How to read a wallet's recent form: game calls, expected hits, and why a hot streak is not in the score
A wallet's recent form is its record on game moneylines over the last 14 days — 9–3 on MLB moneylines, 12 settled, 7.1 expected — printed on every verdict page and in Scout's "Form (14d)" column. It exists because the CopyGrade Score is structurally blind to a sports hot streak, and it is deliberately not a score input, because a hot fortnight among hundreds of wallets happens by chance. This guide defines what a call is, states every threshold, explains the one number that makes a record readable — the hits the prices paid predicted — and shows how often a 9–3 happens to a coin.
Why can't the score see a hot streak?
Because the score's profit-and-loss engine closes a position only when the wallet sells. That is the right rule for most of Polymarket — a trader who buys at 40¢ and sells at 55¢ has banked a round trip, and a sequence of round trips is what edge, drawdown and consistency are measured on. Sports wallets do something else: they buy a side before the game, hold through the final, and redeem the winning shares at $1. No sell ever happens, so no round trip is ever recorded, and a wallet that has quietly hit fifteen of twenty moneylines this month shows the score exactly nothing.
Stored positions cannot fill the gap either. Once a winning position is redeemed it leaves the wallet's Polymarket book entirely, while a losing position stays on it at $0 until the wallet bothers to clear it. Read a sports wallet's open positions and you see its losers and none of its winners — a survivorship bias pointing the wrong way. The only unbiased path is the one the form uses: take the wallet's pre-kickoff position straight from the trade tape, then settle it against the game's actual winner.
None of this makes the score wrong about sports wallets; it makes the score silent about the part of their record that lives between kickoff and redemption. The farming check, the edge estimate and the drawdown history still apply in full. Form is the extra column, not a replacement grade.
What counts as a call?
A call is the side a wallet was net long on a game's moneyline at kickoff: its pre-kickoff buys minus its pre-kickoff sells on each outcome, the larger net side winning the row. It is the Graded Slate's rule applied to one wallet instead of a cohort, and every part of it is deliberate.
| Rule | Threshold | Why |
|---|---|---|
| Market | The game's moneyline only | One binary question per game, one settlement — a hit means the same thing in every row |
| Timing | Fills before the scheduled kickoff only | Pre-kickoff buys are the only fills a copier could plausibly have mirrored at a comparable price; an in-play exit is a different bet |
| Size | Net position of $250 or more at the prices paid | A $40 flutter is not a position; the floor is the slate's own |
| Leagues | NFL, college football, MLB — the leagues with a slug-matched moneyline per game | Where a game, a market and a final can be tied together without guessing |
| Window | Trailing 14 days | Two NFL slates, a dozen MLB game days — enough calls to read, few enough that the record is recent |
| Settlement | The game's final; pending until one is readable | A postponed or voided game is neither a hit nor a miss |
A wallet that traded a game's spread or total but not its moneyline made no call. A wallet that bought both sides and ended flat made no call. A wallet that bought the underdog at 35¢ and sold it at 48¢ before first pitch made no call either: it traded the game, it did not hold a view into it.
Once a call settles, the ledger never rewrites it. A wallet that is re-graded next week keeps last week's calls exactly as they were.
What does "expected hits" mean, and why is it the number that matters?
Expected hits is the sum of the prices the wallet paid. A moneyline price is the market's probability: a side bought at 64¢ was expected to win 64% of the time. Add those prices up across a wallet's settled calls and you have how many hits the market itself predicted — the base rate the record must be read against, instead of 50%.
Two wallets, both 8–2:
| Calls | Average price | Expected hits | Hits over expected | |
|---|---|---|---|---|
| Wallet A — bought favourites | 10 | 80¢ | 8.0 | 0.0 |
| Wallet B — bought coin-flips | 10 | 45¢ | 4.5 | +3.5 |
Wallet A did precisely what the market told it to. Wallet B beat the prices by three and a half games, which is the beginning of a signal. Both print the same 80% hit rate, and only the expected-hits line separates them. This is the sports version of the lesson every data cut on this site keeps landing on: a win rate is not an edge — a record counts signs, and edge lives in the price.
The same arithmetic protects you in the other direction. A wallet at 4–6 on underdogs bought at an average of 30¢ was expected to hit three; it beat its prices by one and made money doing it. Its hit rate looks awful and its form is fine.
What is the hot bar?
A form reads as hot — and can raise the opt-in hot-streak alert — only when every one of four gates clears, all of them on the settled sample (pending calls neither help nor hurt):
| Gate | Threshold | What it rules out |
|---|---|---|
| Sample | At least 8 settled calls | A 3–0 |
| Hit rate | 70% or better | A record that is merely above average |
| Result | Net positive after settlement | A 70% hit rate on heavy favourites that still lost money |
| Against the prices | Hits at least 1.5 above expected | Ten −400 favourites that hit eight times — what the market expected |
Under five settled calls a form is labelled thin and is never hot, whatever it reads.
The alert then applies three gates the form cannot, because they read the wallet rather than the record: a clean farming check, a CopyGrade Score of 65 or higher (the Candidate band), and at least 30 days under observation. A hot fortnight on a farming-flagged wallet, a low-band wallet, or a wallet first seen last week is not a notification — it is precisely the case where a hot record was most likely manufactured, cherry-picked, or simply lucky. And like every notification on this site it is off until you switch it on, under Account → Notifications; nothing is ever turned on for an account that did not turn it on itself.
How often does a coin go 8–2?
Often enough that a hot record on its own is a reading, not a finding. If a wallet's calls carried no information at all — a fair coin on every game — the chance of clearing each bar by luck alone is:
| Record | Chance for a coin |
|---|---|
| 6 of 8 or better (a 75% hit rate at the sample floor) | 14.5% (about 1 in 7) |
| 7 of 10 or better | 17.2% (about 1 in 6) |
| 8 of 10 or better | 5.5% (about 1 in 18) |
| 10 of 14 or better | 9.0% (about 1 in 11) |
Now multiply by the number of wallets being watched. On 2026-09-18, three weeks into the ledger, 239 graded wallets had at least one settled call of $250 or more in the trailing 14 days; 67 had the five that make a form readable, and 43 had the eight the hot bar requires. At the sample floor, chance alone puts about one wallet in seven over 70% — call it six of the 43. The observed count was ten. Seven of those ten were also net positive, and exactly three cleared the prices-paid margin as well — two on MLB, one on college football. All three carried a farming-risk flag and scored below the Candidate band, so on that date no wallet on the site qualified for the hot-streak alert at all.
That last figure is the point of the fourth gate. Of the 31 MLB wallets with eight or more settled calls, eight beat their prices by 1.5 hits or more, but only four had a hit rate of 70% — the two halves of the bar catch different wallets, and a wallet that clears both is rarer than either. A hit-rate screen alone would have surfaced favourite-buyers; a beat-the-prices screen alone would have surfaced underdog-buyers on a good week. Requiring both, plus a positive result, is how the bar stays honest with a small sample. (Those counts are one dated read of a ledger that recomputes twice a day, and the trailing window moves with it.)
What does a whole population's form look like?
Aggregated, the ledger answers a question no single form can: do graded wallets, as a group, beat the prices they pay on game moneylines? Over its first three weeks — 2026-08-27 to 2026-09-18 — the MLB ledger held 1,429 settled calls of $250 or more by 140 graded wallets across 287 games. Those calls hit 727 times against 747.7 expected at the prices paid: about twenty-one games short of the market's own forecast, on an average buy price of 52¢. Split by the price paid:
| Price paid | Calls | Hit rate | Average price | Hits vs expected |
|---|---|---|---|---|
| Under 40¢ (underdogs) | 152 | 28.3% | 35.4¢ | 43 vs 53.8 |
| 40–60¢ (near coin-flips) | 900 | 47.3% | 50.1¢ | 426 vs 451.0 |
| 60¢ and above (favourites) | 377 | 68.4% | 64.4¢ | 258 vs 242.9 |
The graded population's underdog calls hit less often than their prices implied, its favourite calls hit slightly more often, and its coin-flip calls — nearly two-thirds of the ledger — came in a little under. That is what a population with no net information over the closing price looks like once you read it against the prices instead of against 50%, and it is why an individual form is presented beside its expected-hits line rather than beside a hit rate. The MLB board cut carries the same ledger split by farming cohort, and the dollar side of it.
How to use a form
- Read it against expected hits, never against 50%. 9–3, 7.1 expected is two games of signal; 9–3, 8.8 expected is a wallet that bought favourites.
- Check the sample before the record. Under five settled calls the card says thin; under eight nothing is hot. A record built over a fortnight is one reading — the season-versus-bracket guide has the error bars at every sample size.
- Treat it as a reason to open the verdict, not as a verdict. Form does not move the score, and the score's gates — the farming check above all — still decide whether a wallet belongs on a shortlist. A hot flagged wallet is still a flagged wallet.
- Open the calls. The verdict page lists each call with its league, date, side, price and result, so a 9–3 can be read game by game: which side, at what price, against what outcome. A record made of 80¢ favourites reads differently from one made of 45¢ coin-flips, and the list is where you find out.
- Let the alert do the watching, if you want it. The hot-streak notification fires weekly per wallet while a form stays hot, and only for wallets that also clear the farming, band and tenure gates. It is opt-in and Pro; what it delivers is a prompt to look, not a recommendation.
Form is display, not score. Nothing on the card feeds the CopyGrade Score, every surface that renders it says so, and the public wallet page shows the record and the expected-hits line without stakes or dollar results. CopyGrade is analysis-only — it never executes trades, holds funds, or custodies keys. A CopyGrade Score is a documented research opinion, and a farming-risk flag is an algorithmic risk assessment, not an accusation of conduct. Not financial advice.