Unrated Races

This morning I mentioned (in my Morning News Post) that I currently don't publish the ratings for any races where I can only rate less than 40% of all horses in that race.

My thinking, when I started Focus Ratings, was that if I was not able to rate enough horses in a race then there was more risk to that race.

I did some very basic analysis and, to make things easier for me, I set up the spreadsheet to round the percentage of runners that I could rate to the nearest 10%.

My Mistake

That was mistake number one - I should have rounded the percentage of runners that I could rate to the nearest percent.

The analysis was that the strike rate dropped off below 40% - hence the 39% figure.

However, since then (I started Focus Ratings on the 7th of January 2013), the ratings have improved in complexity and accuracy as I've refined the code.

That work mainly occurred up to the 7th of July 2013.

Since then I've done bug fixes (and will continue to do so - with 120,000 lines of code, and growing every day, there will always be bugs; especially code written by me!)

But, since the introduction of the Builder I can't really change how I work out R4 as, if I do so, it might break your systems and as you'd be betting your children's inheritance on those system selections, they'd lose and you'd be forced to send your kids down the pit, auction off one of your mother in law's kidneys, sell your wife (or husband) to the white slave trade and, as a last resort, get an evening job, down the docks, pleasuring sailors for 50p a time.

And I'd feel more than a little guilty about that.

Self Learning Ratings

Now, the ratings are self learning (in 2 ways)...

1). They rely heavily on a horse's previous rated races (and this is the biggie!)

2). The R2 level of the ratings is constantly being refined as the core data increases.

So, there's logic to publishing the currently unrated races - more data for Cortex (which is what I call the logic engine behind Focus Ratings) equals more accuracy so...

Not only do we get more wins (from the currently unrated races) but, we also get more accuracy across the board.

Analysis

I've done some analysis on the unrated races since the 7th of July 2013.

Since the 7th of July 2013 there have been 661 unrated races.

168 were won by the top rated horse = 25.42% Strike Rate.

In 542 of those races we had a second rated horse.

87 of those races were won by it = 16.05% Strike Rate.

In 352 of those races we had a third rated horse.

27 of those were won by it.

Now, these figures understate the reality as I haven't been able to take non-runners into account.

The average price of a top three rated horse (which wins) in the unrated races is 4.28 (ISP.)

And that's just looking at win strike rates.

For the place bettors, things are better.

From the 661 unrated races the top rated horse came 1st, 2nd or 3rd in 423 of them (which equals a 63.99% Strike Rate.)

The top three rated horses from those races came 1st, 2nd or 3rd 722 times (from 661 races.)

Decision Time

So, it's decision time.

My feelings

I am very keen to add the currently unrated races as...

a). They'll give more winners to those of you who like to back winners.

b). There's no logical reason not to rate those races.

c). The additional data will help Focus Ratings learn more about the horses and, thus, will improve the overall accuracy of the ratings in the months and years to come.

d). For you Place Betters, you'll get more than one extra place result from each, currently unrated, race.

Caveats

The downside is...

a). It may slightly reduce the top level strike rate for the top rated horse in the short term - on the other hand, if it gives us a strike rate of 25.42%, it may help us on days when we do badly (strike rate wise.)

b). I'll have to identify those races that were previously unrated on the PDFs by putting the Race Detail line in italics, just so that you can spot those races.

c). I'll have to add a new RC0 level of 4 to further identify those races.

d). I'll need to add a small amount of code (less than 1,000 lines) to the Builder to ensure that your systems don't get broken (I'm already having nightmares about you having to go down the docks etc.) The default will be not to include the previously unrated races but...

You could always build new systems and decide to include those races.

Over to you

And so, it's over to you.

If any of you can come up with a decent reason why I shouldn't start including the previously unrated races, please let me know.

Otherwise, I'll draw up a schedule to start including these races.

I can only do one system change a day (if I do more, it all goes horribly wrong and that stresses the hell out of me!)

So it might take a week to get it all in place.

But...

I think that it's the way forward - the main thing for me is that the extra data will improve the accuracy of the currently rated races and...

Give you more opportunities to win.

One last thing...

And I do apologise for the picture below...

It's just that it makes me laugh and, after 11 hours of number crunching, I do need a little chuckle.

And it further reminds me that I really can't cock this one up (and end up behind the bar with my Creme de Menthe spilled on the floor.)

However...

If I can rate 80% of all horses in a 20 horse handicap, that means that I am showing you 3 selections from a field of 16.

If I can only rate one horse from a five horse non-handicap (in other words, a previously unrated race), that means that I am really picking out the only horse that has some form (and market attention.)

If you had to bet (and, remember, you don't have to bet on every race), which one would you put your sixpence on?

Once again, if you've got anything to say about Focus Ratings, please wander over to Focus Ratings Review and have your say.

As always...

My kindest regards