Catch up on the latest updates and improvements to the system.
Thanks to Jacob for alerting us to Brighton and Hove stadium no longer holding races on Sundays - see Hove racing calendar.
We will run some trial predictions for our courses on Sundays and add the best to the service.
Pleased to say we have resolved the issue with past results. We now have the results for October and November available on Full Results.
We have all our previous results since 2015 but need to process these into the new format to be displayed on the site. Please contact us if you need any specific results.
We have identified an issue with our results service. Results for Sheffied were incorrectly being attributed to other courses. We are working hard to remedy the underlying fault and repair our historical data.
The tips and previous day results are unaffected. Please accept our apologies for any inconvience caused.
The full result history for our tips in June and July are now available on the Results page. Full Results
We have made good progress with our system this year. At our featured veneus we have delivered strike rates of 24% at Nottingham and Newcastle, and 22% at Hove. We will continue to improve our predictions and are working on adding more courses to our system.
After a catastrophic failure we are very pleased to confirm that our unique Artificial Intelligence tipping service is back and better than ever. Please accept our apologies for the disruption. The complete set of results is now up to date. Full Results
We have fixed a bug and now the results for November for our free greyhound racing tips are on the results page. Our apologies for the delay.
We now have available all the results of our tips since July 2015.
Let us know what you think and if you spot any problems.
Yesterday we implemented a fix on the results page. There was an issue with races that included a
substitution or for when only the first 3 places were known at the time the site was updated.
These results are now displayed as follows
We thought this was really cool (but we like this sort of thing) Tinker With a Neural Network Right Here in Your Browser
We have implemented a usability fix for visiters on mobile devices to our tips page.
Let us know if there is anything you would like to see improved on the site Contact Us.
Latest updates have improved the results pages for mobile users. As requested we have added a chart of the monthly result data along with the graph.
We have reintroduced tips for Wimbledon to cover the Greyhound Derby (semi-finals tonight). We do have not past history for predictions at Wimbledon so we are really pleased to see it pick the two favourites for their heats. Remember, these tips do not include the betting so these predictions are entirely based on form.
If tips at Wimbledon prove accurate we will leave them running. Swindon, however, has dropped off the list - we can never say why but the system sometimes just stops working for some courses.
From tomorrow our latest improvement to the system will be introduced. In testing so far this year 29% at Nottingham, 28% Sheffield,
26% Newcastle and Brighton & Hove.
We have replaced the Newcastle tips with a more precise model following feedback it was too vague to be of use.
We have reintroduced predictions for races at Brighton and Hove from today. We have also added an extra chart to the statistics page - results for the past month. The chart has a black line for random selection 16.6% - 1/6. The chart only shows results where they was a clear favourite predicted. The tips for Newcastle use a Bayes classifier which predicts general rather than absolute positions.
We have implemented an improvement to the tips page ready for tomorrow's predictions. Only tips for the required venue are now loaded avoiding lots of scrolling for mobile users.
Our redesigned site is finally ready to go live. This new version will better handle different devices and provides a cleaner less clutered look and feel.
One of the first methods we experimented with for GreyhoundPredictions was a Naive Bayes classifier. Bayes theorem is used to separate the dogs according to the characteristics of their previous results. It makes the assumption that there is no relationship between any of these characteristics. This may sound counter intuitive but can deliver some really useful results.
We found that for UK greyhound racing Naive Bayes Classifier can make a good first cut of the days races and to filter out outlier results from other machine learning techniques.
We really liked this simple explanation of Bayes Theorem using Lego.
The Theory That Would Not Die is a good read on Bayes - it makes some pretty wild claims but covers the history of the theory well.