactly, the market has to be geopolitical or the fill has to be a maker fill.
3. Check the entry price for roundness. You choose the stake, the book chooses the price. Both being round means someone chose both.
4. Check the multiple. A 20x from a $0.05 entry is 18.75x after the crypto taker fee. The longshot corner carries the highest fee share on the venue, 6.65% of stake at $0.05.
5. On Kalshi, check the fee line exists. roundup of a positive number is at least one cent. A zero fee on a taker fill is not possible.
6. Check the trade is a trade that exists. No within-market YES and NO lock on Polymarket. The two asks sum to $1.00 plus the spread by construction.
7. Check whether the position settled or is merely marked. An open position is a screenshot, not a result.
8. Hover the link. A ?r= parameter or a redirect through the poster's own domain means the video is inventory.
9. Ask which programme. The Polymarket in-product route is gated at $10,000 of the referrer's own lifetime volume. Kalshi's is US-only, pays non-withdrawable credits and expires in 7 days.

What we could not verify

• The claim this page was commissioned around. "Polymarket paid dozens to post videos of themselves winning with fake bets" appears in our research corpus with no document behind it. Not asserted here.
• Order book depth at the time of any fill. The Gamma API we use exposes best bid and best ask only, and we keep no historical depth store. We cannot tell you how much size a given market could absorb, so we cannot give you a dollar figure above which a claimed fill becomes implausible. Anyone who says they verified depth from a screenshot is guessing at the same thing we are.
• Whether Polymarket rounds fractional fees. Kalshi's formula rounds up explicitly. docs.polymarket.com/programs/builders/fees has never been fetched by us. At small size, rounding is the difference between the tables above and what you actually pay.
• Whether either venue charges anything at settlement. Every figure here charges one fee, at entry, on the taking side. We hold no document saying the settlement side is free.
• Whether Kalshi charges makers. Their newsroom and their help centre contradict each other in our reading, one saying resting orders are fee-exempt and the other saying maker fees are charged. We hold no URL for the newsroom statement and nobody has reconciled it publicly.
• Contract granularity on Polymarket. Our log records prices to four decimal places, but we have not documented whether fills can be fractional, so the worked example rounds down to whole contracts.
• What disclosure rules apply to paid promotion of event contracts, anywhere. We have not read a rule, a guidance document or an enforcement action on it. This page therefore makes no claim about what a creator is legally required to say.
• Which platform the 174-of-214 BBB figure belongs to. The count is real and in our corpus. The profile it came from is not named in our record. The same gap applies to the Trustpilot 1.9 out of 5.
• How many of our own paper positions are currently open. Our scanner note records five open positions from 28 July 2026. Our trade log records 10 trades with 6 closed, which implies four. We have not reconciled the two, so the five prices above are published as five entry observations and no open-position count is claimed around them.
• A source URL for Kalshi's named partner list. CNN, CNBC, FOX, Madison Square Garden, the Chicago Blackhawks and Men in Blazers appear in our corpus without a link behind them. We have not fetched a Kalshi page that lists them.
• What kalshi.com/affiliates says. Four fetch attempts, HTTP 429 every time. We have never seen the page.
• Whether a US visitor can act on any of these links at all. The United States sits in the close-only tier on Polymarket's geoblock endpoint, on frontend and API both, as read on 5 August 2026, and polymarket.us discloses no operator, no registration status and no state list.
• Whether five entry prices generalise. They do not. n = 5.

FAQ

How do you tell if a prediction market winning video is staged? Run the payout backwards. Settlement pays $1.00 per contract, so a $10,000 payout is exactly 10,000 contracts. Multiply by the entry price shown and add the taker fee, rate x contracts x P x (1 - P). In a crypto market, $2,000 entered at $0.20 owes $112, so a caption claiming "$8,000 profit" is short by $112.

Does Polymarket charge a fee on a winning trade? It charges the taker at entry, and the maker never. The coefficient is per category: 0.07 crypto, 0.05 on sports, economics, culture, weather and other, 0.04 on finance, politics, mentions and tech, and 0.00 on geopolitical, read from each market's feeSchedule object on 4 August 2026. Whether any settlement-side charge exists, we have not confirmed.

Can you really turn $100 into $2,000 on a prediction market? Arithmetically yes, at an entry of $0.05, which buys 2,000 contracts paying $1.00 each. The crypto taker fee is 0.07 x 2,000 x 0.05 x 0.95 = $6.65, so $106.65 leaves the account and the real multiple is 18.75x rather than 20x. Cheap contracts carry the venue's highest fee share: 6.65% of stake at $0.05.

Why are the numbers in these videos always round? Because a round stake and a round entry price cannot both be chosen. You pick the stake, the book picks the price. The five entries our paper trader opened on 28 July 2026 all run to four decimals: $0.9296, $0.8784, $0.8573, $0.8693 and $0.2965. That is five observations, not a rate, and we will not publish a percentage we have not measured.

Does Polymarket pay people to promote it? It runs two documented programmes. The in-product referral pays 10% of net trading fees direct and 5% indirect, gated behind $10,000 of the referrer's own lifetime volume and ending 30 days after signup or at Platinum tier. A separate Dub partner programme pays $0.01 per click and $10 per funded first deposit, with no publisher gate stated on the page.

Does Kalshi pay influencers? Not through anything we could find. The help centre programme is a user refer-a-friend scheme paying non-withdrawable credits that expire in 7 days, US-only, with a 72-hour pre-deposit code window. kalshi.com/affiliates returned HTTP 429 on four attempts and we have never read it. The documented partners are CNN, CNBC, FOX, Madison Square Garden, the Chicago Blackhawks and Men in Blazers.

What does a real fill actually look like? A price the book gave you, not the price you had in mind. Take liquidity and you pay the ask plus a taker fee. Rest an order and you pay zero on Polymarket, but you wear adverse selection, because the fastest fills come from whoever knows more than you do. On Kalshi the fee rounds up to the whole cent, so 1 contract at $0.50 costs $0.52.

Can you arbitrage YES and NO for a guaranteed win? Not inside a single Polymarket market. The Gamma API exposes best bid and best ask only, and the NO price is the arithmetic complement of YES, so the two asks sum to $1.00 plus the spread by construction. Our own scanner was built to find that lock and found none. It surfaces momentum instead.

Why would anyone bother staging one? Because the payout shape rewards it. Earning $100 through Polymarket's in-product referral means the referred person pays $1,000 in taker fees inside 30 days, which at $0.50 in a crypto market is $28,571 of notional. The Dub route pays $10 per funded first deposit outright. Volume and signups are the product, not accuracy.

Where to go next

• Maker versus taker, why a free limit order is not a costless one
• Polymarket fees, the category-aware calculator, all eleven rates
• Our trade log, including the position that could not win at entry
• Methodology, what we check, when, and what we refuse to publish