LinusP wrote: ↑Tue Dec 05, 2023 8:23 pm
When you model something you are looking to quantify what is normal under specific conditions. In my opinion the variables you have mentioned are a bit too abstract / not really going to tell you anything, at the end of the day we are after value and we want lots of it.
If we keep things simple you could look at how much a price should move from its starting price as a function of time remaining (or metres if you have TPD) By modelling this you would then be able to put a number on the min/max price a runner should be trading at a specific distance / time etc. The hard part comes down to having this data available and in a format that makes it possible.
To put some context on the idea of value inplay, if you simply lay the lowest priced runner every second you won’t actually lose that much money (assuming you are getting the mid price / reasonable bet sizes) You then need to swing things into value, for example by using our inplay/sp model described above to remove the negative EV bets.
We are currently ignoring how this EV relates to matching and things start to get very tricky and counter intuitive but I have probably said enough.
Had to look up what ‘dobbing’ is, just seems like another wishy washy idea that ignores how you actually make money in this game, value. You describe it as simple but to be profitable you would be placing a value bet before the race and ideally during which is level 11.
So many questions!
I was up till 2am this morning trying to disseminate (if thats the right word) your reply with ChatGPT, and gave a detailed reply in my head, but of course cannot remember that now lol. I'll try and give this a go...
Seaching online last night for anything relevant to dobbing and creating statistical models I came across this YouTube video fom Racing Bet Data
https://youtu.be/xwzaOPLrY3I?si=NoVCLaAKFs8QDYi9 which basically gives me the result of my 'abstract' variables in my initial post. Great, job done, we can all go home!
However, this doesn't help me to understand how and why to model data. My reasons for this post were actually two fold. One being due to the poor quality of racing recently and wanting to expand my 'toolbox', but also because I have been looking at Associate Trader jobs with gaming companies and researching each job role requirement individually, one of which is to create models, and I wanted to at least have some basic idea of how its done, so if I was to apply for one of these roles they are not met with a blank stare as my brain crashes lol. So what better way to learn than to build one myself.
Unfortunately ChatGPT gave me equally complicated answers which didn't help, but like a moth to a flame, now I have a desire to NEED to know. Looking at sports models for beginners I came across another YouTube video where he was using NFL data to work out the probabilities of the outsider/underdog teams beating the favs and winning the match. Thinking along those same lines, I was going to say something similar i.e. probabilites of 'selected horse' shortening in running under specific race conditions, but as I re-read your first line re quantifying normality under specific conditions I can actually see how abstract my suggestion was.
I am going to have to think about this because in order for a selctions price to shorten in running people need to back it, whether thats because its value, which itself requires many variables to identify value, or because the race is unfolding in a favourable way for the 'selected horse', something which I dont think can be quantified prior to it happening live in a race, there's literally millions of possibilities of how a race can unfold.
In your TPD example, is it even possible to say a selection trades between IPMAX/IPMIN e.g. 70% of the time, to give you a window of opportunity, without knowing what race conditions make up that 70%? Do in play races really repeat themselves enough to provide repeatable price patterns when one fav can be a completely different beast to the next. Not to mentions riders!
At present, even though I do add exceptional horses to my tracker, I admittedly do not look at how those horses price move during a race. Im either keeping them onside as potential B2L's or looking for signs of struggling for L2B's. To me every race is unique at the start, because even if a horse has been a successful back or lay last time, I've found quite often it can be the complete opposite next time out, as the trainers try and position them for a win.
I will try and simplify what I am asking so that it is easier to quantify and come back...
Just to clarify and see if I have understood your example correctly
If we keep things simple you could look at how much a price should move from its starting price as a function of time remaining (or metres if you have TPD) By modelling this you would then be able to put a number on the min/max price a runner should be trading at a specific distance / time etc.
.
Is this you "quantifing something normal under specific conditions", as a base to work from there?
TIA
