AI hockey betting prompts that start with the goalie
One position decides more of a hockey price than any factor in any other sport. If your prompt does not know who is starting in goal, everything it produces afterwards is decoration.
What a hockey prompt has to get right
Hockey is a low-event sport with the highest single-game variance of the major leagues: the better team loses often enough that a prompt judged on a handful of picks tells you nothing. That is an argument for measuring over a real sample — and for a prompt that outputs a probability rather than a verdict, so you can check calibration instead of just counting wins.
Structurally, three things carry most of the signal: who is in goal, how each side performs on the power play and penalty kill, and whether recent results match the underlying shot and expected-goal share. The v1 prompt below is built on the first two; v2 is built on the third, and is designed to find teams whose record is worse than their play.
Confirmed starting goaltender
Save percentage and workload for the goalie who is actually starting, not the team's average. If the starter is unconfirmed, the correct answer is to wait — tell the prompt that explicitly.
Special teams
Power-play and penalty-kill efficiency plus how many penalties each side takes. Special teams decide a large share of goals in a low-scoring game, and they move totals as much as they move winners.
Shot share versus results
Corsi/Fenwick share and expected goals at even strength. A team controlling play but losing games is the classic hockey value spot; a team winning on unsustainable shooting is the classic trap.
Low-event games and empty nets
Tight defensive matchups favour the under, but late empty-net goals push totals over. A prompt that reasons about totals without mentioning empty nets is missing a real, recurring effect.
Hockey prompts v1 and v2 — and how they differ
The same model, two instruction sets, two different betting personalities. Run both on the same matches; that comparison is the only thing that settles the argument.
| Version | Focus | Style | Best for |
|---|---|---|---|
| v1 | Confirmed starter and power play | Disciplined | Moneyline, unders |
| v2 | Underlying play versus results | Analytics value | Value where results lag play |
You are a hockey betting analyst. Game: {home} vs {away}, {league}, {date}. Line: {odds}.
Start with the confirmed starting goaltenders and their save percentage and recent workload. If a starter is not confirmed, answer "no bet" and stop.
Then weight power-play and penalty-kill efficiency, penalties drawn and taken, and home ice. Low-event matchups favour the under, but account for empty-net goals late.
Output exactly:
1) Moneyline pick + win probability for both sides (sum 100%)
2) Puck-line lean
3) Over/Under total pick against the posted number
4) Confidence 1-10
5) One-line reasoning tied to the market
Do not use certain language: single-game hockey variance is high.
You are an analytics-driven hockey analyst. For {home} vs {away} ({league}, {date}):
Use even-strength shot-attempt share (Corsi/Fenwick), expected goals and finishing over the last 10 games, plus the confirmed goaltenders. Identify whether either team's record is better or worse than its underlying play, and whether the market has priced the record or the play.
Output exactly:
1) Winner + win probability for both sides (sum 100%)
2) Best value market and the edge against {odds}
3) Predicted total goals
4) Confidence 1-10
5) The decisive factor in one line
Flag unsustainable shooting percentages explicitly. If the starter is unconfirmed, answer "no bet".
Placeholders in braces are filled automatically when you run a prompt from a match in the AI Lab. Pasting into your own chat window works too — just replace them by hand.
What to feed the model, and what a usable answer looks like
Feed it this
- Confirmed starting goaltenders with save percentage and recent workload.
- Power-play and penalty-kill percentages, plus penalties drawn and taken.
- Even-strength shot share and expected goals over the last 10 games.
- Rest, travel and back-to-back flags — including whether the backup is likely.
- The line: moneyline, puck line and the total (usually 5.5) with over/under prices.
- Injuries on defence and to top-line forwards.
Good output has
- A moneyline win probability for each side summing to 100%.
- A puck-line lean, stated as a lean rather than a certainty.
- A total pick against the posted number, with empty-net risk acknowledged.
- Confidence 1-10 and one decisive factor.
- A mandatory "no bet" if the starting goaltender is not confirmed.
- No claim of certainty — hockey variance makes confident language a warning sign.
Where hockey prompts usually go wrong
- Predicting before goalie confirmation.
- Reading a win streak built on high shooting percentage as strength.
- Ignoring empty-net goals when picking unders.
- Judging a prompt on 20 picks in the highest-variance sport on the board.
Odds, model context and market drift for each fixture are on the hockey games with odds and AI picks board, so most of the input list above can be copied straight from the match page.
Measure both versions before you trust either
Store both versions
Save v1 and v2 as separate prompts in the AI Lab so every run is attributed to a version instead of blurring together.
Run them on the same matches
Pick fixtures from the hockey board and lock both forecasts before start. Same slate, same information, no hindsight.
Judge on ROI, not hit-rate
A value prompt taking underdogs will always look worse on hit-rate and can still be the profitable one. Settlement and scoring are automatic once the match finishes.
The AI Lab starts on the $19 tier with one sport and five stored prompts, which is enough for a full v1-versus-v2 comparison in hockey. Open a free trial to run it on today's card, or read the prompt library overview for the shared structure behind every sport.
Hockey prompt questions
Why does the hockey prompt insist on a confirmed goalie?
Because the difference between a starter and a backup is worth more than most other factors combined, and lines move sharply when the starter is announced. A prompt that answers anyway will produce a confident number built on the wrong player, which is worse than no answer.
Are unders really the default in low-event games?
They are a bias to test, not a rule to trust. Tight matchups suppress scoring, but empty-net goals and special-team bursts push totals over more often than people expect. Let the prompt state the case and let the AI Lab tell you whether your version of it actually profits.
How large a sample do I need before trusting a hockey prompt?
Larger than you would like. Hockey has the highest single-game variance among the major sports, so a 20-pick sample is close to meaningless. Track it across a season segment and compare ROI against a second version running on the same games.
Prompts for the rest of the board
Find out which hockey prompt actually wins
Start the 5-day AI Lab trial without a card. Bring your own AI key, run v1 and v2 on today's hockey card, and let the dashboard settle it on a virtual $10,000 bank.
AI Lab $19/mo after the trial · cancel anytime · Full Access $49/mo for all sports and the official 7-AI consensus