Player prop markets are 2 to 4 times softer than game lines. Books that hire teams of traders for spreads and totals run prop pricing through a thinner layer. Smaller models, fewer adjustments, slower reactions to lineup news. TheOddsAPI surfaces those gaps for operators. NBA points, rebounds, assists, PRA, threes, blocks, steals, and double-doubles from up to 50 sportsbooks with Pinnacle as the sharp anchor.
50
Sportsbooks priced per market
8
NBA prop markets covered
T-30
Snapshot trail per edge
San Antonio Spurs at Minnesota Timberwolves, T-30 minutes to tip
Five prop edges across four players from four different soft books on the same line. The soft books are paying 11 to 12 cents more than Pinnacle. Captured live from prop_edge_snapshots.
The Under-heavy mix in this snapshot is not a directional bias claim. The Knicks at 76ers snapshot earlier the same evening (case study below) was Over-heavy at BetRivers. Soft books misprice in both directions on different nights. The structural point is that the soft-book layer is mispriced, not that one direction is reliably better.
Pinnacle accepts large action, holds the lowest margin in the industry, and adjusts prop lines from the same models that price their game totals. When Pinnacle quotes a player's points line at -121, the implied probability of the Over is 54.8 percent. When DraftKings quotes the same Over at -109, DraftKings is implying 52.2 percent. The 2.6-point probability gap is real money over thousands of bets.
Most retail books run NBA props from a small internal team or a consensus feed that lags Pinnacle by minutes to hours. The line is rarely re-derived from a model in real time.
A late scratch that drops a starter's minutes by 25 percent moves Pinnacle within seconds. Soft books often wait until pre-game lineups are confirmed by media reports, which can be 30 to 60 minutes after Pinnacle has already moved.
Soft books default to wider margins on props than on game lines. Their pricing buffer absorbs forecasting error, but the buffer also leaves room for traders to extract that margin when they identify which side of the line is wrong.
The structural result: prop markets are where sharp pricing data has the highest dollar-per-call ROI. The same 50-book dataset that powers TheOddsAPI's standard odds endpoint extends to player props with no additional integration work.
01
Call TheOddsAPI's /props/ endpoint for an event. The response includes every book's lines for the eight NBA prop markets, with Pinnacle present when available.
02
For each market, locate Pinnacle's line and price. Pinnacle is the baseline for fair probability. If Pinnacle has not yet quoted a market, skip it. There is no sharp anchor to compare against.
03
For each soft book offering the same player, market, direction, and point, calculate the cent-value gap against Pinnacle. The cent value works across the +100 / -100 boundary cleanly.
04
Discard candidates below your threshold. TheOddsAPI's /edges/ endpoint defaults to a min_edge of 50 (filters market microstructure noise). Set min_edge=10 for prop-scale gaps. What remains is sorted by edge score.
A book paying 11 cents more than Pinnacle on a 50/50 prop translates to roughly 2.5 percent expected value before juice. Compounded across hundreds of bets per season, the cumulative return is the difference between a hobbyist and an operator.
Player props refresh approximately every 5 minutes for events starting within 24 hours. TheOddsAPI's edge detection runs server-side on the same cadence as the parent sport's refresh tick (30 seconds for NBA). When a new prop quote enters our cache, the corresponding edge appears in /edges/ within roughly 30 seconds.
End-to-end edge surfacing latency is the edge detection cycle plus the time since the last props refresh. Worst case is around 5 minutes 30 seconds. Best case is around 30 seconds. Typical is a few minutes. Operators running quant systems on NBA props get edges fast enough to act before retail flow corrects them, but the prop refresh ceiling is not 30 seconds. Be honest with yourself about latency when sizing positions.
Map every soft-book NBA prop quote to Pinnacle and surface the gap. Python, ~40 lines.
import requests
# Step 1: Fetch NBA player props for an event from TheOddsAPI
event_id = "YOUR_EVENT_ID"
response = requests.get(
"https://api.theoddsapi.com/props/",
params={
"apiKey": "YOUR_KEY",
"sport_key": "basketball_nba",
"event_id": event_id,
"markets": "player_points,player_rebounds,player_assists,player_points_rebounds_assists,player_threes,player_blocks,player_steals,player_double_double",
},
)
event = response.json()
# Step 2: Index Pinnacle's lines by (market, player, direction)
PINNACLE = "pinnacle"
props = event.get("props", [])
pinnacle_index = {}
for prop_market in props:
market_key = prop_market["market"]
pinnacle_book = next((b for b in prop_market["books"] if b["book"] == PINNACLE), None)
if not pinnacle_book:
continue
for o in pinnacle_book["outcomes"]:
key = (market_key, o["description"], o["name"])
pinnacle_index[key] = (o.get("point"), o["price"])
if not pinnacle_index:
raise SystemExit("Pinnacle not quoting this event yet")
# Step 3: Compare every other book to Pinnacle on the same outcome
def cent_value(price):
if price > 0: return price - 100
if price < 0: return 100 + price
return 0
EDGE_THRESHOLD_CENTS = 10
edges = []
for prop_market in props:
market_key = prop_market["market"]
for book in prop_market["books"]:
if book["book"] == PINNACLE:
continue
for o in book["outcomes"]:
key = (market_key, o["description"], o["name"])
if key not in pinnacle_index:
continue
sharp_point, sharp_price = pinnacle_index[key]
if o.get("point") != sharp_point:
continue # different line, separate calculation
price_gap = cent_value(o["price"]) - cent_value(sharp_price)
if price_gap > EDGE_THRESHOLD_CENTS:
edges.append({
"player": o["description"],
"market": market_key,
"direction": o["name"],
"soft_book": book["book"],
"soft_price": o["price"],
"pinnacle": sharp_price,
"gap_cents": price_gap,
})
# Step 4: Sort by gap, surface the best
edges.sort(key=lambda e: e["gap_cents"], reverse=True)
for e in edges[:5]:
print(f"{e['player']:25s} {e['market']:18s} {e['direction']:5s} "
f"{e['soft_book']:15s} {e['soft_price']:+5d} vs PIN {e['pinnacle']:+5d} "
f"({e['gap_cents']:+d}c)")
For operators running this every 30 seconds across the NBA slate, TheOddsAPI's /edges/ endpoint pre-computes the full list server-side. One call returns a sorted list of every Pinnacle-anchored edge across every NBA event currently quoted, including both standard and prop edges in the same response.
curl 'https://api.theoddsapi.com/edges/?sport_key=basketball_nba&min_edge=10' \
-H 'x-api-key: YOUR_KEY'NBA injury and rest reports drop on a schedule that varies by team and broadcaster. Pinnacle adjusts within seconds. Soft books often wait until pre-game lineup confirmations from beat reporters, which can lag by 30 to 90 minutes. Every minute of lag is opportunity.
Game-line markets at every major book are priced by senior trading teams with multiple model inputs. Player prop markets at the same books are typically priced by smaller teams running thinner models. The gap shows up most on lower-profile players.
Pinnacle's alternate-line pricing is internally consistent with their primary line. Soft books often offer alternate lines that are mispriced relative to their own primary line, creating internal arbitrage as well as Pinnacle-vs-soft edges.
Retail prop bettors gravitate toward Over markets on star players. Soft books shade prices up on those Overs to capture retail margin. Pinnacle does not. Result: Under markets on popular stars are systematically underpriced at the soft books relative to Pinnacle.
| Single-Book Feed | TheOddsAPI | |
|---|---|---|
| Books quoted per market | 1 | Up to 50 (30+ on most NBA prop markets) |
| Pinnacle as benchmark | Sometimes | Always when Pinnacle is quoting |
| Markets covered | Player points typically | Points, rebounds, assists, PRA, threes, blocks, steals, double-double |
| Refresh latency | Varies, often 5+ minutes | Edge surfacing within minutes |
| Edge detection | Build it yourself | Pre-computed at /edges/ |
| Snapshot trail for backtest | None | T-30 captures stored in prop_edge_snapshots |
| Code integration | Custom per book | Single REST API, JSON, any language |
The snapshot trail is operator-specific. Every prop edge surfaced by TheOddsAPI is captured at T-30 to a timestamped table (prop_edge_snapshots), so operators can pull historical snapshots and apply their own settlement logic to compute a real cumulative ROI. Automated grading on Business+ tiers ships once the prop grading service is live.
A second example from earlier the same evening. Pulled directly from prop_edge_snapshots, captured 30 minutes before tip-off.
New York Knicks at Philadelphia 76ers
BetRivers was quoting Over markets on rotation players at consistently better prices than Pinnacle. McBride's Over 10.5 at BetRivers paid 12 cents more than Pinnacle's identical line. Operators running NBA prop edge detection on TheOddsAPI saw this within 30 seconds of the price posting and had a 30-minute window to act before tip-off.
The pattern matters more than the specific edge. BetRivers was systematically mispricing rotation-player Overs that night. An operator scanning all 50 books rather than one or two would have caught the pattern within minutes. An operator with no Pinnacle benchmark would have nothing to compare BetRivers against.
Eight NBA player prop markets, every one priced by every book that quotes it, with Pinnacle present when Pinnacle is in the feed.
player_points
Over/Under on player point totals.
player_rebounds
Over/Under on player rebound totals.
player_assists
Over/Under on player assist totals.
player_points_rebounds_assists
Combined PRA totals.
player_threes
Over/Under on three-pointers made.
player_blocks
Over/Under on blocks.
player_steals
Over/Under on steals.
player_double_double
Yes/No on whether a player records a double-double.
Coverage extends to MLB and NHL props on the same endpoint pattern, so operators expanding from NBA into other sports do not need to rewrite integration logic.
Prop edge detection is not risk-free. Understanding the limits protects your capital.
Some books quote regulation-only player point totals. Others include overtime. The same player's "Over 22.5 points" line can carry different settlement rules across books. TheOddsAPI exposes the raw data feed without papering over this, so operators must read book-specific settlement rules before placing real money.
Soft books that price props softly often impose lower betting limits on bettors who consistently identify edges. Edge that exists at the line will not always exist at the size you want to bet. Plan execution around per-book limits.
Early in the NBA season, soft-book prop pricing is loosest because models have less recent data. Edges narrow as the season matures. By the playoffs, when Pinnacle and the sharpest books have re-trained on current-season data, the typical edge size shrinks. The structural advantage remains, but the magnitude compresses.
Player prop pushes are common at .5 and .0 lines. TheOddsAPI's prop_edge_snapshots table captures every Pinnacle-anchored prop edge at T-30 minutes for every covered NBA, NHL, and MLB game, building a timestamped trail. Automated grading of those snapshots (with full push and void semantics) is on the roadmap for Business+ tiers and ships once the grading service is live for prop markets. Operators who want to grade today can pull snapshots and apply their own settlement logic.
Multi-sport prop coverage across NBA, MLB, and NHL.
Methodology behind the sharp anchor used in every edge calculation.
Backtest h2h, spreads, and totals against settled outcomes. Player props historical coverage is on the roadmap.
Free tier includes 25 requests per day for evaluation. Pro unlocks all 24 sports plus h2h, spreads, and totals. Business unlocks player props, the /edges/ endpoint with prop edges, the T-30 snapshot trail, and edge surfacing within minutes for production prop trading.