Counter-Trading: How to Read the Fade Board Without Fooling Yourself
The Fade Board screens high-volume Polymarket wallets that are net down over the window you choose, and shows what they still hold. We tested the inverse of their settled trades and it came out negative, so this is a research screen and not an edge. Here is what it is honestly good for.
What does the Fade Board actually measure?
The Fade Board answers one narrow question: among the wallets doing the most volume on Polymarket right now, which ones are furthest underwater, and what are they still holding? It reads the public volume leaderboard for the window you select — seven days, thirty days, or all time — keeps the wallets that are net down by more than five hundred dollars, ranks the twenty deepest, and then reads each of those wallets' live positions to surface the single largest one that is worth at least five hundred dollars and priced between three and ninety-seven cents.
Each row shows the wallet, its PnL over that window, the position it currently holds with the side and the price, and a link to the opposite side of that position on Polymarket. On a straight binary market the other side of a Yes is a No. On a categorical market it is whatever the venue lists as the opposing outcome, which is usually a field rather than a simple negation, so the board reads that from the venue rather than guessing it. The header totals the combined losses and the combined volume churned by the wallets on screen.
It is Polymarket only. Kalshi publishes no wallet identities, so there is no equivalent board for it, and anything we show from Kalshi elsewhere in the product is anonymous flow rather than a person. Everything on this board is public on-chain data about named addresses.
One honest gap in the data is worth knowing. When a wallet's positions call fails or times out, the row says positions unavailable rather than reporting nothing held, because a dead upstream and a genuinely empty book are different facts and the page refuses to confuse them.
Does fading these wallets make money? No — and here is the test.
This is the section most articles on this topic will not write, so read it before the rest. We ran the obvious backtest: take these wallets' settled trades, invert every one of them, and see what the inverse book returns. It came out negative. Not marginal, not fee-dependent — negative. The Fade Board therefore ships with no claimed edge, and nothing in this playbook should be read as a suggestion that taking the other side of these wallets is profitable.
The reason is a behavioural asymmetry that cuts against the naive fade. High-volume losing wallets tend to realize their winners early and hold their losers to settlement. The trades that end up in their settled record are therefore skewed toward the ones they closed by choice, and inverting a set of trades that was already selected by the trader's own exit behaviour does not give you the mirror image of their PnL. It gives you a different, worse book.
There is a second reason, and it is arithmetic rather than psychology. A negative PnL over a window is a result, not a model of behaviour. It tells you a wallet lost money; it does not tell you the wallet is systematically wrong about price in a way you can exploit. Volume is expensive — spread and fees on a large number of trades will drag a break-even trader deep into the red without any of their reads being wrong. A wallet can appear on this board for being active and unlucky rather than for being bad.
So treat every row as a fact about the past, not a prediction about the position on it. A losing record does not make a wallet's current position wrong, and the page says exactly that under the table. If you want a claim you can lean on, the Sharp Leaderboard and Consensus are built to measure repeatable skill, and this board deliberately is not.
What is the Fade Board honestly good for?
The first use is de-confirmation, and it is the strongest one. You have a thesis and you are about to take a side. Open the Fade Board, switch to the window that matches your holding period, and look at who is stacked on the other side. If the opposing flow is largely wallets that are net down and churning volume, that is weak evidence for your read rather than against it. If the wallets on the other side of your trade are nowhere on this board, you have learned that the side you were about to fade is not obviously weak money — which is worth knowing before you size.
The second use is deflating a big number. A very large open position looks like conviction, and on a thin book it can move price on its own. This board is a standing reminder that a wallet can be enormous and simultaneously down six figures, so size is not information. When you see a big print in the Live Feed and feel the pull to follow it, checking whether the address is on this board costs ten seconds.
The third use is generating research candidates for markets you would otherwise never look at. The positions column surfaces live markets that real money is currently sitting in. Follow the market link, read the resolution criteria, then open the Market Analyzer to pull every signal we hold on that market onto one screen, and let that decide the trade rather than the identity of who is holding it.
What the board is not for is a mechanical rule. There is no version of read the top row, click the other side, size it that survives contact with the test in the previous section. Every row needs its own research, and most rows should end in no trade.
How do you work through a row step by step?
Set the window first, because it changes the population completely. The seven-day view is dominated by wallets having a bad week and includes traders whose thirty-day and all-time records are fine. The all-time view is the closest thing to a durable record and the slowest to change. Match the window to the horizon of the trade you are considering, and be aware that a wallet that is deep red over seven days may be up over the year.
Read the position, not the wallet. Click through to the market on Polymarket and read the resolution criteria and the current book before you form any opinion about the trade. A position that looks foolish at eighty cents may be a hedge against something the wallet holds elsewhere, or the second leg of a spread you cannot see from one row, or a market maker's inventory. The board shows one position per wallet — the largest — so you are looking at a fragment of a book by construction.
Cross-check the other side. If you are seriously considering the opposite outcome, look at whether Consensus shows tracked sharp capital leaning that way, whether Insider Radar has flagged a fresh concentrated wallet in the same market, and whether the Arbitrage Scanner shows the same event priced differently on Kalshi. If the answer to all three is no, the only thing supporting the trade is that someone with a bad month is on the other side, and section two explains why that is not enough.
Then check the clock. The Event Calendar shows what resolves when across both venues, and a position that is a week from settlement is a very different trade from one with three months of carry, regardless of the price.
How should you size if you take the other side anyway?
Each row carries a link into the Kelly Bankroll Calculator, and the way it is wired is a deliberate statement about the strength of the signal. It opens the calculator at the price of the other side of the position — one minus the holder's current price — with the probability field seeded to exactly that same number. Zero assumed edge. The calculator will answer no trade until you type in a probability you can actually defend, because the fade signal carries no modelled probability and the tool refuses to pretend it does.
That is the discipline in miniature. If you cannot state, in a sentence, why the true probability differs from the market price, there is no trade here — there is only a stranger's bad month. When you can state it, the calculator prices in the venue fee, applies your chosen Kelly fraction, and caps the stake at the per-market maximum you set. Use a quarter or a half. Full Kelly assumes your probability is exactly right, and on a screen with no tested edge behind it, that assumption is unearned.
Run the sensitivity ladder before you commit. It re-prices the same trade at your estimate plus or minus a few points, and if two points of wobble flips it to no trade, the position is a coin flip with extra steps. The variance check below it simulates repeating the trade at your current sizing so you can see the spread of outcomes — including the risk of losing half the bankroll — that any edge, real or imagined, actually produces.
Cap correlation and cap the category. If three of your contrarian positions all resolve on the same event, they are one position. And never commit capital you cannot afford to lose: prediction market shares settle at zero as routinely as they settle at a dollar.
Where counter-trading goes wrong
The largest failure is the one this whole article is built around: assuming a losing record transfers to the next position. It does not, we measured it, and the measurement was negative. Every other failure mode below is downstream of that one.
The second failure is mistaking informed flow for weak flow. A cluster of buying that looks like a crowd can be early, correct positioning ahead of information the price has not caught up to. Insider Radar exists partly for this — it surfaces fresh wallets taking large concentrated positions — and when it lights up on the side you were about to trade against, that is a reason to pause rather than press.
The third is category error. In markets with a knowable numerical input, the flow you would be fading may simply be reading the input correctly. Kalshi's daily temperature markets are the clearest case, and the Weather Edge finder is the right tool there: it prices each bucket off a forecast rather than off anybody's record. Fading a wallet that happens to have read the forecast right is not contrarian, it is just wrong.
The rest are structural. Thin books and wide spreads eat any thin edge before settlement. Correlated positions all lose together when one macro event moves a whole cluster of markets. Small samples lie, so a wallet's weak window may be variance. And the board itself only reads the largest position per wallet from a leaderboard of the top fifty by volume, so it is a screen over a slice of the venue, not a census of it. Nothing here is financial advice, nothing here promises profit, and past performance does not guarantee future results.
Educational content, not financial advice. Past performance does not guarantee future results.