Comprehensive Historical Horse Racing + Betfair Exchange Dataset for XGBoost

[Attention]
I need someone who already has this kind of dataset prepared. not from scratch.

I’m looking for an experienced horse-racing data specialist to provide a clean, comprehensive runner-level historical dataset for training an XGBoost model to predict horse win probability for pre-race Betfair Exchange analysis.

Each row should represent one horse’s run in one race. I need as much reliable historical information as possible about the horse, jockey, trainer, race and pre-race market.

Horse information:

Horse name and unique ID
Age, sex
Official/handicap rating
Weight carried
Previous form/results
Speed/performance ratings
Career runs, wins and places
Earnings/prize money
Days since last run
Course/distance/going record
Recent performance/form
Any other useful historical horse variables

Jockey information:

Name/ID
Career and seasonal runs, wins and places
Win/place percentage
Average finishing position
Recent form
Course/distance performance
Horse-jockey history where available

Trainer information:

Name/ID
Career/season runs, wins and places
Win/place percentage
Recent form
Course/distance performance
Trainer-jockey and trainer-horse history where available

Race information:

Race ID and date/time
Course/track
Race type
Grade/class
Distance
Going/track condition
Surface
Field size
Draw/stall
Prize money
Handicap/non-handicap
Age restrictions and weight conditions
Any other available race-level information

Results:

Finishing position
Winner indicator
Finish time
Winning/finishing margin
Official result
Speed/performance ratings

Betfair Exchange pre-race market data:

I only use pre-race information, so historical timestamped Exchange data before the scheduled off is particularly important.

Ideally, provide the full pre-race price history, including at minimum snapshots around:

10 minutes before the off
5 minutes before
2 minutes before
1 minute before
30 seconds before

For every runner/timestamp, provide where available:

Back price
Lay price
Available back/lay volume
Traded price
Traded volume
Market depth/liquidity
Exact timestamp
Betfair market ID
Runner/selection ID
Scheduled race start time

Full timestamped market data is strongly preferred because I want to calculate price movement, volatility, liquidity and other pre-race features myself.

Data requirements:

Minimum 5,000 unique races
Preferably several years of UK/Ireland racing
One row per horse per race
No duplicate runner/race records
Stable IDs for races, horses, jockeys and trainers where available
CSV or Parquet
Clean numeric columns
Complete data dictionary
Documentation of preprocessing/imputation
Clear source and licensing information

Critical: all predictive features must be based only on information available before the race. No post-race information should be used to construct features or historical statistics in a way that causes data leakage.

I am interested in the most comprehensive reliable dataset available, not simply the largest number of rows. If you have additional useful horse, jockey, trainer, race, rating or Betfair variables, please include them.

Please provide a sample of the actual dataset, complete column list/data dictionary, source, coverage/date range, approximate number of races/runners, Betfair price resolution, licensing terms, delivery time and price.