R22Published

How Much Do NBA Odds Seen at 9 p.m. in Taiwan Change Before Tip-off?

From 9 p.m. Taiwan time to 5 minutes before the scheduled tip-off, how many percentage points the win probability implied by Polymarket moneyline odds moved, how often it moved a lot, and how often the favorite flipped.

Result status
Preliminary results
Last updated
Start date
Duration
Day 1 of 1
Plan registered
Illustration of a moon in the night sky with a slightly wavy line running to a morning sun, showing how odds change from the night before to tip-off
Image:Illustration (made by Mysports.AI)

How Much NBA Odds Move From 9 p.m. Taiwan Time to Tip-off: Key Findings

Finding so far
Median change in implied win probability from 9 p.m. Taiwan time to 5 minutes before the scheduled tip-off:2.0pts

987 games, 2025-26 regular season (tipping off 6 a.m.–noon Taiwan time); 95% range 2–2 (prices move in 1-cent steps and many games moved by exactly 2); mean 2.7 (2.5–2.9).

  • 127 games moved by more than 5 points: 12.9% (10.7%–15.1%); 26 games moved by more than 10 (2.6%).
  • Favorite flipped: 41 of the 974 games with a clear favorite at 9 p.m., 4.2% (2.9%–5.7%).
  • Split into three groups by the 9 p.m. favorite's chance: median change 2.5 points when the favorite was at 50%–60% at 9 p.m., 2.0 at 60%–75%, and 1.0 at 75% or more (descriptive, not tested).
  • Official injury reports: 97.6% of games that moved more than 5 points had a status change in that time, versus 90.1% of games that moved 5 points or less; for a change to or from Out, 89.0% versus 70.8%. This only shows they happened together (descriptive, not tested).
  • US afternoon games (tipping off overnight in Taiwan), 68 games, counted separately: median 2.0 points, 10.3% moved more than 5 points (small sample, range 4.2%–17.6%).

How to use this yourself

  1. On Polymarket, look at the single-game “who will win” market (moneyline) for an NBA game: the price next to each team is what this page calls the odds, and Polymarket’s help page explains where it comes from. A price of $0.60 = about a 60% chance = odds of about 1.67 (odds = 1 ÷ price). The number on screen is usually the midpoint of the best buy and sell prices, while this page used trade prices, so they can differ slightly. Note the time you looked: this page’s numbers are changes from 9 p.m. Taiwan time to 5 minutes before the scheduled tip-off. Open source ↗

  2. Use this page's numbers to judge how much it may move: half of the games moved by 2 points or less (for example, 60% becoming 58%–62%), about 1 in 8 moved by more than 5 points, and the favorite flipped in about 1 game in 24.

  3. See which group the 9 p.m. favorite's chance falls in and compare with this page's group numbers: at 50%–60%, the past median change was 2.5 points; at 60%–75%, 2.0; at 75% or more, 1.0 (descriptive, not tested).

  4. On the NBA's official injury report (a PDF; times are US Eastern), the Current Status column shows Out = will not play and Questionable = may play. This page found that almost every game has some status change between the evening and tip-off (97.6% of games that moved more than 5 points and 90.1% of those that moved 5 or less), so a change on the report does not mean the odds will move a lot; the two simply happened together. Open source ↗

These are past distributions from the 2025-26 regular season and Polymarket trade prices, with no guarantee for any game; trade prices are not the quotes you could trade at on screen, and other platforms' odds may differ. This page is not betting advice.

In plain words

Why study this?

Answers whether odds checked the night before still hold at tip-off, and whether to check again: it shows what size of change is common and what is rare.

What do we want to know?

Most NBA games tip off on Taiwan mornings, so many fans look at the odds the evening before. How much do the odds seen at 9 p.m. change by 5 minutes before tip-off? If they usually barely move, checking the night before is enough; if they often move a lot, it is worth checking again.

How do we study it?

Using public Polymarket trade-by-trade data, we looked at 987 games from the 2025-26 regular season (tipping off between 6 a.m. and noon Taiwan time) at two moments: 9 p.m. Taiwan time the evening before, and 5 minutes before the official scheduled tip-off. For each moment we took the median of all trade prices in the previous 10 minutes (bigger trades count more), turned it into the home team's win probability, and took the difference. We also checked the NBA's official injury reports for player status changes in between. The plan and the analysis code were timestamped before we looked at any results.

What have we found?

The median change in implied win probability was 2.0 percentage points: half of the games moved by 2 points or less. About 1 game in 8 moved by more than 5 points (12.9%), and in about 1 game in 24 the favorite flipped (4.2%). Games whose 9 p.m. favorite was at 50%–60% had a larger median change than games whose favorite was at 75% or more (descriptive, not tested).

What this means

For half of the games, the odds seen the night before moved by 2 percentage points or less by tip-off (for example, a 60% chance, odds about 1.67, becoming 62%, odds about 1.61). But about 1 game in 8 moves by more than 5 points and the favorite flips in about 1 game in 24, so the odds seen the night before are not always the odds at tip-off.

What can’t we say yet?

What we cannot tell yet is why prices move. Almost every game had an injury-report status change between the evening and tip-off (97.6% of big movers, 90.1% of small ones), so we can only say the two happened together, not that the news moved the price. Also, these odds are estimated from trade prices, not the quotes you could trade at on screen; it is one season and Polymarket only, and there is no comparable trade-level data for the new season to test it again.

Terms used on this page

Implied win probability
A Polymarket price is the market's estimate of the chance to win: a price of $0.60 = about a 60% chance = odds of about 1.67 (odds = 1 ÷ price).
Percentage point
The unit for the difference between two probabilities. Going from 60% to 62% is a change of 2 percentage points.
Favorite flip
The team with more than a 50% chance at 9 p.m. drops below 50% by tip-off, so the other team becomes the favorite.
Injury report status
Whether each player can play, per the NBA's official report: Out, Doubtful, Questionable, Probable or Available.
Scheduled tip-off
The official listed start time; the actual tip-off is usually a few minutes later.

Detailed results

How much the implied win probability moved from 9 p.m. Taiwan time to 5 minutes before the scheduled tip-off

987 games, 2025-26 regular season. Numbers are the share of games in each band.

Less than 1 point18.8 %
1 to under 2 points26.0 %
2 to 5 points42.2 %
Over 5 to 10 points10.2 %
Over 10 points2.6 %

1 point is about the change from odds of 1.67 to 1.64 (a 60% chance becoming 61%).

HypothesesFor deeper reading
  • Q1Pre-registered main result (an estimate, not a test): median 2.0 percentage points (95% range 2.0–2.0); half of the games moved by 2 points or lessExploratory
  • Q2Moved by more than 5 points: 12.9% (10.7%–15.1%), about 1 game in 8 (descriptive, not tested)Exploratory
  • Q3Favorite flipped: 4.2% (2.9%–5.7%), about 1 game in 24 (descriptive, not tested)Exploratory
  • Q497.6% of games that moved more than 5 points had an injury-report status change in that time, but so did 90.1% of games that moved 5 points or less; almost every game has one, so this only shows they happened together, not that the news moved the priceExploratory

NBA Odds Moves From 9 p.m. Taiwan Time to Tip-off: Research Methods

Research questions

  • From 9 p.m. Taiwan time to 5 minutes before the scheduled tip-off, what is the median change (up or down) in the implied win probability, in percentage points?
  • What share of games move by more than 5 percentage points? (descriptive, not tested)
  • In what share of games does the favorite flip? (descriptive, not tested)
  • Among games that moved more than 5 points, what share had a status change in the official injury report during that time? (descriptive, not causal)

Method summary

  • There is one main result: the median, over all games, of the absolute difference in the home team's implied win probability between the two moments. The home and away changes are the same size (the two prices add up to 1).
  • The odds at a moment = the share-weighted median of all trade prices in the previous 10 minutes; we use the median because trade prices bounce between the buy and sell price and there are occasional odd prints. These are trade prices, not the quotes you could trade at on screen. With the last trade, the average, a 30-minute window or 15 minutes before tip-off instead, the median stays around 2 points and the share moving more than 5 points stays at 12.2%–14.4%.
  • The 95% ranges come from resampling whole evenings 10,000 times (145 evenings). Polymarket prices move in 1-cent steps, so changes are almost always whole percentage points, and many games moved by exactly 2, which is why the range for the median is 2–2. This is the resampling range of the median, not a sign that every game moved by 2 (see the table for the spread); the pre-listed mean is 2.7 points (2.5–2.9).

Data and limitations

Limitations

  • Retrospective, not blind: the games were long over, and we had used the same season's prices to check how accurate the odds were at different times and how prices reacted to injury news, but had not measured how much each game moved between these two moments.
  • One season only (2025-26 regular season) and Polymarket only; the dataset is missing 21 days and covers only part of Polymarket trading (first-generation contracts).
  • We used the official scheduled start, not the actual tip-off; games where the official and Polymarket start times disagreed (postponed or moved games) were left out, and those may move more. The actual tip-off is usually later than scheduled, but trades from after the start could still slip into the 5-minutes-before window (moving it to 15 minutes before did not change the conclusion).
  • Injury status changes are not weighted by how important the player is, and almost every game has one, so we can only say they happened together, not that they moved the price.
  • 9 p.m. is a fixed moment and may not match when you actually look; the results apply only to NBA moneylines on Polymarket, and other platforms may differ.

Latest progress

Latest:Main results are in: half of the games moved by 2 points or less, and about 1 in 8 moved by more than 5. The plan and analysis code were timestamped first, and the results were reviewed by an AI tool

Next step:Study completed; no comparable public trade-level data exists for the new season, so there is no forward test; we will reconsider if new public trade-level data appears

Study timeline

  1. Data gate passed (game and trade counts only, no prices looked at)
  2. Plan finalized and timestamped after one round of AI-tool code review; analysis code tested on synthetic data and timestamped before running
  3. Main results in; a threshold-comparison bug was found and fixed, then rerun (main result unchanged), and reviewed by an AI tool

Update log

Version history

  1. v1.0-results

    Main result: median 2.0 points; 12.9% moved more than 5 points (the first run miscounted changes of exactly 5 points as more than 5 and showed 14.6%; fixed and logged as a deviation; after seeing results we also reworded the main sentence from "half moved by less than 2 points" to "half moved by 2 points or less", which is correct when changes come in whole steps)

  2. v1.0

    Plan finalized and timestamped (T0 = 2026-11-19 02:14 UTC); retrospective, not blind

Study record

Plan fingerprint (SHA-256)
Plain-language plan06df78acd4f19720a8e8bab581acd7ba30587d14651643b975218e783d98a7ae
Technical appendix18b2d5db1331fbf4a251c05aeb1e01337449000b3997f3e3728c28d82801b3d8
Corrected analysis code timestamp (stamped before the rerun, after seeing the first results and fixing a threshold-comparison bug)ab2cc063daff8d31a4cc963c33971d2738645ae807c18e04435fe5191eb6d241
Analysis code timestamp (before running)6b58c409564d98693c0ce77e6f9d68f7cae82979313da2e25096d354abf95188
Data gate525ae4ba656bd683b1f183bfe68b72840628e97b568ab668994117c83e11cb26
Timestamped into the public Bitcoin blockchain record with OpenTimestamps on Nov 19, 2026
Plan changes:6 (each recorded in the deviation log, DEVIATIONS.md; not public for now)
Code and data:Not public for now (contains third-party data sources and internal information)

This page is a research log for reference only and is not betting advice or a guarantee of profit.