In the 6 months data I have run there are 7,000 occasions where this field is blank and every one is a loser. This looks very odd.
Also, for DIFFTPAEECLVOL there are 3,000 occasions where there is a number in this field and again every one is a loser.
That does seem odd Michael. I presume your logic is,
SHorBEGt18 measures the horses' best speed rating on slow going' (which applies to AW tracks only).
DiffTpSHorBEGt18 measures the difference from top-rated. If no horses in a race have run on the AW in slow going they would presumably, all have a blank rating. It wouldn't be surprising to find a high number of such races. However, one of the horses must have won such a race so there should be some winners among the blank ratings.
It might be worth isolating some whole races in that period where no runners would have run on the AW on slow going to see what SHorBEGt18 rating the winner has.
Yes, this now looks correct. Thank you.
This is as I would expect. Taking a random six months we get around 65,000 horses running, so we're around 11% of the horses don't have this rating, considering it's only for horses on Slow ground that's not outside expectations.
There's not a guarantee that any of those horses must have previously won a race on slow ground, but even if some of them have, this doesn't mean we'll be able to make a speed figure for the race. To make a speed figure we have to have a standard time, to have standard times, we need to have enough races over the course, distance, race type and going to be able to make an accurate figure. As an example, when we break down these conditions, some have only had 1, 2, 3 or 4 races in our entire database (around 2 million runners). We wouldn't make a standard time for these because it would be so inaccurate as to be pointless, which means we wouldn't be able to make speed figures for the race, and these fields would be blank.
Thanks Michael, that explains why there are so few ratings (a good case for dropping the rating altogether perhaps). Still curious why MC has no winners among the runners that are unrated.
I understand that a lot of horses will not have a rating. What doesn't make any sense is that they all ended up losing.
I have looked at these ratings for this month and they do look more sensible. It looks like it is another issue with the historical data.
Still curious why MC has no winners among the runners that are unrated.
@andrewp sorry not following this?
Please find attached file. Hopefully this will explain. All items where the rating is blank. Every 1 of the 4,313 horses lost. Either I have discovered an exceptional laying system or the data is incorrect.
I just realised what @andrewp was talking about, when he said MC I thought he meant Monte Carlo, but he meant you Michael 😀
There doesn't appear to be a single complete race in the sample Michael. Some other filter must have been applied. I can't believe there are 1,800 odd races and not a single one with all runners unrated.
I think Michael pulled out all horses except those without that rating.
I have looked at the base data in detail and 90% of the horses have a rating of 0, which indicates they are top or joint top rated.
I have looked at a 2 runner race where the horses are rated 0 and neither has run on slow ground before.
Okay I've dug into this a bit further, it's because this going description is almost never used.
2007 - 14 times
2008 - 0 times
2009 - 7 times
2010 - 56 times
2011 - 0 times
2012 - 15 times
2013 - 1 time
2014 - 1 time
2015 - 0 times
2016 - 0 times
2017 - 0 times
2018 - 1 time
2019 - 1 time
As for why they're all, or mainly, pulled up for some reason is probably just chance or perhaps you have stumbled on a very good laying system, but with average odds of 111 I wouldn't want to use it personally 🙂
So of the 4,313 horses, not 1 of them managed to complete the race. Are you seriously saying this is due to chance ?
It's probably academic as clearly this rating has no value whatsoever but,
Over a six month period 3,945 runners without this rating had BSPs indicating an expected 306 winners and not one of them finished a race let alone won it?
Looks like something might have happened post race. For example, if all runners are without a run on slow ground perhaps they are all given a joint zero and this has been removed from non-finishers and replaced with a blank.
We will look into it further next week.
Any update on this ?
I think this may have been a memory limit issue as well. Our developers believe it has been corrected by adjusting the scripts to not cause a memory error on the server.
Querying the database for 2019 there are 349609 horses with NULL that finished the race, and 35939 that didn't.
In 2018 29433 didn't finish from 165994 that did.
I have re-run the data for the 6 months as per my initial analysis and the data is the same, nothing seems to have changed. 4,313 horses, every one didn't finish the race according to the data.
I just exported your race card Michael, and I may be looking at things wrong, but there seem to be lots of finish positions for those horses now.
Please let me know if I'm miss-reading your export.
Is this for the period 1.11.17 to 30.4.18 ?
@michael-clarke yes it is, you should see it in your Exports.