Major Changes to Simple Software Program and Database

[mickjohnson]

When I started Simple Software Racing in 1994, it was based on Davey Towey`s book `The Solidus`, under his direction. We input the results data ourselves from the Racing Post and the Sporting Life newspapers.
Davey didn`t like the All Weather and there were no co-efficents for the Irish courses in his book either. Some of the figures in the book were changed by Davey in Biro in the copy I worked from.
When he disappeared to Ireland, I managed to get a daily data supply, continued with the MS-DOS version and wrote the first Windows version of the software.
Obviously, in those 20 years, courses have changed, distances have been added and rails moved, and, of course, new courses added.
I interpolated and extrapolated the figures I had, to calculate the co-efficients I didn`t have but it was very difficult and didn`t have much confidence in them, but they were as close as I could get.
Since I added the actual speed of a horse, in miles-per-hour during a race, to the database last year (I wanted to see how the age of a horse affected it`s speed), it produced loads of negative ages.
I found thousands of horses that have their names re-used in later, younger, horses. The older horses had I or II added to their names and the newer ones were left the same.
I corrected them all and now check daily, when I get my results, to see if there are any more. I correct them before I send the results out to my customers.
I recently looked at the maximum and minimum speeds of horses in the last 20 years and found even more errors in the database.
There must be a phone call involved somewhere in getting the race times from the course to my data supplier. There were hundreds similar to, 235 seconds instead of 2m 35s (should have been, 155 secs) and 123 seconds instead of 1m 23s (should have been 83 seconds).
I also took out `one horse races`, or `walk-overs` from my calculations, as the horse must still finish to win the race but the times are usually very, very slow and cannot be considered useful performance indicators. One showed up as 17mph, about half top speed!
I had one recent race marked `Too Fast`. I checked the times with the Racing Post and the time, going and distance were correct. I looked at the Sporting Life and the time shown was a few seconds slower. As it didn`t seem to be a particularly fast race, I used the slower Sporting Life time.
I walked the course at Epsom a few years ago and noticed that, whatever the official going, some parts of the course were hard at the top of the hills and soggy at the bottom.
I realised that I needed to quantify the actual going coefficient for each distance and each official going value as different race distances at each course were run over different parts of the course. Having the actual horse speeds in the database helped me calculate them all.
The motivation of the changes I have made to the program and database, was the change of the Poly Track at Wolverhampton to the Tapeta surface this year.
I needed to calculate all the coefficients of the new surface from scratch, the going, standard times, everything.
I discussed it with some of my customers and decided to rename the old Poly Track races` coefficients `Wolverhamptom(PT)` and leave the new Tapeta as `Wolverhampton(AW)` and recalculate all the
coefficients using just my own calculations.
I recalculated all the going coefficients for each distance and going using the speeds of the winners of each race (except walk-overs, see above!).
I recalculated all the standard times using, again, just the speed of the winners.
I also calculated a Jockey co-efficient. Whether it is skill and experience, or wrecklessness and inexperience, I guessed that some jockeys ride faster than others.
They might take advantage of faster bits of a course that they know very well. They might only ride faster horses. Whatever! I did notice that some `famous` jockeys are significantly slower than average.
I removed the `jockey effect` when calculating the ratings after the race, and added it back in again, with, posibly a different jockey coefficient, to assess a future race.
As I am more confident in the accuracy of my data now, I did that for ALL the races of all the courses since 1994.
These are the figures I use now in my `default` assessment.
It is early day`s but my win hit rate percentage so far this month is pretty good and I only finished all the calculations a week ago!
I think I`ve made the correct decision.

[mickjohnson]

One of my customers spotted that I had a horse in the database that was over 500 years old! It fitted all the other criteria, ‘Older than its offspring’, etc, but, as it hadn’t run before, it didn’t show up in the usual age checks. I normally subtract the foaling date from the racedate and get the age of the horse at the time of the race, but, it hadn’t run in the UK or Ireland in the last 20 years. That shows up horses that with an incorrect foaling date, with negative ages, but didn’t show up horses that are too old.
I ran a query looking for horses older than my default ‘No Foaling Date’ of 01/01/1900 and found another 27. I corrected them and that showed up another 7 related horses’ errors.
Due to adding a few more queries to the software, I’ve managed to automate the correction of most of the errors, as they arrive, including the renaming of older horses (adding an I at the end) and adding the new horses’ details.
It has all worked well, so far!

[Author]

Posts