|
Neural Ratings
Hi,
I found this site recently and been reading the old posts. A lot of useful information there. I’m interested in systems using the neural ratings from AAP R&S. I’ve been recording these ratings for over 14months now as well as the information on the Unitab website, prices, unitab rating, last 3 starts, etc. I’ve developed several systems based on this and was wondering if others are doing something similar. Rather than giving each criteria a weighting and summing them, I’ve recorded each value (normalised to 100) and the horse’s ranking for each criteria. This enables me to have a rule like, ‘Top 4 in the Current Form Score with a value > 75’, for example. I normalised them to 100 as there is sometimes great variation from one criteria to another, particularly with the Time rating. If anyone is familiar with these ratings, the prizemoney and career performance ratings are the best indicators of where the winners come from. I’ve got about 10 different systems that have made a good profit in the 6 months that I’ve been betting them. (Unfortunately, I’ve also been betting unproven systems at the same time which have made a loss and seen the balance pretty much the same now as it was 6 months ago after turning over about $100k.) Here is a very simple system: Metropolitan venues only Distance = 1200m Highest Time Rater (ignore race if 2 or more = top rater) <3 first starters <4 first up (first uppers don’t have a Time rating) Has a 20% SR from over 700 bets and 12% pot (Unitab div). Would be > 20% with best tote. I would be happy to run any other set of rules if you would like or just hear from anyone using the neurals on a regular basis. Cheers |
Getting back to the use of neurals, I've always put emphasis on the JT rating. How does it correlate to winners? I have generally ignored the timeform rating as it has a very high variance, particularly on weekdays but maybe I'm missing something. Cheers.
|
Hi Dirk,
Personally, I don't really use JT at all. However, nearly all of my systems need both the jockey and trainer scores to be highest or 90% of the highest. This usually results iin the selections being in the top 3 or 4 in the market. So I need a SR of 40-50% to make it profitable. I normalised each score to 100 before I store it in the database to make it easier to build systems around. Unfortunately, this means I can't use the DLR value to find out the days since last run as it is a direct correlation. EG a score of 30 (with weighting of 5) means it's last run was 9-10 days ago. Your right about the Time value. I think it is only useful at a certain distance range. Cheers |
Chinbok,
I have found setting the following values has been ok during the week. Trainer set to 3. Current Form set to 3. Prize Money set to 5. Wet Track rating set to 0 DLR set to 0 JT set to 0 All the rest set to 1. It doesn't work well on Saturdays. Essentially it is looking for good trainers that have a horse who has won good prizemoney and is in form. Current Form can sometimes skew the figures as it is like the time value and doesn't always come out worthwhile. But I found leaving it off produced a worse result. Can you run a test for above or if thats too hard can you check out horse must be in top 2 trainers, top 2 prize Money and current form in top 2. Thanks |
Hi Wesmip,
I can't remeber why but I didn't include JA, TA and JT ranking, only the score. So I've used a TA score of 90 for your test. I've also excluded 2yo and races with >6 first starters. Too many unknowns. In addition to these, are rules are top 2 prizemoney and top 2 current form. Past 14 months or so: All days: 3524 bets 1056 wins SR 30% Pot -10.7% Excl Sat: 2690 bets 808 wins SR 30% Pot -10.2% Appears to be no different on Saturdays. In fact Wednesdays were worst, SR 25%. Other days from 28-32% SR. Pot (loss) figures are based on Unitab divs. Best tote would give around a 10% better return on average. Which takes it close to break even. Pretty good for this many bets. Really needs more filters. I usually exclude fillies, fillies and mares, and mares races as they usually underperform. Over the 7 days, selections in these races performed: Bets 411 Wins 101 SR 24.6 Pot -28.4 Still needs more work though. Cheers |
Thanks
Chinbok,
Hmm, thanks for doing that . I guess I don't usually bet every race with it and I am much more selective. I usually use this as a guide to the top 2-3 horses to look at. But its good to see some figures, at least it had a strike rate of 30% too bad about the POT. If I think of anything else I will let you know. |
Quote:
Do you have any figures on what type of race work BEST? |
Chinbok,
Can you email me at aussiestable@yahoo.com.au I have another test which I was thinking about but I haven't done any preliminary testing. Try top 2 prize money and top ranked for distance as per AAP Neurals. The distance should be distance >= 1600m. Can you also excluded 2yo and races with >3 first starters. I am looking for horses that run well over the longer distance (thus top rated) and are a higher class (based on prize money) If possible can you do this for fields of less than 14 horses. As I haven't done any testing at all this might just bomb badly.... Thanks |
Hi,
The following rules: Top 2 $, Top ranked Dist, No 2yo, <4 1st starters, dist >=1600, field size <14, produced: Bets 2476 Wins 529 SR 21.4% Pot -12.4% Needs more. |
Sportz,
Don't really know what race types work best. I exclude 2yo races becuase early in the season there are alot of first starters and late in the season everyone is on them so you need a very good SR. Always exclude fillies - too inconsistent. I've heard this from other ratings providers as well. Sometimes F&M and Mares. I don't like to base systems on different classes. I think if it's a good system it should work just as well on maidens/class 1's as group races. Always exclude jumps races. I suppose the remainder are what work best for me. Cheers BTW, how do you quote part of another post? |
| All times are GMT +10. The time now is 03:43 AM. |
Powered by: vBulletin Version 3.0.3
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.