- This topic has 93 replies, 14 voices, and was last updated 19 years, 5 months ago by
.
- Posts
Once we had it we could work out what to do with it! <br> cormack15 Posted on 9:53 pm on Nov. 27, 2006
<br>I think that could be a mistake, you will need to know first which of these factors are of most value .. you could check this using an inferencing model.
- Total Posts 17716
<br>Here’s a wish list of items I’d want to know as a minimum about each race and horse who has contested a particular race (and I’d want ten years worth of data) –
Going<br>Number runners <br>Finishing position of the horse<br>SP<br>Racing Post Rating achieved in the race<br>Topspeed achieved in the race<br>Previous best RPR<br>Previous best Topspeed rating<br>Last three RPR’s (in same code) prior to race<br>Last 3 topspeed ratings (same code) prior to race<br>Trainer<br>Age<br>Weight carried<br>Draw (flat)<br>Previous best RPR over Course (returning null if not run over course)?<br>Previous best RPR over Distance (as above)?<br>Previous best RPR on going (as above)?<br>Days since last ran<br>Number previous runs that season<br>Number previous runs in total<br>Race completion percentage (jumps)<br>
Jesus H Christ, Corm!<br> Do you really think all this is necessary?:o <br>
Hi
I absolutely agree that we must isolate the most pertinent race trends. I/We will only use these to produce the ratings.
"All the possibly useful trend angles must be considered, then the relevant ones ticked for any given race. In this way the program will have a generic nature."
If cormack wants to balance this against an alternate approach, this will be a good test.
The Betfair example, (see previous link), has a hollow ring to it. We can do better.  It was just an illustration of the currently available, trends analysis display.
This week will have to resort to a pencil and paper to produce any figures.  It’s sure to whet our appetite.
Having said that, you? might surprise me.
byefrom<br>carlisle<br> <br>
(Edited by carlisle at 8:35 am on Nov. 28, 2006)
Cant see how analysing all those factors would lead to a conclusion that would be any different from market expectation over the long term.
lol – my wish list was just that! Intended more as a starting/discussion point.
Carlisle – I’m a bit confused (not for the first time)??
"If cormack wants to balance this against an alternate approach, this will be a good test."
I thought we were aiming for the SAME approach, arrived at after group discussion/debate?
Apart from being an interesting experiment in Race Trend analysis this thread may well prove an excellent study for those looking at the difficulties of effective communication!
Do we need a ‘project manager’ for this exercise? Carlisle – how about you taking that co-ordinating role?
Hi cormack
this is what effective communication is all about.  Nice to chat with someone interested in getting to the root of the matter.
Sorry if I am coming across in a confused way.  I just don’t want forum members to think I am being bossy and ignoring other peoples postings.  It’s good manners to acknowledge all GOOD contributions and maintain the thrust of the thread……. (in my book)
Seems to me that you would like to produce your trends via statistical analysis.  My view is "why keep a dog and bark yourself."  Lets use "the 3 Trendmeister" as our starting point. Add to and manipulate this data then present it in a <br>more useful & palettable display.
I’ll consider myself the Captain then, okay.  What have you done………..?
byefrom<br>carlisle
"There’s another side to the praise coin."
(Edited by carlisle at 10:47 am on Nov. 28, 2006)
A quick and dirty screen-scraping script pointed towards the RP’s form pages would pull most of that data out whilst you have a cup of tea.
OK Carlisle – good man.
I’ve done nothing so far!!!
"Lets use "the 3 Trendmeister" as our starting point. Add to and manipulate this data then present it in a <br>more useful & palettable display."
OK – so what data, exactly, are we talking about here, how do we get our hands on it and who will compile/collate it initially.
Gareth – you sound like a man in the know. I’d be interested in finding out about ‘screen scraping script’ development. Maybe you want to PM me.
Carlisle – you mention ‘3 articles’ in the opening post on this thread and that you could copy those. Can you copy them and send tehm to me – I’ll pm you my address, etc.
The record of winners LTO in the race being reviewed is sure to be high up on the trend list, in the last 10 years how many won having been successful in their last race, how many total losers having contested the race being reviewed won LTO? WE could apply filters to improve the % the then use Likelihood ratios to form a probability based on trends.
Is that the kind of thing we are looking at?
(Edited by Factorman at 1:12 pm on Nov. 28, 2006)
Hello chaps,<br>This is an angle I am very interested in and would be very willing to help. Unfortunately my downfall is an inablility to effectively extract the data from the available databases.<br>I can however, provide fairly strong statistical analysis on sample data sets, if they can be provided. If this can be useful to you guys then let me know and i’m sure i can fit it in.<br>Cormack has my email address, otherwise i’ll check this thread for any updates.
Nice work so far though.
Regards,<br>Ken.
Great Ken – I know we’ve spoken in the past about possible collaboration so this seems a good opportunity.
cormack15,
on a tangential matter, might it be an idea to set up a section of the forum where people can upload programs and applications that might be of interest to other members (assuming people would be interested in sharing)?
i use excel to a certain extent (betting records, pricing races etc) – its excellent for repetitive calculations… but i know i am vastly underusing its capabilities… i tried a few years back to use it to download from the Post website but couldn’t get round the password! this is mentioned here again and arouses my curiosity anew…
just a suggestion
Hi nore
you touch on some interesting stuff, I will get back to you at some point.
I have produced some scans. They are all about the Hennessey. They will give us an insight into professional trends analysis.  I will do this weeks on Wednesday.
Send me your email if you want to view them.  I am new to scanning so I hope they are okay.
byefrom<br>carlisle
Nore – I’ll have a think about that, seems a good idea though.
OK Carlisle et al – I’ve finally done something!
I think the first step is to identify a generic set of data that we feel should be examined for every race (as per Carlile’s statement on Page 1 – All the possibly useful trend angles must be considered, then the relevant ones ticked for any given race. In this way the program will have a generic nature.
I’ve reviewed the three trendors (get it!) and in their Henessey write ups here are the factors they highlight as key –
CT – <br>Class of race previously won<br>LTO form figure<br>No. Chase Runs<br>Distance previously won at<br>Age<br>Strike Rate<br>Official rating<br>Weight<br>Days since last ran<br>Running with a penalty?
AW adds -<br>Previous form in H’Caps<br>Influence of trial races (good ones + poor)<br>Trainer form with seasonal debutantes<br>No. of runs that season
SW  adds – <br>Type of race run in LTO<br>Raceform rating<br>Course form
In addition in the other races he reviewed on sat CT also highlighted – <br>Racing Post Rating<br>RP rating in relation to top rated<br>Price/position in betting market
Summary<br>Now, they may well look at other factors and have identified only those as key but I think there is the basis of a pretty comprehensive list there. I think we need a concensus on which items to include in our analysis before then moving onto the next step – working out how/who will get the historical data extracted.
I have also given some thought to the development of a trend ‘rating’ and think that’s an exciting concept. Don’t think it’s been done before so, a great idea Carlisle! More on that anon. Lets get stage 1 out of the way first. (and we don’t need to beat ourselves up and agonise too much over which data to include  – lets just make a decision, go with it and then we can pull back and add/take away data later as we see how things pan out).
ÂÂÂ
<br>
Hi cormack
your last posting sounds good. We can take our time with this project. It will be ace if we can eventually end up with a stand alone race trend rating.
This can need be viewed alongside any form ratings.
byefrom<br>carlisle
ps I’ll work on some stuff.  Scans anyone?
(Edited by carlisle at 12:28 pm on Nov. 30, 2006)
OK – the ‘easy’ data to get at includes –
Weight carried<br>Age<br>Trainer<br>Official rating<br>Starting Price
These snippets (some of them key) are all available from the result of the race as shown on the RP site so we should include those.
The next level of data requires us to examine the pre-race record of each horse so means interrogating the results sheet back a level. This can be done manually or by an automated process.
Now, a team of four/five of us could feasibly dig for quite a bit of data without inordinately laborious amounts of work.
However if someone can write a script to uncork the next level of data then let them stand up now!
My suggestions would be that we look at ways to extract the following which, from my limited knowledge of data extraction, may not be that difficult –
Strike rate<br>Days since last run<br>Distance previously won at<br>Class of race previously won<br>Best previous RPR<br>Last three form figures
More tricky perhaps might be the following (but useful nevertheless)
No. Runs that season<br>Previous Course form<br>Whether had won on the prevailing ground before
<br>Might be a combination of hard graft and computer work to assemble the data.
Thoughts please – perhaps we should pick a race (the King George?) as a model to work on?
- You must be logged in to reply to this topic.