AI sports predictions · The open record
Know what the model said.
Before you know who won.
A forecast and a fair price for covered games, compared against the best price across sportsbooks — and locked before the result. The misses stay in the record.
7-day free trial · Everything included · Plans from $59.99/mo
The record
1 hour before · last 7 days
- TENNISTennis796 right · 349 wrong70% right
- SOCCERSoccer226 right · 212 wrong52% right
- NHLNHL20 right · 18 wrong53% right
- NBANBA23 right · 10 wrong70% right
- NFLNFL7 right · 7 wrong14 games so far
- CFBCollege football6 right · 2 wrong8 games so far
- Last 7 days
- 1078–598
- Right
- 64%
Today’s board
1 game forecast todayFirst up 8:00 PM ET
| Time | Matchup | Mkt | Seg | Line | Best | Edge | Str | Band | |||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Upcoming · 1 | |||||||||||
| 8:00 PM | MIL @ TB | ML | FG | — | -136BR | Model columns are in every plan — start a free trial to see them | |||||
Upcoming · 1
- MIL @ TB8:00 PMML · FG
- Best
- -136 BR
- Edge
- Str
- Band
- TRIAL
Forecast accuracy.
How often a frozen forecast named the winner, checkpoint by checkpoint. Accuracy only — no price and no units here; the units record prices each forecast separately.
last 30 days
24 hours out
57.5%
1197 of 2081 graded
60 minutes out
66.8%
4323 of 6468 graded
In play
80.1%
4267 of 5329 graded
An in-play forecast is taken after the game has started and already reflects the score, so it is not comparable to the ones published before it. A hit rate is not a return either: a correct forecast at a short price can still lose money over a season.
Strongest segment · before kickoff
70.8%
3309–1368
TENNIS · 60-minute main
4677 graded
Once the game is under way
84.7% (3525–637) — TENNIS · Live in-game, after set 1 · 4162 graded. A live forecast knows the score, so it is an easier call than a pre-game one. We keep the two records apart rather than blending them into one flattering number.
One slice of the published record, not the overall figure. Every checkpoint and every sport is broken out in full on the complete record, losses included.
The record is open
Every result belongs in the story.
Every published forecast is scored twice: did it name the winner, and what would 1 unit on it have returned at the best sportsbook price when it was published. Both records include the misses. Like any market there are winning and losing stretches; what matters is that the forecast is published before you know which one you are in.
30 days · 60-minute forecasts
Forecast accuracy
4,322–2,145
66.8% over 6,467 graded
7169 recorded forecasts in this cohort.
30 days · 60-minute forecasts at the best price
Forecast units
4,016–2,030
Return -27.16u · -0.4% ROI
6,049 of 6,470 settled forecasts priced at a sportsbook.
Units are hypothetical: each forecast is settled as 1 unit at the best sportsbook price on record when it was published — no wagers were executed.
Read a row
Twelve columns. One question.
Every row on the board answers the same thing: is the price you can get better than the price the model says this is worth? Here is that row, taken apart.
Matchup
BAL @ DET
Mkt
ML · FG
Fair
-148
Best
-118 FD
Edge
+15.8
EV
+0.13
Str
100
Band
GEM
- MatchupThe game
- Away @ home, with the live score once it starts.
- MktThe market
- Which bet this row is about. ML is the moneyline; FG is the full game, F5 the first five innings.
- FairWhat the model thinks it is worth
- The model’s win probability expressed as a price. This is an opinion, not an offer — you cannot bet it.
- BestWhat you can actually get
- The best price across the sportsbooks we read, and which book is showing it. Stale books are excluded.
- EdgeThe gap between the two
- How far the market’s price sits from the model’s, in percentage points. This is the whole reason the row exists.
- EVWhat the gap is worth
- Expected value per $1 at the listed price, if the model is right. Edge says how far apart; EV says how much that distance pays.
- StrOne number, 0 to 100
- Strength compresses the whole row into a single score, so a slate can be scanned down one column.
- BandWhere that number lands
- The named tier for the score. GEM is the top of the ladder and the rarest.
The strength ladder · 0–100
- None
- 0–39
- Small
- 40–59
- Good
- 60–79
- Strong
- 80–99
- Gem
- 100
A higher band means a wider gap between the model and the market — not a higher chance of winning. Most rows on a normal slate land in None or Small, and the board says so rather than manufacturing an edge. Illustrative example
The instruments
Five surfaces. One slate.
Five surfaces, one slate — with each capability's coverage stated sport by sport in the matrix below.
AThe board
One row per market
Every game on the slate, every market on the game, priced against the book. Filter to gems, live or upcoming; sort by strength or by edge.
6 sports
BGame Center
Stay with the game
Observed scores, periods and plays beside the original forecast — so a result never quietly replaces what was said beforehand.
Live scores
CCheckpoints
Published, then frozen
An early read 24 hours out and the main forecast at 60 minutes, calculated independently. Once published, neither is rewritten.
24h · 60m
DSharpDeck AI
The numbers explained
A grounded thesis for the decision, and questions you can ask of it. The figures come from the model — the AI reads them, it does not invent them.
Pro
EThe record
Including the misses
Every forecast scored for accuracy and settled at the best price when it was published, graded from official results, with losses, ties and voids all in view.
Open to all
Not a sportsbook
SharpDeck publishes analysis. It places no wagers and guarantees no outcome.
Coverage, sport by sport
What actually works,
sport by sport.
Six sports, one public record. Every forecast SharpDeck publishes is graded against the result — and NFL's pregame model passed a held-out evaluation on 543 games before it shipped.
How to read the labels — “Validated model” means the model passed a held-out evaluation before release — NFL's pregame model beat its baseline on 543 games it had never seen. “Limited” means the forecast is published and graded in the public record, but not on a model release that passed review; MLB's pregame forecast is the de-vigged market consensus.
- Boards & scores
- Available
- Model forecast
- Validated model
- Live analysis
- Coming soon
- AI analysis
- Coming soon
- Boards & scores
- Available
- Model forecast
- Limited
- Live analysis
- Available
- AI analysis
- Available
- Boards & scores
- Available
- Model forecast
- Limited
- Live analysis
- Coming soon
- AI analysis
- Coming soon
- Boards & scores
- Available
- Model forecast
- Limited
- Live analysis
- Coming soon
- AI analysis
- Coming soon
- Boards & scores
- Available
- Model forecast
- Limited
- Live analysis
- Coming soon
- AI analysis
- Coming soon
- Boards & scores
- Available
- Model forecast
- Limited
- Live analysis
- Coming soon
- AI analysis
- Coming soon
Forecast and price
A forecast is a starting point.
A likely winner and a worthwhile price are different things. SharpDeck shows the forecast, the fair price it implies and the best price across sportsbooks side by side, and records each forecast with the price on offer when it was published.
A forecast has a history
How it works- 1
24 hours before
The early view
A first assessment, recorded with the inputs available then.
- 2
60 minutes before
The main forecast
A separate calculation. The original selection stays locked.
- 3
During the game
The live observation
Updated analysis where supported, kept separate from both originals.
- 4
After the result
The complete record
Each published checkpoint is graded against its own selection.
From forecast to the record
- 01
The forecast
An outcome probability, frozen at its checkpoint.
- 02
The price
The best sportsbook price on record when it was published.
- 03
The result
Graded from the official result and settled as 1 unit at that price.
A process illustration. A forecast with no price on record is still graded, and carries no units.
See the AI work
An analyst that explains the numbers.
SharpDeck AI reads the same model output you see and puts it in plain language — what the edge is, why it exists, and when to pass. The probabilities and prices come from the model; the AI never invents them.
Detroit
Moneyline -118
- Model probability
- 69.1%
- Market probability
- 53.4%
- Edge
- +15.8 pts
- Expected value
- +$0.13 / $1
Why SharpDeck likes it
- Team strength
- Starting pitcher
- Recent form
- Home field
Detroit projects at 69.1% to win, but the market is pricing them near 53%. That gap is the whole story here: the model likes the starter's matchup and Detroit's contact profile against this bullpen, and the recorded price hasn't caught up. The edge is real but not enormous — treat it as a genuine opportunity, not a lock.
Why is the edge here and not on the total?
The edge sits on the moneyline because that's where the model and the recorded price disagree most. On the total, the model's number is within a point of the market, so there's nothing to act on — SharpDeck leaves it as a pass rather than manufacturing a play.
A recorded example of the SharpDeck AI analyst on an MLB game. Available on the Pro plan. Figures are illustrative and not a live price or a recommendation to wager.
Access
Try SharpDeck free for 7 days.
Every plan starts with everything — the live model included. Basic is the model on every game; Pro sees it a day early and shows what moved; Elite keeps forecasting after kickoff.
7-day free trial
Basic
The model on every game, and the record behind it.
$59.99/mo
- Model forecasts across six sports
- Win probability on every game
- The market line, best price and fair price beside every forecast
- Game Center for every covered game
- The full forecast record: every result, and what 1 unit returned
- Watchlist and saved decks
Recommended
Pro
Earlier and deeper: see it a day ahead, and see what moved.
$149.99/mo
- Everything in Basic
- The 24-hour early forecast
- How the forecast and the market moved
- The SharpDeck AI analyst on every MLB game
- Price alerts on any game
7-day free trial
Elite
Live: the model keeps forecasting after kickoff.
$249.99/mo
- Everything in Pro
- The live in-game model
- The full live probability history
- Game intelligence: why we forecast it, the matchup and the head-to-head record
- The daily results email, every morning
Every plan starts with 7 days of everything — the live model included. A card is required to start; cancel before day 7 and you are not charged. After that, cancel any time and your plan runs to the end of the period you paid for. A plan changes how much of the forecast you see, never what it says, and every plan shows the same model labels.
After you sign up
Three steps to your first game.
Choose a plan and start a 7-day free trial of everything. A card is required; cancel before the trial ends and you are not charged.
01
Choose your sport
Six boards — MLB, NFL, CFB, Tennis, Golf, Soccer. You land on the one you used last, on today’s slate.
02
Open a game
Every row opens into the price ladder across sportsbooks, then into the full game terminal with the forecast and its checkpoints.
03
Check the record
The public record is open before you sign up and stays open after. Compare what was published against what happened.
Plain answers
Know what you’re looking at.
Straight answers about access, models and the record.
Read our methodology →How does the 7-day free trial work?
Choose Basic, Pro or Elite and enter a card. For the next 7 days you get everything — the live model, the AI analyst and the early forecasts — whichever plan you chose. On day 8 you move to that plan and are billed for it. Cancel before then and you are not charged anything.
Is a forecast a betting recommendation?
No. A forecast estimates the chance of each outcome — from a statistical model, or for MLB pregame the de-vigged consensus of the betting market. SharpDeck shows it beside the fair price it implies and the best sportsbook price, and records how it did, but it does not tell you what to bet. A forecast whose model has not passed review is labelled as such.
Why do some games have no forecast?
Coverage varies by sport and competition. A model may need inputs, licensed data or validation evidence, and a checkpoint that could not be published on time stays missing. SharpDeck explains the state instead of forcing a forecast.
Can a published forecast change?
The early and main forecasts are independent records. Once published, their selections and probabilities stay fixed. Current live analysis is a separate observation, and a settlement correction never rewrites the original forecast.
Which sports are on SharpDeck?
MLB, NFL, Soccer, CFB, Tennis and Golf have dedicated boards or centers. Score, price, model and live-data availability are verified separately; not every listed competition has a provisioned or approved model.
Are losses included in the record?
Yes. Every graded forecast stays in the record, wins and losses alike. Each is scored for accuracy, and settled as 1 unit at the best sportsbook price when it was published; a forecast with no price on record counts for accuracy only. Ties, voids, pending results and settlement reviews are identified rather than counted as wins.
Why do you ask for a card to start the trial?
So the trial is one per person and continues without interruption if you keep it. The card is handled by Stripe; SharpDeck never sees it. Nothing is charged during the 7 days, you get an email three days before the trial ends, and the app shows exactly when it ends and what you will pay after.
Who can get the free trial?
One free trial per person: a confirmed email address, a card that has not had a SharpDeck trial before, and a browser that has not either. Prepaid cards can subscribe but are not eligible for the free trial. If a trial is not available, the start page says so before you enter a card.
Is there a free plan?
No. Every plan starts with a free 7-day trial of everything, then continues as Basic, Pro or Elite. The public performance record stays readable by anyone, without an account.
What is the difference between Basic, Pro and Elite?
Basic is the forecast on every game: win probability, strength and band, with the line, best price, fair price and edge beside it. Pro adds the forecast a day early, how the forecast and market moved, EV and stake size, the AI analyst and alerts. Elite adds the live forecast after kickoff and its full probability history.
Can I change plans?
Yes, from your subscription page. An upgrade takes effect straight away and is prorated. A downgrade waits for the end of the period you have paid for, so you never lose access you have already paid for.
How do I cancel?
From your subscription page. During the trial, cancelling means you are never charged. After that, a cancellation takes effect at the end of the period you have already paid for.
Does paying change what the model says?
No. A plan changes how much of the forecast you see, never what it says. Every plan sees the same forecasts from the same models, the same model labels and the same published record.
Does SharpDeck place bets or guarantee results?
No. SharpDeck is a sports analytics product, not a sportsbook. It does not place wagers. Sports outcomes are uncertain and past results do not guarantee future outcomes.
Field notes
Guides to reading the numbers.
- GuidesExpected value in sports betting, explained with real numbersWhat EV means, how to calculate it from a probability and a price, and why a +EV bet can still lose.Read the guide →
- GuidesHow to read American odds: minus, plus and what they payA plain guide to moneyline odds: what the minus and plus signs mean, how much you win, and how to convert them.Read the guide →
- GuidesImplied probability: what a betting price believesTurn any price into a probability, see the sportsbook margin hiding in it, and compare it with a forecast.Read the guide →
A clearer view starts here
Get closer
to the game.
Explore the boards, understand the forecasts and follow the record — with everything switched on for your first 7 days.
Card required · nothing charged for 7 days · cancel any time