Badge rules · updated Oct 14, 2026, 7:15 a.m. ET
How badges work
In short
A badge is a short fact about a forecaster's past picks, given by a published rule to everyone the same way. It is not a recommendation and says nothing about the next pick.
Badges are facts about past picks, given by published rules.
The rules
- The catalog defines 64 kinds of badges; 31 are shown now. Each kind has a published rule, a minimum number of decided picks and a window: one NFL week, the season, or the live record since Week 6.
- Badges are computed for every tracked forecaster at once, with every data refresh. Nobody gets a badge by hand, and a new kind is never written to fit one person.
- A share is wins divided by wins plus losses; pushes are left out. A badge that compares a forecaster with the field also needs a share above the field average and more wins than losses.
- Any known loss ends a run of correct picks, including a pick read after the game. Games that start at the same time count as one block, because their order is unknown.
- Week badges are given after the week's last game and kept with their date.
- Badges built on at least 80% of the same picks are merged and shown once; the stronger one stays.
- A profile shows up to 3 badges under Standing: the best level in the largest field, the rarest one by our chance check, and a week badge or the line of picks recorded before kickoff. Every other badge is under All badges, and each one opens with the picks behind it, losses included.
- Outlet badges describe a panel as a whole and are kept apart from the badges of people.
- Badges are plain text. We do not mark them up as ratings or awards for search engines.
Every kind with its rule, holders and chance check: all badges.
Where pick times come from
Every badge is labeled with where the times of its picks come from. A badge can carry two labels when its picks come from two sources.
- Recorded before kickoff
- We saw the pick on the outlet's page before kickoff. This is the live record. Can give: Every badge, including levels (No.1 to Top 10) and the Verified line.
- Timestamped before kickoff (platform)
- We read the pick after the game, but the platform it was made on (Tallysight) records the time of each pick, and that time is before kickoff. Can give: Fact badges and counts, marked with this label. No levels.
- Timestamped before kickoff (web archive copy)
- A web archive copy of the outlet's page, made before kickoff, shows the same pick. Can give: Fact badges and counts, marked with this label. No levels.
- Matches the outlet's published record
- ESPN, CBS Sports and NFL.com publish a running record for each of their experts. Where our recount of the picks we read after the game matches that published record (for that expert and week; for ESPN, the season so far), those picks count. Can give: Fact badges only, marked with this label. No levels.
- read after the game
- Any other pick we read from an outlet's page after the game, with no time stamp and no published record to check against. Can give: No badges. It adds to the total but never counts as a win, and a loss ends a run.
Levels use the live record only, counted from Week 6 of 2026, the first week we recorded in full before kickoff. A pick read after the game still counts against the forecaster in a level: it adds to the total but never as a win.
No.1, Top 3, Top 5, Top 10
Within a category (for example, against-the-spread picks in the live record, or picks in one team's games), forecasters with enough decided picks are ordered by the lower bound of the 95% Wilson interval of their share of picks right. This accounts for uncertainty in short records. The size of the field sets how far down the ladder goes:
| Level | Field of at least |
|---|---|
| No.1 | 5 forecasters |
| Top 3 | 8 forecasters |
| Top 5 | 12 forecasters |
| Top 10 | 20 forecasters |
- A level always shows the size of the field, "of M":
No.1 · ATS picks · 2026, live record since Week 6 · of 12. - Up to three tied at the top share Joint No.1. Four or more tied at the top get "Shared best record · k of M tied", and only if they are no more than a quarter of the field; otherwise nobody gets the level.
- Team badges have only No.1 and Top 3. A week has one No.1 per market (or one shared top record).
Words we use
- Verified
- Only in the line "Verified · N picks recorded before kickoff": we recorded these picks on the outlet's page before kickoff. It says nothing about results, and a pick we did not see before kickoff never carries it.
- Timestamped
- The line "N picks timestamped before kickoff": the time before kickoff comes from the publishing platform or a web archive copy, and we saw the picks later.
- No.1, Joint No.1, Top 3, Top 5, Top 10
- A place in a field of M forecasters, by the ladder above, always with "of M".
- Best
- Only in "Best … in a category · period · of M", for a No.1 in the whole field of that category, and in "Best call": the correct straight-up pick with the lowest market chance in a forecaster's record, shown as a profile line. Never "best" without a category.
- Shared best record
- Four or more forecasters tied at the top, no more than a quarter of the field: "Shared best record · k of M tied".
- Upset Specialist
- Only together with its numbers: "Upset Specialist · x of N underdogs won outright · prices implied y". It needs at least 5 straight-up picks on underdogs, at least 3 of them won, and at least 1.0 more wins than the market prices implied.
- in a row
- Picks right in a row on one market, in kickoff order, shown with the market odds of the run.
- Called it
- On a game page after the game: the forecaster picked the underdog to win straight up, and it did.
- market-adjusted points
- Each pick counted as a 1-point play at the no-margin market price at the time of the pick (for picks added after the game, the price at the start of that NFL week): a win adds 1/p − 1, a loss takes away 1, a push counts 0. Always shown with its sign and the number of picks. These points measure pick results; they do not represent betting returns.
- over picking favorites
- Only in "SU W–L · +d over picking favorites (Wf–Lf)": the record against picking the favorite in every one of the same games, at the same prices.
- Positive points … in k of m weeks
- NFL weeks with five or more scored picks in which the points were above zero.
- Line moved toward the pick
- Only with "k of n": after the pick, the market line moved toward the side picked before kickoff. It says nothing about the result.
- Leads the … panel
- The most market-adjusted points within one outlet's panel, live record, always with "of M".
- Most underdog picks
- Always together with how those picks did; shown only when their points are zero or more.
No badge calls anyone an expert, a guru, a genius or a sharp, and none speaks of money or of the next game.
How often luck alone gives a badge
We test the catalog against chance. The check keeps every forecaster's real picks, draws every game's final score again at random around the market line, 40 times over, and counts the badges each of those replayed seasons would give. It was run on the picks graded by Oct 13, 2026.
- By the rules at the time (picks recorded or timestamped before kickoff): 29 of 41 forecasters with five or more usable graded picks held at least one badge for results. In the chance check, 34.1 did on average.
- With every pick we hold, including those read after the game: 63 of 82 held one, against 71.4 by chance; 41 reached a level (No.1 to Top 10), against 34.3 by chance.
Luck alone hands out about as many badges as the real record does. That is why a badge is worded as a fact about past picks ("did this"), never as a skill ("is the best at this"), and why each badge page shows its own numbers under How common.
Badges of each kind: how many the real record gave, and how many luck gave on average over 40 replayed seasons. Every pick we held at the time of the check counts here, including picks read after the game.
| Badge | Real record | By chance |
|---|---|---|
| Straight-up picks standing | 11 | 10.5 |
| Picks against the spread standing | 1 | 4.8 |
| No.1 of the week | 14 | 17.1 |
| Strong week | 6 | 17.9 |
| Biggest upset called | 26 | 19.2 |
| Correct run | 6 | 7.1 |
| Rare run | 2 | 4.8 |
| Upset calls | 33 | 44.7 |
| Upset Specialist (underdogs that won outright) | 8 | 11.8 |
| Underdogs against the spread | 5 | 4.7 |
| Division games | 17 | 19.2 |
| Prime-time games | 6 | 20.4 |
| Team game record | 6 | 11 |
| Right against the crowd | 66 | 58.9 |
| More right than the market implied | 2 | 8.3 |
| Panel standing | 47 | 42.7 |
| Exact final score | 1 | 0.7 |
| Panel record when most agreed | 1 | 0.1 |
| Best call | 60 | 68.8 |
| All of these | 318 | 372.7 |
Badges added in the October 2026 update (market-adjusted points; favorites, form and panels), checked the same way: 40 replayed seasons, picks recorded or timestamped before kickoff, archive records and checked outlet records; places use the live record only. A dash: nobody qualifies yet, because the live record with prices starts in Week 6.
| Badge | Qualifying | Real record | By chance |
|---|---|---|---|
| Points standing | — | 0 | — |
| Market-adjusted points | 57 | 17 | 19 |
| Better than picking favorites | 23 | 4 | 5.6 |
| Line moved toward the pick | — | 0 | — |
| Consistent weeks | 35 | 4 | 6.5 |
| Recent form | 35 | 4 | 11.6 |
| Panel points leader | — | 0 | — |
| Most underdog picks | — | 0 | — |
Do picks within a titled area do better?
This is the test of the badges themselves. A pick is "within a titled area" when, at the moment it was made, the forecaster already held a badge in that pick's category: the same market, an underdog or favorite pick, a division or prime-time game, one of the two teams, or market-adjusted points, a record better than picking every favorite or recent form in that market. These are the badges shown next to picks on game pages, up to two per pick. We compare how often those picks were right with all other picks.
Live record only (picks we recorded before kickoff), games in the 90 days to Oct 14, 2026. We draw a conclusion only when each group has at least 30 graded picks.
Not enough live picks yet (N=99). Within a titled area: 28.6% right (2 of 7; 95% range 8.2–64.1%; the market expected 56.4%). Other picks: 57.6% right (53 of 92; 95% range 47.4–67.2%; the market expected 49.9%). 142 live picks waiting for their games.
By badge (live record; picks with that badge against picks with no badge):
- Better than picking favorites: 33.3% right (1 of 3; 95% range 6.2–79.2%; the market expected 53.8%); too few picks to compare yet.
- Prime-time games: 0.0% right (0 of 2; 95% range 0.0–65.8%; the market expected 61.2%); too few picks to compare yet.
- Recent form: 0.0% right (0 of 2; 95% range 0.0–65.8%; the market expected 61.2%); too few picks to compare yet.
- Market-adjusted points: 50.0% right (1 of 2; 95% range 9.4–90.5%; the market expected 55.6%); too few picks to compare yet.
For reference only, not the live record: picks timestamped by the publishing platform, archive copies and picks read after the game, same window. For a pick read after the game the moment of the pick is taken as the start of that NFL week.
- Within a titled area: 54.1% right (417 of 771; 95% range 50.6–57.6%; the market expected 60.8%).
- Other picks: 55.4% right (2,354 of 4,246; 95% range 53.9–56.9%; the market expected 56.0%).
- Difference: −1.4 points (95% range −5.2 to +2.4).
On this reference count there is no clear difference.
By badge (reference count; picks with that badge against picks with no badge):
- Market-adjusted points: 54.3% right (355 of 654; 95% range 50.4–58.1%; the market expected 61.0%); no clear difference from picks with no badge: −1.2 points (95% range −5.3 to +2.9).
- Better than picking favorites: 62.2% right (51 of 82; 95% range 51.4–71.9%; the market expected 62.8%); no clear difference from picks with no badge: +6.8 points (95% range −4.2 to +16.6).
- Division games: 43.8% right (32 of 73; 95% range 33.1–55.2%; the market expected 55.1%); did worse than picks with no badge: −11.6 points (95% range −22.5 to −0.1).
- Prime-time games: 14.7% right (5 of 34; 95% range 6.5–30.1%; the market expected 55.4%); did worse than picks with no badge: −40.7 points (95% range −49.1 to −25.2).
- Underdogs against the spread: 55.6% right (10 of 18; 95% range 33.7–75.4%; the market expected 50.0%); too few picks to compare yet.
- Upset Specialist (underdogs that won outright): 33.3% right (2 of 6; 95% range 9.7–70.0%; the market expected 43.5%); too few picks to compare yet.
The market expectation is the average chance the market gave those picks: from the moneyline without the bookmaker margin straight up, one half against the spread and on totals. The ranges are 95% Wilson intervals, and Newcombe's method for the difference. Picks on the same game or by the same person are not independent, so the true range is wider than shown. We publish this check whatever it shows: if titled areas do no better on the live record, we say so here and review the catalog. Updated Oct 14, 2026, 7:15 a.m. ET.
No hindsight
Next to a pick on a game page we show only badges the forecaster already held when the pick was made, counted from games whose result was known by then (four hours after kickoff). For a pick we read after the game, the moment is the start of that NFL week. Badges shown before kickoff are saved with the pre-game version of the game page and are not changed afterwards.
History and corrections
Badges are recounted from the saved picks and final scores with every data refresh. When a badge ends or changes, it moves to History on the holder's page, with the date and the reason in plain words: it no longer meets the rule, a data correction changed the picks behind it, or the catalog changed. A badge merged into a stronger one built on the same picks has not ended. It is named inside the stronger badge and is not listed under History.
Week badges are kept with their date. If a data correction or picks added later for that week change who was first, the badge goes to the new holder and the earlier one is listed under History as revised: a week never shows two different No.1 records. Errors in our data are listed on the corrections page.
All badges · New badges this week · How we record and grade picks