Player Prop Line Shopping Study: 2,763 Props Across 5 Sportsbooks

By David Reilich, Founder of PropsBot.AI. Published July 19, 2026 · Study version 1.0.0 · Study method · Corrections policy

What Did The PropsBot Line-Shopping Study Find?

PropsBot’s July 18, 2026 line-shopping study found that 2,763 MLB and WNBA player-prop points were offered at the exact same line by at least two of five tracked sportsbooks near 90 minutes before start. The median difference between the highest and lowest break-even implied probability was 2.269 percentage points on overs and 1.589 points on unders. The study used a fixed 24-hour UTC window, retained captures 60-120 minutes before start, selected each book’s observation nearest T-90, and required selected cross-book captures to be within 300 seconds. It compared each side separately and used raw implied probability, so sportsbook hold remained in the measured price difference. This was a one-slate price study, not an outcome study: it did not test win rate, closing-line value, realized ROI, or guaranteed savings. Different lines were excluded, and multiple observations could involve the same player, game, market family, or sportsbooks.

<div data-propsbot-line-shopping-study data-theme="dark">
  <a href="https://propsbot.ai/player-prop-line-shopping-study/">Line-Shopping Research by PropsBot.AI</a>
</div>
<script defer src="https://web.propsbot.ai/embeds/line-shopping-study/v1.js"></script>

Quick Answer

PropsBot compared 2,763 MLB and WNBA player-prop points that were available at the exact same line from at least two of five sportsbooks near 90 minutes before the event started. On the over side, the median gap between the highest and lowest implied probability was 2.269 percentage points. The 90th-percentile gap was 5.197 points. On the under side, the median was 1.589 points and the 90th percentile was 3.737.

That is a price study, not a picks study. It does not say which side won, and it does not turn one slate into a season-long savings claim. It measures a narrower question: when the player, event, market, and prop point were identical, how far apart were the prices a bettor could have accepted?

The Result At A Glance

The 2,763 eligible prop points produced more side-level comparisons than prop points because over and under were analyzed separately. Eight prop points had enough valid prices for an under comparison but not an over comparison; many others had an over comparison but lacked two usable under prices.

What A Percentage-Point Gap Means

American odds are awkward to compare by subtraction. A 20-cent difference means something different around -110 than it does on a longshot. We converted every valid price into its break-even implied probability and measured the distance between the highest and lowest figure offered for the same side.

Suppose one book’s price implies a 50% break-even rate and another implies 52.3%. The second price asks the bettor to win more often to break even. The underlying player and line did not change; only the price did. That is the cost line shopping is designed to avoid.

The study uses raw implied probability, not a no-vig estimate. Sportsbook hold is therefore part of the observed difference. That is intentional for this question because the bettor pays the posted price. Readers interested in estimating a margin-free market can use the no-vig fair odds calculator.

How Much Book Coverage Was Available?

The source tracked DraftKings, FanDuel, BetMGM, Caesars, and ESPN BET. Not every book posted every prop at the required time and line. Of the 2,763 eligible prop points, 732 had two books, 648 had three, 1,185 had four, and 198 had all five.

Book-level coverage ranged from 419 eligible prop points at ESPN BET to 2,462 at DraftKings. Those counts describe presence in the study cohort, not sportsbook quality. A book with fewer matching rows may have posted different lines, offered fewer markets, or updated on a different schedule.

MLB And WNBA Results

MLB dominates this slate and should dominate the interpretation. The WNBA sample is useful as a companion signal, but 154 over comparisons and 125 under comparisons are not enough to make a sweeping league-level claim. The honest conclusion is that meaningful same-line price differences were visible in both sports on this date.

Which MLB Markets Showed The Widest Typical Gaps?

Several common batter markets had more than 200 over observations. Batter singles had a 3.326-point median over-side gap, RBI 3.147, hits plus runs plus RBI 2.935, runs scored 2.747, hits 2.676, doubles 2.514, total bases 2.269, and home runs 2.234.

This table is descriptive. It does not prove that singles or RBI will always offer the largest shopping opportunity. Market mix, odds range, sportsbook participation, and one day’s players all affect the result. Smaller pitcher samples are excluded from the comparison table even when their observed median was high.

Why The Median Leads The Headline

Player-prop prices do not have a tidy, symmetric distribution. A few longshot markets can create a large maximum and pull the mean upward even when most comparisons are much closer. The median answers a more practical question: what gap sat in the middle after every eligible observation was sorted from smallest to largest?

We also report the 75th and 90th percentiles so the upper part of the distribution remains visible. The maximum is included for completeness, but it should not be treated as a typical opportunity. A repeat study may show a different extreme simply because one unusual market or stale price appears in the cohort.

There Was No Single “Best Book”

This study ranks prices within each matching prop point; it does not name one sportsbook as the permanent winner. The best over price can be at one book while the best under price or another player’s market is elsewhere. Coverage also changes throughout the day. A sportsbook absent from one comparison may post the market later, use a different prop point, or lead the next comparison.

That is why a useful odds-shopping tool compares the exact market now instead of relying on a season-long logo ranking. The bettor’s question is not which brand is always best. It is which available line and price create the lowest break-even requirement for the side being considered.

Exact Study Method

1. Use the fixed window from July 18, 2026 at 04:00 UTC through July 19 at 04:00 UTC, with the end timestamp excluded.
2. Keep pregame captures made between 60 and 120 minutes before the scheduled start.
3. Define one prop point by sport, event, market, player, and exact line.
4. For each sportsbook, select the observation nearest 90 minutes before the start.
5. If two observations are equally close to 90 minutes, keep the earlier capture.
6. Require at least two sportsbooks offering the exact same prop point.
7. Require the selected cross-book captures to fall within 300 seconds of one another.
8. Analyze over and under separately, and require at least two valid prices for the side.
9. Convert negative American odds with -a / (-a + 100) and positive odds with 100 / (a + 100).
10. Calculate dispersion as the maximum implied probability minus the minimum, in percentage points.
11. Calculate percentiles by linear interpolation at (n - 1)q.

Reproducibility And Data Handling

The production collection contained 13,227,805 snapshots when the versioned export ran. The exporter used a read-only aggregation with secondary-preferred routing and did not write to production. A second independent execution reproduced every audit row and headline distribution.

The research package includes aggregate JSON and CSV results, a deterministic 50-row audit sample, the versioned exporter, and SHA-256 hashes. The audit rows are de-identified. They retain sport, market, book count, timing checks, quote counts, and derived dispersion without publishing player names, event identifiers, prop points, or raw sportsbook prices.

PropsBot does not redistribute the licensed odds feed. That is why the public result is aggregate and the audit identities are hashed. Reporters or research partners who need to inspect the package can request the versioned artifacts through the PropsBot contact page.

Limitations

• One slate: this is a cross-sectional study of July 18, not a season-long estimate.
• Near T-90: it measures prices near 90 minutes before start, not openers, closing lines, or the price a particular bettor actually received.
• Exact-line cohort: books posting different prop points are excluded from the price-dispersion calculation. Line disagreement is a separate shopping problem.
• At least two books: single-book markets are not represented.
• Raw implied probability: sportsbook margin is not removed.
• Related observations: multiple props can come from the same player, game, market family, and sportsbooks. The rows are not independent experiments.
• No outcome claim: the study does not measure win rate, closing-line value, realized ROI, or guaranteed savings.

What Bettors Can Do With This

First, keep the line and price together. Saying that a model likes over 1.5 RBI is incomplete without the odds. Second, compare the same market at more than one book before placing the wager. Third, record the number actually played so the decision can be reviewed later.

PropsBot’s odds shopping guide explains that workflow, while player props today provides the current research entry point. For payout math, use the live parlay calculator or implied probability calculator.

How To Cite This Study

Suggested citation: PropsBot.AI, Player Prop Line Shopping Study: 2,763 Props Across 5 Sportsbooks, version 1.0.0, July 19, 2026. Link to this permanent page and retain the study date, sports, exact-line rule, and T-90 timing in any summary.

Study FAQ

Did PropsBot compare different prop lines?

No. The headline price analysis requires the exact same player, event, market, and prop point. A book at 1.5 and a book at 2.5 are not treated as the same offer.

Does a wider gap prove one sportsbook was wrong?

No. The prices can reflect different hold, risk, customer flow, or update timing. The study shows the bettor faced different break-even requirements, not why each book posted its number.

Why are there fewer under comparisons?

Some player-prop markets were offered primarily or only on the over side, especially longshot event markets. A side enters the distribution only when at least two valid prices were present.

Will PropsBot repeat the study?

The method is versioned so it can be repeated on future MLB, WNBA, and other sport slates. Future reports should be published as separate dated cohorts rather than silently changing this result.

This study is for research and education. Sports betting involves risk, prices change, and no historical market observation guarantees a future result.