Datasets:
College football mainline odds (2026), the dataset behind The Closing Line issue 7
Browse files- README.md +128 -0
- kick_day=2026-08-27/data.parquet +3 -0
- kick_day=2026-08-28/data.parquet +3 -0
- kick_day=2026-08-29/data.parquet +3 -0
- kick_day=2026-09-03/data.parquet +3 -0
- kick_day=2026-09-04/data.parquet +3 -0
- kick_day=2026-09-05/data.parquet +3 -0
- kick_day=2026-09-06/data.parquet +3 -0
- kick_day=2026-09-07/data.parquet +3 -0
- kick_day=2026-09-10/data.parquet +3 -0
- kick_day=2026-09-11/data.parquet +3 -0
- kick_day=2026-09-12/data.parquet +3 -0
- kick_day=2026-09-17/data.parquet +3 -0
- kick_day=2026-09-18/data.parquet +3 -0
- kick_day=2026-09-19/data.parquet +3 -0
- kick_day=2026-09-24/data.parquet +3 -0
- kick_day=2026-09-25/data.parquet +3 -0
- kick_day=2026-09-26/data.parquet +3 -0
- kick_day=2026-10-01/data.parquet +3 -0
- kick_day=2026-10-02/data.parquet +3 -0
- kick_day=2026-10-03/data.parquet +3 -0
- kick_day=2026-10-06/data.parquet +3 -0
- kick_day=2026-10-07/data.parquet +3 -0
- kick_day=2026-10-08/data.parquet +3 -0
- kick_day=2026-10-09/data.parquet +3 -0
- kick_day=2026-10-10/data.parquet +3 -0
- kick_day=2026-10-13/data.parquet +3 -0
- kick_day=2026-10-14/data.parquet +3 -0
- kick_day=2026-10-15/data.parquet +3 -0
- kick_day=2026-10-16/data.parquet +3 -0
- kick_day=2026-10-17/data.parquet +3 -0
- kick_day=2026-10-20/data.parquet +3 -0
- kick_day=2026-10-21/data.parquet +3 -0
- kick_day=2026-10-22/data.parquet +3 -0
- kick_day=2026-10-23/data.parquet +3 -0
- kick_day=2026-10-24/data.parquet +3 -0
- kick_day=2026-10-27/data.parquet +3 -0
- kick_day=2026-10-28/data.parquet +3 -0
- kick_day=2026-10-29/data.parquet +3 -0
- kick_day=2026-10-30/data.parquet +3 -0
- kick_day=2026-10-31/data.parquet +3 -0
- scripts/extract_day.sh +23 -0
- scripts/stage1_compact.sh +46 -0
- scripts/xraw_day.sh +34 -0
- scripts/xtop_day.sql +47 -0
README.md
ADDED
|
@@ -0,0 +1,128 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-4.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
pretty_name: SmartStake College Football Mainline Odds (2026)
|
| 6 |
+
tags:
|
| 7 |
+
- sports-betting
|
| 8 |
+
- sports-analytics
|
| 9 |
+
- college-football
|
| 10 |
+
- ncaaf
|
| 11 |
+
- odds
|
| 12 |
+
- prediction-markets
|
| 13 |
+
size_categories:
|
| 14 |
+
- 100M<n<1B
|
| 15 |
+
task_categories:
|
| 16 |
+
- time-series-forecasting
|
| 17 |
+
configs:
|
| 18 |
+
- config_name: default
|
| 19 |
+
data_files:
|
| 20 |
+
- split: train
|
| 21 |
+
path: "kick_day=*/*.parquet"
|
| 22 |
+
---
|
| 23 |
+
|
| 24 |
+
# SmartStake College Football Mainline Odds (2026)
|
| 25 |
+
|
| 26 |
+
The exact dataset behind the study
|
| 27 |
+
["Who's Actually Sharp on College Football Mainlines?"](https://smartstake.app/learn/sharpest-sportsbooks-cfb-mainlines):
|
| 28 |
+
every pregame price change on college football moneylines, spreads, and totals at the
|
| 29 |
+
twelve sportsbooks, exchanges, and prediction markets in the study, for the 1,067 games
|
| 30 |
+
that kicked off from 27 August through 31 October 2026. A companion to
|
| 31 |
+
[SmartStake MLB Mainline Odds (2026)](https://huggingface.co/datasets/SmartStake/mlb-mainlines).
|
| 32 |
+
|
| 33 |
+
## Coverage
|
| 34 |
+
|
| 35 |
+
- **Games:** 1,067, first kickoff 27 August 2026, last kickoff 31 October 2026 (US time).
|
| 36 |
+
- **Window:** pregame only. Every row's price was posted before kickoff (`ts < start_ms`).
|
| 37 |
+
The earliest quotes date from 7 August.
|
| 38 |
+
- **Markets:** full-game moneyline, spread, and total, at every line each book posted
|
| 39 |
+
(alternate spreads and totals included, as in the study).
|
| 40 |
+
- **Books:** Pinnacle, Bookmaker, BetOnline, Kalshi, Polymarket, Novig, ProphetX, 4Caster,
|
| 41 |
+
FanDuel, DraftKings, Fanatics, and Caesars.
|
| 42 |
+
- **Rows:** 172,700,424 (about 1 GB of Parquet, one file per kickoff day).
|
| 43 |
+
|
| 44 |
+
## One row is a run of one price
|
| 45 |
+
|
| 46 |
+
Books re-send an unchanged price over and over. Each row here is one run: a book's price
|
| 47 |
+
for one selection staying the same from `ts` (first seen) to `ts_last` (last re-sent).
|
| 48 |
+
A new row starts when the price, or whether it is bettable, changes. A price holds until
|
| 49 |
+
the next row for the same selection, so to read a book's price at time T, take the
|
| 50 |
+
latest row with `ts <= T`.
|
| 51 |
+
|
| 52 |
+
**Sportsbooks** (Pinnacle, Bookmaker, BetOnline, Novig, ProphetX, FanDuel, DraftKings,
|
| 53 |
+
Fanatics, Caesars) are at the feed's millisecond resolution.
|
| 54 |
+
|
| 55 |
+
**Kalshi, Polymarket, and 4Caster** show many resting offers at once, so their rows are
|
| 56 |
+
the best price you could actually bet, minute by minute: `ts` and `ts_last` fall on whole
|
| 57 |
+
minutes, and a row with `eff = null` marks a minute with nothing bettable. A 4Caster offer
|
| 58 |
+
counts as gone if it is not re-posted within 90 seconds, and a Kalshi offer under $100 is
|
| 59 |
+
ignored rather than treated as a pulled price.
|
| 60 |
+
|
| 61 |
+
A selection is `(matchKey, market, line, side, book, contract)`.
|
| 62 |
+
|
| 63 |
+
## Schema
|
| 64 |
+
|
| 65 |
+
| column | type | description |
|
| 66 |
+
|---|---|---|
|
| 67 |
+
| `matchKey` | string | Per-game identity. Group on this. |
|
| 68 |
+
| `market` | string | `Moneyline`, `Spread`, or `Total Points`. |
|
| 69 |
+
| `book` | string | Sportsbook, exchange, or prediction market. |
|
| 70 |
+
| `contract` | string | Kalshi's contract id (Kalshi lists each team as its own contract). Empty for every other book. |
|
| 71 |
+
| `side` | string | `A` = home (moneyline, spread) or over (total); `B` = away or under. |
|
| 72 |
+
| `line` | double | `0` for moneylines; the **home** team's spread for spreads (so both sides of a spread share one `line`); the total for totals. |
|
| 73 |
+
| `ts` | int64 | First time this price was seen, epoch milliseconds (UTC). |
|
| 74 |
+
| `ts_last` | int64 | Last time the book re-sent this same price, epoch milliseconds (UTC). |
|
| 75 |
+
| `start_ms` | int64 | Scheduled kickoff, epoch milliseconds (UTC). |
|
| 76 |
+
| `home` | string | Home team. |
|
| 77 |
+
| `away` | string | Away team. |
|
| 78 |
+
| `eff` | double | The bettable decimal price. Null when the price was pulled, or when the book reported a max bet under $100. |
|
| 79 |
+
| `del` | bool | The book flagged the selection as removed while still showing a price (a feed refresh; it usually comes back). |
|
| 80 |
+
| `mx` | double | The book's reported max bet (or size at the price) when the run started, where the feed carries one. |
|
| 81 |
+
| `kick_day` | date | Partition key: the kickoff date, taken as kickoff time minus 8 hours in UTC (so a late-night Pacific kickoff stays on its Saturday). |
|
| 82 |
+
|
| 83 |
+
## How it was built
|
| 84 |
+
|
| 85 |
+
1. `scripts/extract_day.sh` pulls each day of SmartStake's raw odds archive down to
|
| 86 |
+
pregame NCAAF moneylines, spreads, and totals at the twelve books.
|
| 87 |
+
2. `scripts/stage1_compact.sh` turns the sportsbooks' ticks into the runs above. A pulled
|
| 88 |
+
price (null odds, or odds of 1 or less) and a price with a reported max bet under $100
|
| 89 |
+
both become `eff = null`, so neither counts as a price anyone could bet.
|
| 90 |
+
3. `scripts/xraw_day.sh` and `scripts/xtop_day.sql` do the same for the three exchanges,
|
| 91 |
+
keeping each resting offer separate and then taking the best live one each minute.
|
| 92 |
+
|
| 93 |
+
The study's other rules are applied in the analysis, not here: a new exchange price has
|
| 94 |
+
to hold for five minutes before it counts as a move, and a cross showing an edge above 5%
|
| 95 |
+
is thrown out as a bad print.
|
| 96 |
+
|
| 97 |
+
## Notes and caveats
|
| 98 |
+
|
| 99 |
+
- **Odds are raw.** Nothing is de-vigged. The study de-vigs each book's two sides before
|
| 100 |
+
comparing books, except for crossed markets, which use posted prices.
|
| 101 |
+
- **Exchange bots are left in on purpose.** Refill-bot prints are part of what the study
|
| 102 |
+
measures; the five-minute hold above is how the study keeps them from counting as moves.
|
| 103 |
+
- **Sparse feeds.** BetOnline and Fanatics confirm their prices rarely, so a run can last
|
| 104 |
+
hours without the book re-sending it. Check `ts_last` before treating an old price as live.
|
| 105 |
+
- **No results.** Game outcomes are not included.
|
| 106 |
+
|
| 107 |
+
## Loading
|
| 108 |
+
|
| 109 |
+
```python
|
| 110 |
+
import pandas as pd
|
| 111 |
+
df = pd.read_parquet("hf://datasets/SmartStake/cfb-mainlines/kick_day=2026-10-03")
|
| 112 |
+
```
|
| 113 |
+
|
| 114 |
+
```sql
|
| 115 |
+
-- DuckDB, straight from the hub
|
| 116 |
+
SELECT book, count(*) FROM 'hf://datasets/SmartStake/cfb-mainlines/kick_day=*/*.parquet' GROUP BY 1;
|
| 117 |
+
```
|
| 118 |
+
|
| 119 |
+
## Provenance, license, and responsible use
|
| 120 |
+
|
| 121 |
+
Odds were collected from public sportsbook and exchange feeds. Released under
|
| 122 |
+
CC BY 4.0 for research and educational use. This dataset is informational and
|
| 123 |
+
historical: past prices do not predict future outcomes, and betting carries risk of
|
| 124 |
+
loss. Not affiliated with any sportsbook. 21+.
|
| 125 |
+
|
| 126 |
+
## Citation
|
| 127 |
+
|
| 128 |
+
> SmartStake (2026). SmartStake College Football Mainline Odds (2026). Hugging Face.
|
kick_day=2026-08-27/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7b1d01020ae4c28c9c537cadade2c4d0ed8c8221e1049c4e750a63c06052078b
|
| 3 |
+
size 3137046
|
kick_day=2026-08-28/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6d82dce65d7d4e4dd3411d9135963a7ac01c6ec46fcec9bee69431239c6c8c95
|
| 3 |
+
size 4265963
|
kick_day=2026-08-29/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ed61947c582ab7649e3336b53e85c0ea71f638958326cb86e7bf1e3a45425576
|
| 3 |
+
size 27907417
|
kick_day=2026-09-03/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9b9f83bb9c68ab654310814710a7c558c39689b60bf1cfd9bdf65d8a91798594
|
| 3 |
+
size 22336949
|
kick_day=2026-09-04/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:17039986e2228d4f8788b462b642222abdeb9e2aa4d4e69158d3e364e33f2757
|
| 3 |
+
size 17151482
|
kick_day=2026-09-05/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a616bc2871eee8e2cf9ce9c59638dfa4bfc2e23ee9d8ddf8a70e38793c82975d
|
| 3 |
+
size 120075599
|
kick_day=2026-09-06/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:539d2ec5a8f2c94710033a25cafa0b3d9b51dd2e33cb6533c502b291dcbbdcc6
|
| 3 |
+
size 6395831
|
kick_day=2026-09-07/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1cc95c6b62f47b70ba82f4298f3d72fb2c5148349a66360689b9a22f548bfdb0
|
| 3 |
+
size 2015416
|
kick_day=2026-09-10/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:26dedfab2f5b5a2bd32df6aa7b18980ba1cc203ab59c9740730a25b66a2ce3b4
|
| 3 |
+
size 1164353
|
kick_day=2026-09-11/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cf500f14fd2e6712a51240a2934c726310e86aff0061bf9be967540762b59c4d
|
| 3 |
+
size 13029860
|
kick_day=2026-09-12/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ed4ce85f85d1b7f511222a443cea5bba681f80da735047ab6f1c1873e13933da
|
| 3 |
+
size 194248888
|
kick_day=2026-09-17/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1255a0d2a35608e1d860b06d27e80ad8064f4e3097e4f9a1a62034fc96bc0933
|
| 3 |
+
size 1570731
|
kick_day=2026-09-18/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d77593044be1d8b0cf6cb7ec8b306181d86474d20a657e7b8301c7f150b1ad96
|
| 3 |
+
size 5517951
|
kick_day=2026-09-19/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:65d0bcb23d445f0568a5d0643fd06faf36b5655aae074ff14fbed264784316d4
|
| 3 |
+
size 101866156
|
kick_day=2026-09-24/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ad9eaba3562c3d6d935ac66821ce8da747d8c589860af55eba809080634b1148
|
| 3 |
+
size 1419866
|
kick_day=2026-09-25/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d69e3d8ab5db441a74b2e8503187d5b161d555c000abe026182b1f50d4ec7c73
|
| 3 |
+
size 5881150
|
kick_day=2026-09-26/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e5620be2f557e5413acf16f8696c9afda5df2eefabe7dfbe8cb74c3a4c29109a
|
| 3 |
+
size 76639343
|
kick_day=2026-10-01/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:590d85a3c563e4253c566789df5cd74566993f1dbbe90a8313cc83f7480404cd
|
| 3 |
+
size 2702365
|
kick_day=2026-10-02/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:67ca046fb501254c3d0716ee74c023cd9e90f8f135af5489f291e04403d6537c
|
| 3 |
+
size 5052120
|
kick_day=2026-10-03/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1a84929e10614e67dcd976cac6417d51f30e089b2483acd97d843e4e3d6a4086
|
| 3 |
+
size 75484841
|
kick_day=2026-10-06/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a9244f13bbd1cbeb3368823c6c7ac2a5427c845af1355409bca5b1bd83000493
|
| 3 |
+
size 959833
|
kick_day=2026-10-07/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e4793e228ec5858bb464874da643653a80578d6a1f86db3b66144bb149463349
|
| 3 |
+
size 2604273
|
kick_day=2026-10-08/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:60e35469db9069a1f3af3d5d0965866bd8913c12f181fafb6d69c83d28a00ef8
|
| 3 |
+
size 5133934
|
kick_day=2026-10-09/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1f032c17f9a6e39ea401d8026a59771e0c6a60ecac7339a192426483b3ce4a4c
|
| 3 |
+
size 7811026
|
kick_day=2026-10-10/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8f46f402e7d9cd87a2981560630e2b55385f38fde98dfedf7b65f89b3a9f8b81
|
| 3 |
+
size 65868476
|
kick_day=2026-10-13/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2b2cb1ad03512221b2392921c0e6a77197140cb49417ab5b4e87b7f378112f03
|
| 3 |
+
size 2117645
|
kick_day=2026-10-14/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:84e48b44fe08edb4fa39ba5b2b89a7ce8d40e37f309da53dcce3dd27dd3184ab
|
| 3 |
+
size 1977519
|
kick_day=2026-10-15/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:29b440f20072d64b102f1230120312c45c21e0f8287f558b1021d6e484f08710
|
| 3 |
+
size 4592122
|
kick_day=2026-10-16/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4f763d996cf5b44234e106fc01695752c5f07921914fec3abe81dba34277da20
|
| 3 |
+
size 4439758
|
kick_day=2026-10-17/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:989748adaf4a9d722890855b393b0f0aa3816349471dcc949618ab8973672614
|
| 3 |
+
size 65620081
|
kick_day=2026-10-20/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b8a896dced7ba0c4a939cde8e7031b695a652f73de851024d1f4fe95a65572e8
|
| 3 |
+
size 2949313
|
kick_day=2026-10-21/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:86d0ef3f57fdc139acb8a215f77becac0c8848c619bcc7b3cb8c1bf9a79486fd
|
| 3 |
+
size 1043685
|
kick_day=2026-10-22/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7c66d5977f98eaed70f99d3fd37600e9638b78b027460c072475060f7e344db3
|
| 3 |
+
size 3678154
|
kick_day=2026-10-23/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c8c7b940a59cf5a13cf9814b7624619cc96a99ce70e908ebf6b26da173300cb4
|
| 3 |
+
size 7370365
|
kick_day=2026-10-24/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d5f3f82554971539f6e0c863bee823ab62e8b54c99bb169d29ff140747d7bfee
|
| 3 |
+
size 73063837
|
kick_day=2026-10-27/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e07258a7f4d17edbb45a7f11b24e348e2d7a826a2b4c3cdb123ca7730b520998
|
| 3 |
+
size 2432582
|
kick_day=2026-10-28/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3c4968c7e970c2fd3216d6300da4667daab1d892a702618c520525573de3a8cb
|
| 3 |
+
size 2659096
|
kick_day=2026-10-29/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:00781d25d108923766c02529416af371975ba0982510a387573c081aaf01a17e
|
| 3 |
+
size 3667500
|
kick_day=2026-10-30/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4d982ec90ab8efb285e6d8ecb54d3fc72be8c9031f4dd46c4a026c2c8fde3413
|
| 3 |
+
size 5960504
|
kick_day=2026-10-31/data.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b557a3eee1facd16cdee83e7e2bcc6140c3aa6585bb1685c5ff5c22212f73c82
|
| 3 |
+
size 80140520
|
scripts/extract_day.sh
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/bin/bash
|
| 2 |
+
# Extract pregame CFB full-game mainlines for the 12 study books from one day partition.
|
| 3 |
+
# usage: extract_day.sh 2026-09-12
|
| 4 |
+
D=$1
|
| 5 |
+
SRC=s3://smartstake-bet-updates-archive/bet_updates_archive/bet_update_v1/date=$D
|
| 6 |
+
FINAL=/home/stats/cfb_cross/cache/day=$D.parquet; OUT=/home/stats/cfb_cross/cache/tmp_$D.parquet
|
| 7 |
+
duckdb -init /home/stats/.duckdbrc_noview -c "
|
| 8 |
+
COPY (
|
| 9 |
+
SELECT matchKey, lineKey, market, bookmaker, timestamp,
|
| 10 |
+
epoch_ms(startDate::TIMESTAMP) AS start_ms, startDate,
|
| 11 |
+
homeCompetitor, awayCompetitor, selectionCompetitor, selectionLine,
|
| 12 |
+
selectionPoints, handicapSelectionPoints, odds, maxBetSize,
|
| 13 |
+
isAlt, isDeleted, nativeMarketId, nativeSelectionId, eventName, source
|
| 14 |
+
FROM read_parquet('$SRC/*.parquet', union_by_name=true)
|
| 15 |
+
WHERE league = 'ncaaf'
|
| 16 |
+
AND market IN ('Moneyline','Spread','Total Points')
|
| 17 |
+
AND bookmaker IN ('bookmaker','pinnacle','novig','prophetx','kalshi','polymarket',
|
| 18 |
+
'betonline','fanduel','4caster','draftkings','fanatics','caesars')
|
| 19 |
+
AND NOT coalesce(isPlayer,false)
|
| 20 |
+
AND timestamp < epoch_ms(startDate::TIMESTAMP)
|
| 21 |
+
) TO '$OUT' (FORMAT parquet, COMPRESSION zstd);
|
| 22 |
+
"
|
| 23 |
+
[ -s "$OUT" ] && mv "$OUT" "$FINAL"
|
scripts/stage1_compact.sh
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/bin/bash
|
| 2 |
+
# Compact one extracted day into effective-price change rows, partitioned by kickoff day.
|
| 3 |
+
# side: A = home/over, B = away/under. line: 0 (ML), home spread number, or total number.
|
| 4 |
+
# eff: bettable decimal price, NULL if pulled (odds null) or maxBetSize reported < $100.
|
| 5 |
+
# del: isDeleted row that still carries odds (feed refresh cycle; expires later if not re-added).
|
| 6 |
+
# ts/ts_last: first and last sighting of a run of identical prices (polled books re-log unchanged prices).
|
| 7 |
+
# contract: Kalshi nativeMarketId (team-specific contracts), '' for every other book.
|
| 8 |
+
D=$1
|
| 9 |
+
IN=/home/stats/cfb_cross/cache/day=$D.parquet
|
| 10 |
+
duckdb -init /home/stats/.duckdbrc_noview -c "
|
| 11 |
+
COPY (
|
| 12 |
+
WITH b AS (
|
| 13 |
+
SELECT matchKey, market, bookmaker AS book,
|
| 14 |
+
CASE WHEN bookmaker='kalshi' THEN nativeMarketId ELSE '' END AS contract,
|
| 15 |
+
CASE WHEN selectionLine IN ('hometeam','over') THEN 'A' ELSE 'B' END AS side,
|
| 16 |
+
CASE market WHEN 'Moneyline' THEN 0.0
|
| 17 |
+
WHEN 'Spread' THEN CASE WHEN selectionLine='hometeam' THEN selectionPoints ELSE -selectionPoints END
|
| 18 |
+
ELSE selectionPoints END AS line,
|
| 19 |
+
timestamp AS ts, start_ms, homeCompetitor AS home, awayCompetitor AS away,
|
| 20 |
+
CASE WHEN odds IS NULL OR odds <= 1 THEN NULL
|
| 21 |
+
WHEN maxBetSize IS NOT NULL AND maxBetSize < 100 THEN NULL
|
| 22 |
+
ELSE odds END AS eff,
|
| 23 |
+
(isDeleted AND odds IS NOT NULL) AS del,
|
| 24 |
+
maxBetSize AS mx,
|
| 25 |
+
strftime(make_timestamp((start_ms - 8*3600000)*1000), '%Y-%m-%d') AS kick_day
|
| 26 |
+
FROM read_parquet('$IN')
|
| 27 |
+
WHERE selectionLine IN ('hometeam','awayteam','over','under')
|
| 28 |
+
AND (market = 'Moneyline' OR selectionPoints IS NOT NULL)
|
| 29 |
+
),
|
| 30 |
+
d AS (
|
| 31 |
+
SELECT *, CASE WHEN eff IS NOT DISTINCT FROM lag(eff) OVER w AND del IS NOT DISTINCT FROM lag(del) OVER w
|
| 32 |
+
THEN 0 ELSE 1 END AS chg
|
| 33 |
+
FROM b WINDOW w AS (PARTITION BY matchKey, market, line, side, book, contract ORDER BY ts)
|
| 34 |
+
),
|
| 35 |
+
r AS (
|
| 36 |
+
SELECT *, sum(chg) OVER (PARTITION BY matchKey, market, line, side, book, contract ORDER BY ts
|
| 37 |
+
ROWS UNBOUNDED PRECEDING) AS run
|
| 38 |
+
FROM d
|
| 39 |
+
)
|
| 40 |
+
-- One row per run of identical (eff, del): ts = first sighting, ts_last = last re-log of that price.
|
| 41 |
+
SELECT matchKey, market, book, contract, side, line, min(ts) AS ts, max(ts) AS ts_last,
|
| 42 |
+
any_value(start_ms) start_ms, any_value(home) home, any_value(away) away,
|
| 43 |
+
any_value(eff) eff, any_value(del) del, arg_min(mx, ts) mx, any_value(kick_day) kick_day
|
| 44 |
+
FROM r GROUP BY matchKey, market, book, contract, side, line, run
|
| 45 |
+
) TO '/home/stats/cfb_cross/stage1' (FORMAT parquet, COMPRESSION zstd, PARTITION_BY (kick_day), FILENAME_PATTERN 'src_${D}_{i}', OVERWRITE_OR_IGNORE);
|
| 46 |
+
"
|
scripts/xraw_day.sh
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/bin/bash
|
| 2 |
+
# Exchange fix, pass 1: pull one extracted day's exchange rows (4caster, polymarket, kalshi) into
|
| 3 |
+
# xraw/, partitioned by kickoff day like stage1. Usage: xraw_day.sh <sport dir> <day>
|
| 4 |
+
# lvl: one resting offer (4caster: nativeSelectionId) or price level (polymarket: "token@price");
|
| 5 |
+
# kalshi sends only its current top of book, so the whole contract side is one stream ('').
|
| 6 |
+
# eff: bettable price at that offer/level, NULL once deleted, nulled or under $100.
|
| 7 |
+
# Kalshi rows under $100 are dropped instead: they are $1-type orders sitting on top of the real
|
| 8 |
+
# price, and treating them as "pulled" made Kalshi blink in and out.
|
| 9 |
+
S=$1; D=$2
|
| 10 |
+
duckdb -bail -init /dev/null -c "
|
| 11 |
+
SET memory_limit='6GB'; SET threads=2;
|
| 12 |
+
COPY (
|
| 13 |
+
SELECT matchKey, market, bookmaker AS book,
|
| 14 |
+
CASE WHEN bookmaker='kalshi' THEN nativeMarketId ELSE '' END AS contract,
|
| 15 |
+
CASE WHEN bookmaker='kalshi' THEN '' ELSE coalesce(nativeSelectionId, '') END AS lvl,
|
| 16 |
+
CASE WHEN selectionLine IN ('hometeam','over') THEN 'A' ELSE 'B' END AS side,
|
| 17 |
+
CASE market WHEN 'Moneyline' THEN 0.0
|
| 18 |
+
WHEN 'Spread' THEN CASE WHEN selectionLine='hometeam' THEN selectionPoints ELSE -selectionPoints END
|
| 19 |
+
ELSE selectionPoints END AS line,
|
| 20 |
+
timestamp AS ts, start_ms, homeCompetitor AS home, awayCompetitor AS away,
|
| 21 |
+
CASE WHEN odds IS NULL OR odds <= 1 OR isDeleted THEN NULL
|
| 22 |
+
WHEN maxBetSize IS NOT NULL AND maxBetSize < 100 THEN NULL
|
| 23 |
+
ELSE odds END AS eff,
|
| 24 |
+
maxBetSize AS mx,
|
| 25 |
+
strftime(make_timestamp((start_ms - 8*3600000)*1000), '%Y-%m-%d') AS kick_day
|
| 26 |
+
FROM read_parquet('/home/stats/${S}/cache/day=$D.parquet')
|
| 27 |
+
WHERE bookmaker IN ('4caster','polymarket','kalshi')
|
| 28 |
+
AND market IN ('Moneyline','Spread','Total Points')
|
| 29 |
+
AND selectionLine IN ('hometeam','awayteam','over','under')
|
| 30 |
+
AND (market = 'Moneyline' OR selectionPoints IS NOT NULL)
|
| 31 |
+
AND timestamp < start_ms
|
| 32 |
+
AND NOT (bookmaker = 'kalshi' AND odds > 1 AND maxBetSize IS NOT NULL AND maxBetSize < 100)
|
| 33 |
+
) TO '/home/stats/${S}/xraw' (FORMAT parquet, COMPRESSION zstd, PARTITION_BY (kick_day), FILENAME_PATTERN 'src_${D}_{i}', OVERWRITE_OR_IGNORE);
|
| 34 |
+
"
|
scripts/xtop_day.sql
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
-- Exchange fix, pass 2: best bettable price per exchange stream, minute by minute, for one kickoff day.
|
| 2 |
+
-- Tokens: @S@ (nfl_cross | cfb_cross), @K@ (kick_day)
|
| 3 |
+
-- Output: xtop/kick_day=@K@/ in the stage1 row shape (one row per run of an unchanged best price), so
|
| 4 |
+
-- the crosses step reads it in place of the exchanges' stage1 rows and the other studies build on it.
|
| 5 |
+
SET memory_limit='12GB'; SET threads=4;
|
| 6 |
+
|
| 7 |
+
-- 1. Each offer / price level is live from its row until its next row (an update, or a delete = NULL eff).
|
| 8 |
+
-- 4caster re-logs its whole book every 60 s, so an offer not re-logged within 90 s is gone.
|
| 9 |
+
CREATE TEMP TABLE iv AS
|
| 10 |
+
SELECT matchKey, market, book, contract, side, line, start_ms, home, away, eff, mx,
|
| 11 |
+
ts // 60000 AS m0,
|
| 12 |
+
(CASE WHEN book = '4caster' THEN least(coalesce(nxt, start_ms), ts + 90000)
|
| 13 |
+
ELSE coalesce(nxt, start_ms) END - 1) // 60000 AS m1
|
| 14 |
+
FROM (SELECT *, lead(ts) OVER (PARTITION BY matchKey, market, book, contract, lvl, side, line ORDER BY ts) AS nxt
|
| 15 |
+
FROM read_parquet('/home/stats/@S@/xraw/kick_day=@K@/*.parquet'))
|
| 16 |
+
WHERE eff IS NOT NULL;
|
| 17 |
+
|
| 18 |
+
-- 2. Best live price (and the stake behind it) in every minute any offer was live.
|
| 19 |
+
CREATE TEMP TABLE top AS
|
| 20 |
+
SELECT matchKey, market, book, contract, side, line, m, max(eff) AS eff, arg_max(mx, eff) AS mx,
|
| 21 |
+
any_value(start_ms) start_ms, any_value(home) home, any_value(away) away
|
| 22 |
+
FROM (SELECT *, unnest(range(m0, m1 + 1)) AS m FROM iv WHERE m1 >= m0)
|
| 23 |
+
GROUP BY matchKey, market, book, contract, side, line, m;
|
| 24 |
+
|
| 25 |
+
-- 3. Runs of an unchanged best price; a minute with nothing live ends the run with a NULL row.
|
| 26 |
+
CREATE TEMP TABLE runs AS
|
| 27 |
+
WITH g AS (
|
| 28 |
+
SELECT *, lag(eff) OVER w AS prev_eff, lag(m) OVER w AS prev_m FROM top
|
| 29 |
+
WINDOW w AS (PARTITION BY matchKey, market, book, contract, side, line ORDER BY m)),
|
| 30 |
+
c AS (SELECT *, CASE WHEN prev_m = m - 1 AND prev_eff = eff THEN 0 ELSE 1 END AS chg FROM g),
|
| 31 |
+
r AS (SELECT *, sum(chg) OVER (PARTITION BY matchKey, market, book, contract, side, line ORDER BY m
|
| 32 |
+
ROWS UNBOUNDED PRECEDING) AS run FROM c)
|
| 33 |
+
SELECT matchKey, market, book, contract, side, line, run, min(m) m_first, max(m) m_last,
|
| 34 |
+
any_value(eff) eff, arg_min(mx, m) mx, any_value(start_ms) start_ms, any_value(home) home, any_value(away) away
|
| 35 |
+
FROM r GROUP BY matchKey, market, book, contract, side, line, run;
|
| 36 |
+
|
| 37 |
+
COPY (
|
| 38 |
+
SELECT matchKey, market, book, contract, side, line, m_first * 60000 AS ts, m_last * 60000 AS ts_last,
|
| 39 |
+
start_ms, home, away, eff, false AS del, mx, '@K@' AS kick_day
|
| 40 |
+
FROM runs
|
| 41 |
+
UNION ALL
|
| 42 |
+
-- nothing live the minute after this run (and the next run doesn't start right away)
|
| 43 |
+
SELECT matchKey, market, book, contract, side, line, (m_last + 1) * 60000, (m_last + 1) * 60000,
|
| 44 |
+
start_ms, home, away, NULL, false, NULL, '@K@'
|
| 45 |
+
FROM (SELECT *, lead(m_first) OVER (PARTITION BY matchKey, market, book, contract, side, line ORDER BY m_first) nxt FROM runs)
|
| 46 |
+
WHERE coalesce(nxt, 9e18) > m_last + 1 AND (m_last + 1) * 60000 < start_ms
|
| 47 |
+
) TO '/home/stats/@S@/xtop/kick_day=@K@/xtop.parquet' (FORMAT parquet, COMPRESSION zstd);
|