Our public record
Every pick the PropAnvil app suggested, graded from official box scores. Wins, losses, and what we haven’t proven yet.
Updated Nov 22, 9:23 AM ET · refreshes about every 10 minutes
Our picks, by label
Every Top pick, Strong, Lean and Close call our pick rule gives, logged by the server 3+ hours before the game (nobody chooses what counts) and graded from official box scores, from Nov 6, 9:00 AM ET. “Needs” is the hit rate that pick’s app needs to break even on its cheapest entry. A label needs about 100 graded picks before its record means much.
| Label | Record | Hit | Needs | |
|---|---|---|---|---|
| Top pick | 28–18 | 60.9% | 48.4% | Too few to judge (46) |
| Strong | 168–108 | 60.9% | 49.3% | Ahead of what it needs |
| Lean | 267–245 | 52.1% | 50.5% | About even so far |
| Close call | 353–307 | 53.5% | 55.1% | About even so far |
| All picks, every app | 456–363 | 55.7% | 49.7% | Ahead of what it needs |
| Dabble (payout estimated) | 405–321 | 55.8% | 49.3% | Ahead of what it needs |
| Underdog | 40–26 | 60.6% | 52.3% | Too few to judge (66) |
| PrizePicks | 11–16 | 40.7% | 54.2% | Too few to judge (27) |
| 🔁 Dabble picks, same line on PrizePicks (not added above) | 2–1 | 66.7% | 54.2% | Too few to judge (3) |
| Top pick + Strong since Nov 6 | 211–141 | 59.9% | 49.0% | Ahead of what it needs |
The last row counts only Top picks and Strong picks, the labels we suggest today. The rows above keep every labelled pick from Nov 6.
Picks are research, not bets. Past results don’t promise future ones.
Every day, every pick
Each day’s labelled picks (Top pick, Strong and Lean while it was a label), counted once per player, stat and app, graded after the games by the game’s date. Losing days stay in. “Needs” is what that day’s picks needed to break even; “Total” is the running record.
| Day | Record | Hit | Needs | Total |
|---|---|---|---|---|
| Sat, 11/21 | 9–3 | 75.0% | 49.2% | 456–363 (55.7%) |
| Fri, 11/20 | 9–5 | 64.3% | 47.1% | 447–360 (55.4%) |
| Thu, 11/19 | 13–9 | 59.1% | 49.7% | 438–355 (55.2%) |
| Wed, 11/18 | 34–26 | 56.7% | 48.3% | 425–346 (55.1%) |
| Tue, 11/17 | 44–42 | 51.2% | 48.5% | 391–320 (55.0%) |
| Mon, 11/16 | 34–36 | 48.6% | 50.6% | 347–278 (55.5%) |
| Sun, 11/15 | 88–72 | 55.0% | 52.1% | 313–242 (56.4%) |
| Sat, 11/14 | 15–17 | 46.9% | 49.7% | 225–170 (57.0%) |
| Fri, 11/13 | 26–25 | 51.0% | 49.1% | 210–153 (57.9%) |
| Thu, 11/12 | 26–30 | 46.4% | 50.0% | 184–128 (59.0%) |
| Wed, 11/11 | 39–25 | 60.9% | 49.7% | 158–98 (61.7%) |
| Tue, 11/10 | 33–27 | 55.0% | 48.9% | 119–73 (62.0%) |
| Mon, 11/9 | 37–21 | 63.8% | 49.6% | 86–46 (65.2%) |
| Sun, 11/8 | 19–10 | 65.5% | 50.0% | 49–25 (66.2%) |
| Sat, 11/7 | 14–9 | 60.9% | 48.1% | 30–15 (66.7%) |
| Fri, 11/6 | 16–6 | 72.7% | 45.2% | 16–6 (72.7%) |
Most picks are on Dabble, whose payouts are estimated. Pushes and players who didn’t play are left out. Picks still waiting on a final stat are added when they’re graded.
Beat the closing line
The “closing line” is the betting market’s final price when the game starts, the sharpest public estimate of the true odds. If the market keeps moving toward our picks after we make them, that is the quickest honest sign of a real edge. It shows up in about 100 picks, long before wins and losses settle.
57% of 390 suggested picks beat the close, by +0.4 points of win chance on average.
Measured by our server on every call the model made with an edge (not only picks someone tracked): 57% of 1,960 beat the close.
| When the pick was made | Beat the close | Picks |
|---|---|---|
| 12+ hours before | 54% | 158 |
| 3–12 hours before | 63% | 104 |
| 1–3 hours before | 60% | 35 |
| Under 1 hour before | 81% | 27 |
Buckets under 20 picks are shown as “too few”.
Testers’ tracked picks by app
Pick’em picks the app suggested that testers tracked, graded from official box scores. Each pick counts once, however many people played it. “Needs” is the hit rate that app’s payouts require to break even.
| App | Record | Hit | Needs |
|---|---|---|---|
| Dabbleestimated payoutsAhead so far, not proven | 454–263 | 63.3% | ≈52.0% |
| PrizePicksToo few picks to judge (66) | 30–36 | 45.5% | 53.6% |
| UnderdogToo few picks to judge (46) | 27–19 | 58.7% | 51.7% |
| parlayplayToo few picks to judge (6) | 3–3 | 50.0% | 52.3% |
| DK Pick6Too few picks to judge (2) | 1–1 | 50.0% | 54.9% |
| SleeperToo few picks to judge (2) | 1–1 | 50.0% | 54.9% |
| BetrToo few picks to judge (1) | 1–0 | 100.0% | 54.9% |
Dabble doesn’t publish its payouts in our data feed, so its break-even is our estimate. That’s why its record is kept apart from the headline.
Testers’ tracked picks by sport
| Sport | Record | Hit | Needs |
|---|---|---|---|
| MLBAhead so far, not proven | 374–212 | 63.8% | 54.0% |
| NBAAhead so far, not proven | 67–39 | 63.2% | 44.5% |
| NFLToo few picks to judge (69) | 36–33 | 52.2% | 49.7% |
| NHLToo few picks to judge (47) | 25–22 | 53.2% | 48.3% |
| CFBToo few picks to judge (24) | 11–13 | 45.8% | 53.4% |
| WNBAToo few picks to judge (8) | 4–4 | 50.0% | 54.8% |
Includes every app, Dabble too.
Shadow record: every call, misses included
Since October 2, 2026, the server records every prop the model prices in the hours before its game, including the ones the app doesn’t suggest, and grades all of them from the box score. Nobody picks which calls count.
40,000 calls recorded · 33,558 graded so far
- Pick’em calls the app would suggest (the rule in force at the time: since Oct 21, that entry’s break-even + 3 points on a trusted app with a confirmed payout): 229–178 (56.3% hit vs 57.3% the model said)
- Pick’em calls with an edge the app didn’t suggest: 2741–1621 (62.8% hit vs 61.4% the model said)
- Sportsbook bets with an edge (not suggested right now): 637–1064 (37.4% hit vs 46.7% the model said)
- Game bets (moneylines, spreads, totals): 197 recorded, 28–59 graded
What’s not proven yet
- Profit on apps with known payouts. We call it proven after 500+ graded picks with a win rate clearly above break-even (a 90% confidence test). So far (official, the server’s record of every pick the app suggests): 175 of 500 picks; testers have tracked 123.
- Dabble profit. Its payouts are estimated, so even a good record there can’t prove profit.
- Sportsbook bets. They are down 17.5 units and paused while we review their shadow record.
- Are the win chances honest? Replayed on past games, a sport passes when hits land within 2 points of what the model said: NFL: on target (50.6% vs 52.4% said) · CFB: on target (52.7% vs 52.4% said) · NBA: on target (51.2% vs 52.8% said) · CBB: on target (51.8% vs 52.4% said) · WNBA: off (48.9% vs 52.1% said) · MLB: on target (57.1% vs 57.3% said) · NHL: on target (52.4% vs 52.4% said). This is a test on past games, not live picks.
- Any single pick. Every number here is an estimate. Good picks lose often; a 60% pick misses 4 times in 10.
How we count
- Only picks the app itself suggested count, not picks people chose on their own.
- Each pick counts once, no matter how many people played it.
- Graded from official box scores by our server (1,401 graded that way so far). Pushes and players who didn’t play are left out.
- No names, accounts or personal data are shown here.