Class/form horses

[Graeme]

Hmmm - what do VDW and AI have in common - they both taste like Marmite - you will either love it or hate it!

JennyK I wonder if your friend uses AI to help him or her code or if they are talented to do it old school.

For me that is the true benefit of AI to do things I simply do not have the inherent skill to do - as for VDW - well I don't like Marmite But each to their own and live and let live

 

[Tufnel]

Hmmm - what do VDW and AI have in common - they both taste like Marmite - you will either love it or hate it!

JennyK I wonder if your friend uses AI to help him or her code or if they are talented to do it old school.

For me that is the true benefit of AI to do things I simply do not have the inherent skill to do - as for VDW - well I don't like Marmite But each to their own and live and let live

I was just going to say the same thing to JennyK . I used AI to design my database and then the code to use it and display races laid out as I designed them. The ARs are calc'd every time I load a new day's results, as are other stats and info about the previous opposition in races. It even does the consistency ratings and Erin (as I saw them).

But the issue gets tricky when the source of the data changes at their end. There's a very good github project that has code and help on getting the daily cards and also the race results, etc. When the source changed a while back it caused problems but the project updated their code to account for this. But it's just part and parcel of the ongoing maintenance that is required from time to time.

It probably took me about 3 weeks to get mine up and running and I often had to solve issues by checking things with different AI engines. At some point during the design and coding, the AI can trash your code when it goes off on it's tangent and it get's frustrating. But what I found was switching to a good rival bot could solve problems the other one had inflicted. As bots have improved that is less of a problem now and OpenAI seems very stable for me now and it's recall on my database and project is pretty reliable.

That's coding though. It's a right or wrong equation. Using AI to predict sports results, etc is very different and this is where it gets very unreliable, especially if you allow it to grab info around the web rather than a dedicated data set on your drive. Many a time I thought it had cracked a VDW angle or race evaluation only for it to confess it had put in placeholders or used dummy data because it couldn't actually get the right info from racing sites. And even if you save your instructions for a chatbot and paste them in the next day, it will still reinterpret them and send you down a blind alley.

 

[Graeme]

T Tufnel depends on what information changes in the source I guess I try to use column headers instead of column references for example so if the order of data changes the lookups still work but appreciate that does not sort issues like scraping websites etc

 

[rave154]

But the issue gets tricky when the source of the data changes at their end. There's a very good github project that has code and help on getting the daily cards and also the race results, etc. When the source changed a while back it caused problems but the project updated their code to account for this. But it's just part and parcel of the ongoing maintenance that is required from time to time.

i know the feeling,

you should see my code to parse out the data from the Racing Post results html

 

[mlmrob]

an excellent insight to handicapping by @

mlmrob on wizards VDW thread.

That was some years I ago I believe, will have to refresh my memory now, as I have to most days.

 

[JennyK]

" I wonder if your friend uses AI to help him or her code or if they are talented to do it old school."

He has been a self-employed Excel specialist for the ten or more years I've known him. Graeme, so in your terms old school. Whether he now uses AI for the projects he is commissioned to do, I don't know but next time I'm in touch I'll ask him. Oddly, it did not occur to me to ask him the question about using AI to get VDW ability ratings from the Post's website that I've asked here, so AI never came into mind in his context, but I now it has I imagine he'll know quite a lot about it.

 

[Graeme]

" I wonder if your friend uses AI to help him or her code or if they are talented to do it old school."

He has been a self-employed Excel specialist for the ten or more years I've known him. Graeme, so in your terms old school. Whether he now uses AI for the projects he is commissioned to do, I don't know but next time I'm in touch I'll ask him. Oddly, it did not occur to me to ask him the question about using AI to get VDW ability ratings from the Post's website that I've asked here, so AI never came into mind in his context, but I now it has I imagine he'll know quite a lot about it.

I expect he will - I would have deemed my self old school (still primary school actually) when it comes to Excel but AI has massively increased my skills so I am sure he will use some version of it at some point.

 

[Tufnel]

" I wonder if your friend uses AI to help him or her code or if they are talented to do it old school."

He has been a self-employed Excel specialist for the ten or more years I've known him. Graeme, so in your terms old school. Whether he now uses AI for the projects he is commissioned to do, I don't know but next time I'm in touch I'll ask him. Oddly, it did not occur to me to ask him the question about using AI to get VDW ability ratings from the Post's website that I've asked here, so AI never came into mind in his context, but I now it has I imagine he'll know quite a lot about it.

Does your XL sheet access a separate database or sheet for the form/past results you've copied in?

If so your friend should be able to do a an xlookup and calculate each runner's ability rating based on races they won. They'd need to check race type/code to keep it flat or jumps and obviously sum up the win prizemoney and divide by the number of wins found. But you'd need the career form in the database.

 

[Sol Ski]

JennyK,
Its your game so its your rules, if you enjoy what your doing keep doing it.

 

[JennyK]

The book I use when analysing a race starts with six sheets, T Tufnel, two of which are hidden. The first is for when I am analysing an as yet not run race and is for copying the race card into. I then click a button and the raw data is processed into a second sheet, where if it is a future race I press one option (if I am analysing or re-analysing a past race, I choose the other). Either way, pressing a button causes the book to prune my database so that what remains is just the data rows for the runners, and then processes that data into a third, working sheet (with two grids) and creates a career sheet for each runner.

The three remaining sheets from the original six are a database of the Post's comments for every race for each runner, and two very small hidden reference sheets, one the list of the last year's 3yos I keep tabs on (as per the Lists thread), the other my time-based assessment of the various "power" courses, showing their ranking in order of difficulty of finish (only for 5f, 6f and 7f races) as an aide memoire.

Once I have completed the two grids on the working sheet, via three macros and a data transfer of ARs (in various forms) from another book, I'm ready to start. (To get to this stage takes about five minutes if it is a race not yet run, and about three if I an analysing, or more usually re-analysing, a past race.)

The creation of the two grids on the working sheet includes automatic checking of my last year's 3yo list and flagging in one of the grids if one is running.

I draw data from the Post comments sheet so that it is added to the career sheet for any horse where I feel a need to check that. From each career sheet I can also by macro generate up to four additional sheets for the horse concerned, all of which excludes any non-handicaps. So for handicaps only, these show: in sheet 1the handicaps by year with an analysis of performances for each year based on my ratings, so I get a sense of trajectory; sheet 2 an analysis by distance (for which my main use is, obviously, sorting out whether a horse has a decided preference, judging from results, for 5f or 6f); sheet 3 an analysis by going, and sheet 4, my latest addition, compares handicap performances against market position (which I hope will help me judge when I am on solid ground backing an outsider).

If I really went to town, which I never, ever have felt the need to do, from my original six sheet book I lose one (the database) and gain potentially five sheets per runner, which if I were ever foolish enough to try to do the Cambridgeshire could theoretically mean a book of over 150 sheets! Typically, I only generate the supplementary career sheets for two or three runners that the top grid of the working sheet shows me are possibles, and then usually only one or two, so for a ten runner event the final book contains 15-20 sheets.

I know enough to make small adjustments to some of the coding, for example 3yo only races have a specific colour code so that I know if a race in a horse's career sheet was a 3yo only, and literally only yesterday I changed the colour coding for that. But I know my limitations and mostly have no idea what the line after line of VBA means or does, just that it works.

 

[Tufnel]

The book I use when analysing a race starts with six sheets, T Tufnel, two of which are hidden. The first is for when I am analysing an as yet not run race and is for copying the race card into. I then click a button and the raw data is processed into a second sheet, where if it is a future race I press one option (if I am analysing or re-analysing a past race, I choose the other). Either way, pressing a button causes the book to prune my database so that what remains is just the data rows for the runners, and then processes that data into a third, working sheet (with two grids) and creates a career sheet for each runner.

The three remaining sheets from the original six are a database of the Post's comments for every race for each runner, and two very small hidden reference sheets, one the list of the last year's 3yos I keep tabs on (as per the Lists thread), the other my time-based assessment of the various "power" courses, showing their ranking in order of difficulty of finish (only for 5f, 6f and 7f races) as an aide memoire.

Once I have completed the two grids on the working sheet, via three macros and a data transfer of ARs (in various forms) from another book, I'm ready to start. (To get to this stage takes about five minutes if it is a race not yet run, and about three if I an analysing, or more usually re-analysing, a past race.)

The creation of the two grids on the working sheet includes automatic checking of my last year's 3yo list and flagging in one of the grids if one is running.

I draw data from the Post comments sheet so that it is added to the career sheet for any horse where I feel a need to check that. From each career sheet I can also by macro generate up to four additional sheets for the horse concerned, all of which excludes any non-handicaps. So for handicaps only, these show: in sheet 1the handicaps by year with an analysis of performances for each year based on my ratings, so I get a sense of trajectory; sheet 2 an analysis by distance (for which my main use is, obviously, sorting out whether a horse has a decided preference, judging from results, for 5f or 6f); sheet 3 an analysis by going, and sheet 4, my latest addition, compares handicap performances against market position (which I hope will help me judge when I am on solid ground backing an outsider).

If I really went to town, which I never, ever have felt the need to do, from my original six sheet book I lose one (the database) and gain potentially five sheets per runner, which if I were ever foolish enough to try to do the Cambridgeshire could theoretically mean a book of over 150 sheets! Typically, I only generate the supplementary career sheets for two or three runners that the top grid of the working sheet shows me are possibles, and then usually only one or two, so for a ten runner event the final book contains 15-20 sheets.

I know enough to make small adjustments to some of the coding, for example 3yo only races have a specific colour code so that I know if a race in a horse's career sheet was a 3yo only, and literally only yesterday I changed the colour coding for that. But I know my limitations and mostly have no idea what the line after line of VBA means or does, just that it works.

Ok so it sounds like the comments are a data sheet on their own and the other past race data is on another sheet. Presumably that data has a field with the prize money to the winner of each race in the data?

If that’s the case then it is possible to automate the calculation of ARs and have them appear on your cards sheet for each runner.

Mine are calculated automatically when I update results. The ARs are stored for each runner as they are today and also for each runner in a race as they were before the race was run. But that’s because I’m using a dedicated database that sits on my drive and is accessed at speed via my app project code.

But as I say if the data is on one of your sheets it’s definitely viable to do the ARs with a button click in XL. In fact I’d be sure given the info ChatGPT could sort the code to do it in seconds. It’s just a case of having the structure and field names.

 

[JennyK]

I have no problem generating ARs from my databases, T Tufnel. The problem is that some handicappers have wins not in my databases, such as Meydan, Bahrain and sometimes even the US and France. Although I could obviously expand my database to take on board foreign runs, I've not thought it worthwhile as relatively few handicappers' ARs are affected. The only source I have found which appears to include all relevant overseas races is the Post, hence my enquiry about the feasibility of getting them from there by AI. In practice, though, it is perhaps two or three in a field of twenty in a race like tomorrow's Portland where it really matters to check, and I think I can do that manually as quick as AI could, especially if in the light of Leodis's warning about accuracy I'd need to check the results manually anyway.

 

[Tufnel]

I have no problem generating ARs from my databases, T Tufnel. The problem is that some handicappers have wins not in my databases, such as Meydan, Bahrain and sometimes even the US and France. Although I could obviously expand my database to take on board foreign runs, I've not thought it worthwhile as relatively few handicappers' ARs are affected. The only source I have found which appears to include all relevant overseas races is the Post, hence my enquiry about the feasibility of getting them from there by AI. In practice, though, it is perhaps two or three in a field of twenty in a race like tomorrow's Portland where it really matters to check, and I think I can do that manually as quick as AI could, especially if in the light of Leodis's warning about accuracy I'd need to check the results manually anyway.

Ok I understand. I just spent 15 mins or so finding a way to to get the ARs from the RP using just the url for a racecard. The reason most bots can't get the info is the RP's horse profiles are generated dynamically with API calls not embedded in the HTML. Once I pointed the AI to the data that deals with all that it has been able to retrieve the ARs for runners using the RP's racecard URL. I'll get it to give me the code to generate an exe file that will allow anyone to copy in a race URL and get the ARs instantly and they be generated in an XL sheet or CSV file.

 

[JennyK]

T Tufnel

I've just processed the Portland, and checked all the ARs generated from my database with the Post. Unsurprisingly for this class of race several had overseas wins and I've amended my grid accordingly.

What would be interesting, if you can generate the ARs for the race via AI, would be for us to compare. Obviously I could have made an arithmetical error when using the Post data but that is easily checked if there are any differences. Assuming mine are correct, such a comparison would at least provide one test of the AI's accuracy.

I am going to leave the detailed analysis of the race until tomorrow, but my initial reaction to the data in my grids is that nothing shouts bet. Only one consistent "form" horse in the field out of twenty-one runners, in my view; Twilight Calls. Coincidentally TC is 11th in the ability ratings ranking, so exactly in the middle. I don't think VDW ever went that low in his "certainties", "outstanding bets" and stated bets examples.