After a Public Pick Is Posted, Do the Odds Move Its Way?
After free NBA picks from PTT, Yuncai Xiaowei (a Taiwanese tipster) and Predictem are posted, does the backed team become more favored on Polymarket before tip-off? Pick data are collected by betlead.win.
Study details
On this page
Key findings
Comparable games (the same side, home or away, at a similar price, in other games over the same window) show 50%—almost the same. Based on the picks whose odds moved among 1,706 picks in the 2025-26 NBA regular season; each source’s share is computed and then averaged. Pick data collected by betlead.win (same team as Mysports).
- How small: a price moving from 0.55 (odds about 1.82) to 0.56 (about 1.79) is 1 percentage point; net of the usual drift the average here was −0.002 point, with a range of −0.23 to +0.23.
- No sign does not mean no effect: the data have limited ability to detect changes below about 0.3 percentage point (a rough estimate made after seeing the results). Nor does it mean picks are useless or the market is perfectly right.
- When picks backed underdogs, the backed team’s win chance rose about 0.3 point on average—but comparable underdogs rose by about the same over the same window, so this looks like a usual pre-game drift for underdogs.
- Re-running under several other conditions (such as odds 15 minutes before tip-off, only two of the sources, or adding the playoffs) keeps results near zero.
In plain words
What do we want to know?
Every day people post free NBA picks online. If a tipster truly sees something before the market, then after the pick appears and before tip-off, the market should increasingly favor the team they backed; if they are just guessing, market moves should be unrelated to the pick.
How do we study it?
We use Polymarket prices (shown as odds): a price of $0.60 ≈ a 60% chance ≈ odds of about 1.67. “Odds moving toward the pick” means the backed team’s win chance rises and its odds fall—for example, the price goes from 0.55 to 0.58 and the odds from about 1.82 to about 1.72. We collect public picks with posting times (about 1,700 in 2025-26) and measure how many percentage points the backed team’s win chance changes from when the pick is confirmed to exist until 5 minutes before tip-off. Because odds may drift before games anyway, we first subtract the average move of the comparable side in other games at the same time and similar price; only what remains counts. Each of the three sources gets its own average, then we average those, so the busiest source does not dominate.
After a pick is posted, which way do the odds go?
Hypothetical: after a pick on Team A, its odds go from 2.10 to 1.95. Lower odds mean more favored, so this counts as moving toward the pick.
- 01Record when the pick was posted
From free public picks collected by betlead.win, record each pick’s posting time and the team it backs.
- 02Check the odds at posting time
Find that team’s Polymarket odds at the moment the pick was posted.
- 03Check the closing odds
The last odds before tip-off usually reflect the most information.
- 04Compare the direction
Check whether the odds moved toward the backed team after the pick, compared with the move for simply backing the favorite in the same game.
2.10 and 1.95 are hypothetical. The plan is finalized and analysis is under way, with no results yet; odds moving toward a pick does not mean the tipster has inside information or that following it would profit.
What have we found?
There is no sign that odds move toward the picks. Between posting and tip-off, the backed team’s odds fell (more favored) 49% of the time versus 50% for comparable games; net of the usual drift, the implied win chance changed by −0.002 percentage point on average—essentially zero. Re-running under several conditions also stays near zero. The 2026-27 season will re-test the same method.
Taiwanese member picks (sub-study)
We also save Playsport members’ public picks every day and only check whether the pick’s direction matches later line moves. Because posting times are unknown, a match does not mean the tipster knew first. Only aggregate statistics are published, with no names.
What can’t we say yet?
Even if odds do move toward picks, the tipster may simply have seen public news earlier (such as injuries), or many people may follow the pick and push the price; we cannot tell these apart. The study also does not answer whether following picks makes money. With only three sources, results represent only those sources.
Terms used on this page
- Public pick
- A free game prediction posted online, such as “Lakers to win” or “Lakers −3.5”.
- Posting time
- When a pick was published. We take the later of the posting and last-edit times; picks edited after tip-off are excluded.
- Usual drift
- Odds may move in a consistent direction before games even without picks. We estimate it from the comparable side in other games at the same time and similar price, then subtract it.
- Percentage point
- The difference between two probabilities. For example, 50% to 51% is an increase of 1 percentage point.
- Playsport
- A Taiwanese sports-betting community where members post game predictions.
Detailed results
Share whose odds fell (became more favored) between posting and tip-off
Only cases where the odds moved; each source’s share is computed and then averaged.
Per the pre-registered plan | 2025-26 NBA regular season | Polymarket odds every 5 minutes | Pick data collected by betlead.win.
HypothesesFor deeper reading
- H1After a pick is posted and before tip-off, net of the usual drift, the backed team’s win chance rises on average (its odds fall) → no sign (−0.002 point, range −0.23 to +0.23)Insufficient evidence
- H1b(Secondary) The backed team moves more than simply backing the favorite in the same game → holds numerically (+0.35 point), but this is a mechanical price comparison driven by underdogs’ usual pre-game rise; it does not mean picks beat backing the favoriteSupported in this sample
- P1(2026-27 re-test) The new season shows the same direction with the same methodPending
Background and why it matters
If tipsters truly see something before the market, the market should increasingly favor the teams they back after the pick appears. This separates real insight from luck better than hit rates alone, and helps a model decide which public picks are worth considering.
Latest progress
Latest：2025-26 main results are in and corrected after external code review; all five data checkpoints passed
Next step：Re-test on the 2026-27 season with the same method (decision on 2027/2/15; until then we only count picks without looking at results)
Study timeline
- Direction set2026-10-20
- Daily Playsport archive started 10/21 (sub-study)2026-10-21
- Redesigned around picks with posting times; terms and novelty checked; plan finalized and timestamped after two external reviews2026-10-21
- 2025-26 analysis done; corrected after external code review2026-10-21
- Re-test on the 2026-27 season (decision on 2027/2/15)In progress
Methods
Method summary
- Only sources whose terms do not prohibit use: the PTT board, Yuncai Xiaowei and Predictem; sites whose terms forbid scraping or use (such as Covers, FOX and Action Network) are excluded.
- Pick time is the later of posting and last edit; picks edited after tip-off are excluded.
- We measure the change, in percentage points, in the backed team’s Polymarket win chance; for spread picks we use the same team’s win chance.
- We subtract the average move in other games at the same time; the 2026-27 season re-tests the same method.
Data and limitations
Data we use (all public and free)
- Public picks: PTT sports-betting board, Yuncai Xiaowei and Predictem (collected by betlead.win, which is run by the same team as Mysports; only side, line and time are used, articles are not reproduced)
- Polymarket NBA single-game odds (every 5 minutes, shared with R1)
- Industry closing lines: minmaxxxer / Hugging Face (CC-BY-4.0)
- Playsport members’ public picks (sub-study; archived daily after games, no paid content)
Limitations
- Last season’s picks were collected afterwards; posting times are self-reported by the sites, and deleted picks are missing.
- There are only three sources, two of them single authors, so results do not represent all online picks.
- Only one market (Polymarket) is used; spread picks use win chance as a proxy, and totals have no time-stamped odds, so only the pick’s line versus the closing line can be compared.
- Odds moving toward a pick does not mean the tipster has inside information or that following it would make money.
Update log
Version history
- v1.0-results2026-10-21
2025-26 main results; all 9 external review comments fixed, pre-registered numbers unchanged
- v1.02026-10-21
Redesigned around picks with posting times; plan finalized and timestamped (T0 = 2026-10-21 08:28 UTC)
Study record
8fbf2797774bdedf97d7f53ce938ee0352c56005a7ef9c8ad87669072e5b8b65a4eef72ae72c594c7e7f789c7022bb95cb35515f6283a35792703ee9c8c5c290This page is a research log for reference only and is not betting advice or a guarantee of profit.