February 20, 2003 at 20:49 #3902adminKeymaster
- Total Posts 739
— Posted by Seagull on 9:16 am on Feb. 10, 2003<br>Dear Peter, <br>With regard to trainers trends in your view how many seasons do you think best to use when forming trends. <br>I think 10 is to many as they become very stale and 3 years is not enough. I think 6 is ideal.
Do you think profitable trainers trends are always better to follow than jockey trends? <br>Thanks for the time you have given.
In my opinion five years seems to give the best results. Ãƒâ€šÃ‚Â I used five years in the Trainer Trends column and base all my Trainer statistics on that period. Ãƒâ€šÃ‚Â However it is important to check for significant changes, such as a main owner leaving a yard, or a change of location. Ãƒâ€šÃ‚Â For example, Jonjo OÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢NeillÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢s move South will significantly change the pattern of his winners with respect to course.
I tend not to follow jockey bookings, though I am looking into this area for a new book.
I hope this answers your question.
— Posted by JAR on 10:39 am on Feb. 10, 2003<br>Dear Dr Peter May
I notice that there are several computer-based forecasting methods on the Internet. <br>Boxform, Compunter, Wizard-Pro, Pro-Punter, Punters-pal, Brimardon and Denville rater – to name a few.
Some require only sparse input, while others require a great deal of input.
Has anyone done a comparison of these forecasting computer programs?
Have any of these programs been proven to work by an independent source? Which are the best? <br>Do you think they have any value?
I have seen a selection of these programs, and I am afraid that I have not been that impressed. Ãƒâ€šÃ‚Â Most seem to work on the standard linear approach of applying weights to key factors which, in the long run, is unlikely to succeed. IÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢m not sure whether anyone has conducted an independent test, perhaps SmartProof did before its demise. Ãƒâ€šÃ‚Â
Personally I would not use any of these programs. Ãƒâ€šÃ‚Â A better approach is to buy the relevant literature that will help you construct your own system, this way you can continue to improve the product over the seasons, incorporating the factors you feel are important. Ãƒâ€šÃ‚Â At present I am working on a method for my 2yo Ratings Software that will generate an odds line based on the ratings and sire and trainer trends. Ãƒâ€šÃ‚Â It uses neural networks and impact values and the results from early tests are encouraging. I feel that hybrid systems of this type will probably offer the best solution since they can borrow from the best of two the approaches and not be restricted to a linear solution. Ãƒâ€šÃ‚Â Naturally you do not need to use Neural Networks or Impact values, for instance you could combine an expert system approach with a statistical method. Ãƒâ€šÃ‚Â I hope this helps.
— Posted by winchy on 1:57 am on Feb. 11, 2003<br>When looking at different sources for speed ratings, <br>Raceform, Timeform, Racingpost etc. Why do they not all give a similar representation of speed trends and speed. Often one rating will look exeptional with one or two different sources but the others seem to miss out. Are they massaging the figures to the individual animal and not really giving us the true rating at all, or keeping it for themselves. <br>Surely all the graph trends should be the same, giving us a similar trend, but when I plot them on a graph, they are not, even if the figures and scales are different.
Speed figures are generally calculated in a standard way, though this leaves a great deal of scope for the handicappers to produce differing figures. Ãƒâ€šÃ‚Â Standard times, for instance, vary from one organisation to another, similarly some will have different approaches to assessing the going allowance. Ãƒâ€šÃ‚Â Unfortunately some ratings providers do massage their figures simply to get a specific horse top-rated. Ãƒâ€šÃ‚Â PostMark is the worst. Ãƒâ€šÃ‚Â Arbitrary amounts are added to the ratings of some horses simply to make sure they come out top, this makes the whole rating process redundant and reduces the service from one of ratings to one of tipping, which is not what the user is expecting. Ãƒâ€šÃ‚Â
With respect to Speed Figures, I firmly believe that the conventional approach is flawed, especially the use of standard times. Ãƒâ€šÃ‚Â More recent approaches do not use standard times, which themselves are mere estimates and only serve to introduce more inexact data into the calculation. Ãƒâ€šÃ‚Â These alternative methods that do not use standard times are the ones to follow.
— Posted by malcolm on 11:03 pm on Feb. 12, 2003<br>dear peter <br>im very interested in learning the art of handicapping <br>and learning how people arrive at there ratings <br>eg postmark in the rp can you recommend any decent <br>literature to help me <br>regards <br>malcolm
Thanks for taking the time to write. Ãƒâ€šÃ‚Â From memory Braddock covers handicapping in his Guide, though for a more up to date approach I would suggest an internet search which also covers the American literature. Ãƒâ€šÃ‚Â I am sorry that I cannot recommend any single publication, though if I come across any IÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢ll be in touch.
— Posted by Bricoman on 11:52 am on Feb. 13, 2003<br>Hi Peter,
When I first started to use the online version of the Racing Post I had three winners at one particular meeting, just by picking the highest Topspeed rating with a Postmark that was within five points of the top Postmark. I just wondered if this was a fluke or is there something to it.
As with all my system’s, I never followed it through to see if would give a consistant profit.
Do you think this would be worth monitoring and are there certain race types I could discount?
This is a reasonable approach, and could work well on the All Weather. Ãƒâ€šÃ‚Â However I would suggest that you monitor it, recording other factors such as going, race grade and number of runners, then analyse the data with respect to these additional variables. The problem is that you will need to do this for about a season unless you can get historical data from the Racing Post. Ãƒâ€šÃ‚Â Another drawback is the reliability of the ratings. Ãƒâ€šÃ‚Â The personnel may change, or the method of calculation may alter, this will, in effect, devalue your results. Ãƒâ€šÃ‚Â Ideally you need to find a ratings provider that does not change his/her methods. Ãƒâ€šÃ‚Â Best of luck with your research.
— Posted by cormack15 on 10:00 pm on Feb. 13, 2003<br>Dr. May – <br>Statistics question, apologies if it’s not your field. I’m interested in calculating the probabilities of a system throwing up a certain number of winners in, say the next 100 or whatever races after it has, say, thrown up a certain number of winners from a given sample. Ãƒâ€šÃ‚Â I suspect it’s down to standard deviations and the like but would like som eadvice and pointers perhaps, <br>Thanks
Yes it is down to standard deviation and the binomial distribution. Ãƒâ€šÃ‚Â You will need to get yourself a statistics text and look at the sampling and confidence interval sections. Ãƒâ€šÃ‚Â However these texts will cover random samples, and your sample is biased in that the future races from no part of the population from which the sample is drawn. Ãƒâ€šÃ‚Â This is a major problem with systems as well. Ãƒâ€šÃ‚Â Though a system may have worked well in the past, and a statistical confidence interval indicates that a profit is almost guaranteed, this is only valid if the sample is taken from the population. Ãƒâ€šÃ‚Â However this cannot be the case with a system since the races in which we will bet are in the future. Whilst it is not strictly appropriate to apply such measures they do provide an indication of the likely spread of profit or winners success rate. Ãƒâ€šÃ‚Â
If you have no luck with the statistics book, please e-mail me directly and I will forward the relevant chapter from ÃƒÂ¢Ã¢â€šÂ¬Ã…â€œForecasting MethodsÃƒÂ¢Ã¢â€šÂ¬Ã‚Â
- You must be logged in to reply to this topic.