DataBase Ratings: November 24th, 2016

Punting Pointers

PDF Download

Excel Download


Welcome to a new free feature on the site:

Ratings for all TAB races every day posted as a free download.

They are produced using a machine-learning ensemble approach, that is the use of 25 separate models to come up with a single, overall prediction. The model uses over 40 variables related to runner-specific factors, race-specific factors and track-specific factors. The model is retrained on a weekly basis to ensure it keeps up with any trends or changes in racing.

We’re expecting people to use them in very different ways, but here are a few pointers:

Rank    SR
1          28%
2          17%
3          14%
4          11%

Quinella in top 4 rated: 40%
Trifecta in top 4 rated: 21%
Top rated loses approximately 3% at Best Tote.

  • Races with unraced runners are excluded
  • Races with a clear spread in the ratings (ie 1 or 2 horses well clear of the rest) can have some good value.
  • The rated prices are there as a guide and not there to be bet as strict overlays, if a horse is rated well clear of the rest (i.e. 20 pts clear) and it is rated $2.80 but the market is trading at $2.70 don’t be afraid to back it because it is under the rated price. In the same way if the market is trading at $1.60 that’s probably too short to have a bet on.
  • Overlays are best used for metro racing and to be more specific the higher class metro races (i.e. stakes races).
  • There are no real trends in distance ranges other than a higher degree of accuracy for races 2300m+.
  • The ratings don’t include variables related to the run the horse will get, e.g. the model won’t know that a horse will map to get caught 3 wide and also obviously there is no information contained about any track bias/patterns that presents itself on the day.
  • Our in-house analysts such as Nathan Snow, Trevor Lawson or Adam Mintz do not use these ratings as the basis for their form study. They go about things their own way and get to know every horse, jockey and trainer as individuals. These ratings were developed externally by a data scientist rather than a form analyst.

Post a comment below if you have any questions at all about how they are generated or how you might use them.

Good luck!