I'm not sure successes describes the bumpy road covered!
To get to this point has involved a number of steps, I'll mention a few odd thoughts then but there are undoubtedly many other paths that can be followed, some of which will be quicker and more successful.
There are some things that haven't changed over time, that have been and I think are essential for taking the path I have. One has been programming. I've written programs for about 40 years, I've learned a good number of computer languages over that time and become competent in them, and I think it would be fair to say I am well and truly bored and fed up with them.... but without being able to program much of my daily work I'd never have got through anything like as much of it as I have. Throw in forcing myself to learn how to set up macros in Excel and suchlike. Every day I run a series of programs that strip results data, rejiggle it, save it as a set of databases (very simple csv's, nothing clever), then go through that data extracting information such as how each runner performed (not just winners). The odd Excel macro is used to sort the course winrate info along the way.
Those programs not only extract the basic data about the runners, races, and trainers/jockeys etc but also give me files I can load into Excel and use the data filtering on to find out win rates for different combinations of info - so I can check, for example, how all my 2nd or 3rd or 4th top rated runners yesterday did, and I can do that over any set of distances, courses, etc whilst also seeing what 'flags' were in play, what the market position was, all sorts of things. These allow me to do things like the 'picks'. where I use Excel to determine that 'top rated in a handicap, was in last three last time out' and similar are worth a bet. When I find things like that I make additions to the ratings program i run in the evening that will pick these runners out and list them in a csv for me to look at.
So not a change - the bedrock of it all is being able to use Excel and write programs. I program in Python 2, which is a very easy language to learn - like all programming languages a small start writing things that do simple things earns the experience needed to write more complex and useful code.
Originally I looked at a couple of books about betting systems, Larkspur method was one, class figures another, and I've since read a few more from the likes of Nick Mordin. I consider them to be distractions, from a betting viewpoint, at the very most I'd pick up a fraction of an idea that would in itself not be any use but perhaps kicked my mind off into another area that I did make some progress in. Once I'd got past reading them for betting and just enjoying the read as a story I got onto speed ratings, which I'd done years ago for greyhound racing, and trawled Youtube looking for clues. I'm pretty certain I found a video from
TheBluesBrother that was, by then, maybe 8 years old or so. I picked up a basic method of compiling them from that, got some info on here from Mike about improving the old method, then got further info such as 'rail movements are important'.
After a few weeks of doing ratings in Excel, and storing them in the RP 'my ratings' bit (and cursing the inability to change the lbs/length values on a per meeting basis) I found I was putting in a couple of hours a day just rating horses manually, and that's when I decided I'd never want to work that hard every day just to rate results. You can't 'skip' days though, so if you don't do the ratings one day you have to go back and add them to the next day's workload. This taught me a bit more bloody mindedness perhaps, for the past 50 years or so I have tended to batter away at things until they are finished to what I consider an appropriate standard - I think that being stubborn is important too. That's what drove me to program it all up, to make as much of the job as easy as I could. It still takes an hour to two hours to process results, but I am also stripping all sorts of analysis out as well, all of which forms the foundation of my efforts.
From then on changes did kick in, because I found the raw ratings probably only give around 20% winners - this varies, as the ratings are always based on past runs of course so when the going changes, for one example, ratings can be quite poor until newer ratings start to appear. The variation in going coupled with the variations in track geometry combine to make consistency in assessing races a difficult business.
First big change was realising that the top rated runners represented a pool of potential winners, and would be the best place to look for horses to bet on. This is one of those simple ideas, 'don't bet on a horse doing something it hasn't managed to do before' as I'd phrase it. To win a race going faster is, I would argue, a significant bonus. A high past speed rating shows what a horse can do in the right circumstances. I see no reason to ignore the top rated runners... all I need to do is find the ones from that shortlist who are going to win! (Easy eh?)
This change led me to start looking into the data to try to identify patterns amongst the top rated - I did checks on how many ratings to look back over, and suchlike, and I spent a lot of time looking at the data with ideas like 'was placed last time out/ did/did not win last time out was in the first x in the betting, had previously won/placed at this class/going - all sorts. What I found was a bit annoying, very very few such ideas, when I checked the results, actually produced notably better win rates.
Perhaps 18 months back there was some emphasis on runners that were setting top ratings in the past week, and also fastest last time out. By focussing on these I began to see patterns I could use - the 'selections' were the result, and for about 3 months each winter, as the soft ground through December-Feb or so stayed relatively unchanged, the NH selections did okay. I never got the handicaps to work though, and a few good runs was all I ever got from them - the usual problem, find an idea that works really well, and as soon as you use it the idea starts to splutter.
I decided that I needed better data - back as this was all starting 4 years ago Mike had asked me to write a program to check race times from Timeform and RP against each other, I had to learn how to scrape websites via my programs (that had me stymied for several days) - and I check these against my HRB data each day. This highlights races I need to check videos for timing, I kind of assume that if all TF and RP (HRB is basically the same as RP) times agree then they're probably correct. Rail moves from the BHA sleepy weasel department, that only leaves standard times.
RP standards this time last year were bad, in my view. I spent months writing more programs, after about 2 months just collecting and editing the data (correcting as many errors as I could - plus trying to ensure I got things like 'outer/inner course' and suchlike correct). Then I compiled my own set of standards and spent another couple of months trying to find the errors in them, and apply corrections. Now I have standards I can work with, and that has boosted my results a little. I find that the 'picks' are working okay - a combination of top rated runners with extra filtering provided by their previous placing - this will always be a bit less than precise, coming third in a 20 runner handicap is probably rather a good pointer while third of three by 25 lengths is not - but such issues should be statistically trivial over a season.
Standards -
O
Outlander did a few courses, initially I added his standards in place of the relevant RP ones, but I decided this was too much of a bodge, and although I was very very loathe to do the work, I knew it would be months of slog, in the end I decided I had to do it.
So although not a change, interacting with people who have a clue is, and has been, important to the development.
I'd also say that interacting on here with other members is important, it helps avoid that 'working in isolation' feeling that can otherwise arise - I doubt
mick and myself have exchanged much info that we subsequently made any use of, but the entertainment value of 'chatting' with him has been significant, I'm not going to do an acceptance speech but others like yourself
markfinn ,
Chesham ,
pete and a good few more (enter blog member list here) have provided company over that journey.
Good records are essential, at least to me - with a record of everything I rated over the years I can test out 'what if' ideas, which I could not do otherwise. I can check back to see if what I did was any good - don't rely on how you think things are going, our minds play tricks, there is nothing like a simple spreadsheet to show you if an idea was profitable or not.
Learn your limits - know how much work you can sustain, there is no point working 6 hours a day at something for a month then giving it up as you find you cannot sustain that level of effort.
Not so much changes, there is a bit about evolution there though - program, learn to check how things went, check the data, improve the data, practise analysing the data until you can look, effectively, for patterns to inform future efforts.
Dave