After an NHL Starting Goalie Is Confirmed, How Long Until the Odds Move?
A study of “goalie confirmations”: after reporters confirm who will start in goal tonight, do the odds on Polymarket change, and how long does the change last? (The first stop of our starting-lineup research.)
Study details
On this page
Key findings
"A fifth" means 18% of the total 3-hour move, not a 20% change in the odds. Example: if the odds slowly fall from 2.000 to 1.994 over 3 hours, they are only at about 1.999 after 15 minutes and about 1.997 after an hour. Added after the main results: average of 889 confirmations (from 689 games); the range is wide (−9% to 43%), and in the first 15 minutes the odds may even briefly move the other way.
- Looking afterwards only at earlier confirmations that can be followed for a full 3 hours: about halfway after an hour on average, and possibly still moving after 3 hours.
- The change is very small: over 3 hours, the win chance implied by the odds moves only about 0.15 percentage points on average, for example from 50% to 50.15%.
- Surprise goalies (not the one people expected) move the odds more at first.
- For our models, what matters is estimating the gap between the two goalies early, not racing the reporter’s post by a few minutes.
In plain words
What do we want to know?
In ice hockey, the starting goalie is the player in goal when the game begins. If the main goalie is swapped for a backup, the team’s chance of winning may drop. Reporters who cover the team usually post a few hours before the game to confirm who will start. We want to know: after the confirmation comes out, do the odds on Polymarket change, and how long does the change last?
How do we study it?
Polymarket is a prediction platform where people buy and sell shares in game results. Take “Team A wins tonight”: the price is between $0 and $1. If Team A wins, each share pays back $1; if they lose, it pays $0. So a price of $0.60 roughly means the people trading think, as a group, that Team A has about a 60% chance to win. That is the crowd’s view, not necessarily the true chance; when news changes that view, the odds change too. A Polymarket price converts directly to odds: a price of $0.60 means about a 60% win chance and odds of about 1.67 (odds = 1 ÷ price). We collected 1,801 goalie confirmations from the 2025-26 season (598 of them were not the expected goalie). We used the time of the reporter’s post as the time the news came out (for the few without a tweet link, we used the time logged on the website), recorded the odds every 5 minutes, and looked at how the odds moved before and after each confirmation.
What have we found?
In this data, the odds barely move at the moment the goalie is confirmed, and only change slowly afterwards. Looking only at confirmations made at least 3 hours 5 minutes before the game: about a fifth of the change happens in the first 15 minutes, and half of the change over the first 3 hours is reached after about 1 hour. How long it takes to finish completely can’t be estimated well this time; the odds may still be moving after 3 hours. The total change is small.
How small is it?
The average change over 3 hours is about 0.15 percentage points. As a number example, that is like going from 50% to 50.15%, not to 65%. This only shows the size; it is not the real change in any game. For our models, what matters about goalie information is estimating the strength gap between the two goalies accurately and putting it in early, not racing the reporter’s post by a few minutes.
What can’t we say yet?
“Halfway in about 1 hour” comes from the average path of 889 confirmations made at least 3 hours before the game. The range is 30 minutes to 2 hours, and it doesn’t mean every game is the same. About 36% of games had the goalie confirmed only in the last half hour before the game (during warm-ups); that is too short to measure “how long.” How often we could find confirmation times varies by team. The study also did not calculate real trading costs.
Terms used on this page
- Polymarket
- A prediction platform where people buy and sell shares in game results. Prices run from $0 to $1 and roughly show the chance the traders think an outcome has — not necessarily the true chance. A price converts directly to odds: odds = 1 ÷ price.
- Starting goalie
- The player in goal when the game begins.
- Surprise
- The confirmed goalie is not the one expected by our pre-game rule — for example, everyone thought the main goalie would start, but it turned out to be the backup.
- Percentage point
- The gap between two percentages. For example, going from 50% to 51% is a rise of 1 percentage point.
- Toward the news
- If news makes a team more likely to win (for example, a stronger goalie is confirmed to start), its odds go down (the implied win chance goes up). That counts as “toward the news” and is recorded as a positive number.
- Sample
- The set of news items or games this study actually analyzed. “Supported in this sample” means it held in this data and still needs checking on new data.
Detailed results
Which times should you compare?
- 01Goalie confirmed
Start at the time the starting goalie is confirmed.
- 0215 minutes later
Measure the initial change in odds.
- 031 hour later
Check whether changes continue; this is not a fixed rule for every game.
- 043 hours later
Compare the same confirmations with a complete observation window.
The timing curve uses only confirmations at least 3 hours 5 minutes before the game. Odds may keep changing afterwards; the final point does not mean the reaction is complete.
How the team’s win chance moved after a goalie confirmation, on average (implied by the odds)
Only the 889 confirmations (from 689 games) made at least 3 hours 5 minutes before the game, so every point uses the same set of confirmations and isn’t cut short by the game starting. The green band is the first 15 minutes after the confirmation: only a little happens there, and the line rises slowly after that.
A description added after seeing the results｜The ranges are wide (for example, 0.03–0.15 percentage points at 1 hour)｜2.5 hours being slightly higher than 3 hours is within the range｜Games where the goalie was confirmed only in the last half hour before the game (about 36%) are too close to the start and are not in this chart.
View detailed chart values
How much the team’s win chance has changed in total at each point after a goalie confirmation (implied by the odds)
The same 889 confirmations (from 689 games, all made at least 3 hours 5 minutes before the game). Unit: percentage points; the brackets show the share of the change over the first 3 hours.
A description added after seeing the results (the original plan only split the time into “the first 15 minutes” and “after that until the game”)｜The share in brackets = how much of the total change over the first 3 hours｜The uncertainty is large: the share at 1 hour could be anywhere from 27% to 104%｜Confirmation times fall into two groups: about 36% come within half an hour before the game (during warm-ups), and about 49% come more than 3 hours before; only the second group can measure “how long.”
HypothesesFor deeper reading
- H1The odds don’t finish adjusting all at once → holds: from 15 minutes after the confirmation until at most 3 hours after it (or 5 minutes before the game), they still move a tiny bit further the same way on average (about 0.08 percentage points of implied win chance); the result is the same with several different data filtersSupported in this sample
- H2Games with fewer people betting are more likely to lag → not enough evidence, and the direction is the opposite; can’t tellInsufficient evidence
- H3After surprise goalie news, the odds are more likely to keep moving → right direction, but under a stricter standard, there isn’t enough evidence to tellInsufficient evidence
- H4The odds move before surprise goalie news comes out → not enough evidence to tell (on average, they barely move in the 1 hour before)Insufficient evidence
- H5(Exploratory) Compared with R1’s NBA injury news → the “lag” after the news is small and about the same for both; looking only at news at least 3 hours 5 minutes before the game, NBA news finishes about 90% of its change within the first 15 minutes, while NHL goalie news finishes only about 20%. This is just a side-by-side description, not a formal comparisonExploratory
Background and why it matters
Starting goalies and starting pitchers have a big effect on who wins, and the time they are announced is clear. Comparing with R1 shows whether the way the odds react to lineup news is special to the NBA or the same across sports.
Latest progress
Latest：Initial analysis done; data filtering method corrected
Next step：Check again on new games from the 2026-27 season; we compare the answers in mid-February 2027
Study timeline
- Research direction set2026-10-20
- Plan finalized and timestamped2026-10-20
- Collecting goalie confirmation times and odds (2025-26 regular season)2026-10-20
- Main analysis done2026-10-20
- Checking again on new 2026-27 games: plan finalized, collecting data every day (verdict on 2027/2/15)In progress2026-10-20
Methods
What we want to know
- Once a goalie is confirmed, do the odds change all at once, change slowly, or go too far and then come back?
- After surprise news (not the goalie people expected), do the odds tend to keep moving more?
- Do the odds move before the news comes out?
Method summary
- One piece of “goalie news” = a reporter who covers the team posting to confirm who will start in goal for that team tonight; the post time is accurate to the millisecond.
- The rule for the “expected starter”: if the team played the day before and the main goalie played → we expect the backup; otherwise → we expect the goalie who has started the most this season. If the confirmed goalie is different from the expected one, it counts as a “surprise.”
- The main test: from 15 minutes to 3 hours after the confirmation (or until 5 minutes before the game), do the odds keep moving toward the news?
- Goalie strength is compared using the NHL’s official save percentage (the share of shots the goalie stops).
- News time: mostly the time of the reporter’s tweet. Of all 2,138 confirmations we found a time for, 124 had no tweet link, so we used the time logged on the website instead (about 2 minutes later on average); 95 of the 1,801 confirmations in the analysis are like this. Recalculating with tweet times only gives the same conclusion.
- A description added after seeing the results: “how long” is recomputed using only the 889 confirmations (from 689 games) made at least 3 hours 5 minutes before the game, so every time point uses the same set of games and isn’t cut short by the game starting. The range is estimated by “re-drawing the games many times.”
Data and limitations
Data we use (all public and free)
Limitations
- How often we could find confirmation times varies by team (goalies who changed teams are often missing), so the results lean toward teams with a stable set of goalies.
- Save percentage is easily affected by luck; about 157 games in March 2026 are missing trading volume data and were not included in the main analysis.
- We only have one odds reading every 5 minutes; reporters’ tweets are not always the first source of the news.
- We did not calculate the gap between buy and sell prices or fees, so we can’t yet tell whether trading on this would make money; the continued move is under 0.1 percentage points, so the room for profit looks limited.
Update log
Version history
- v1.1-forward2026-10-20
The plan to check again on new 2026-27 games is finalized and timestamped, with the rules for what counts as “still holds” written in advance
- v1.0-fix2026-10-20
After another AI tool checked the code: fixed the timing for 44 confirmations posted “less than 20 minutes before the game” and recomputed; “finishes in about 2 hours” didn’t hold up, so it was changed to “halfway in about 1 hour”; added comparison checks
- v1.02026-10-20
Plan finalized and timestamped; finished the main analysis of the 2025-26 regular season (all changes along the way are recorded)
Study record
3b06a91190703537c405416c2624b033a1b4dc756d659d7bee912d2816054bce650106bea9e7948dab65be99250b4aed986a115994491a7bd066ab68f1e28acb01f34eda2b722f7cfb85405f18fdcaa4c21a1d333c311922cc198f1a532ee3c446593c2f24931cd275d3e8c28065ba39d634e405e10570bf7f0705e2ff0ea719This page is a research log for reference only and is not betting advice or a guarantee of profit.