DSLR Profiles

[pawras]

I had some spare time so I had a wee ponder about DSLR profiles by trainer following the discussion in the thread.
Coder/scraper needed

So as an initial stab :-

I looked at all runners and since 2000 and 50% have 24 or less days between runs and 75% have a run within 71 days of their last run

Next in my own horse racing query system, going along the lines of what was talked about on the thread, I defined 10 categories and applied them to the last 5 runs of every horse to come up with a DSLR profile code
A = 1 to 7 days since last run
B = 8 to 14 days since last run
C = 15 to 21 days since last run
D = 22 to 28 days since last run
E = 29 to 35 days since last run
F = 36 to 42 days since last run
G = 43 to 49 days since last run
H = 50 to 56 days since last run
I = 57 to 84 days since last run
J = 85 or greater days since last run

Therefore for example if for the last 5 runs the days since their last run for each race was :-
Prevrun1 = 5 days
Prevrun2 = 30 days
Prevrun3 = 14 days
Prevrun4 = 42 days
Prevrun5 = 100 days

This would give a code of JFBEA , I decided to order it like you see the formline i.e. latest on the right.

However I considered some trainers might like to use 5 prep runs some 2 or some 1. Plus one trainer might use different prep runs for different types of horses.
So I broke the 5 letter code for each runner into 5 separate codes for easier processing/experimentation, therefore JFBEA would become
JFBEA
UFBEA
UUBEA
UUUEA
UUUUA

U signifying that run to be ignored.
Also you the encounter horses switching trainers so made the query only count runs at the current trainer.

Once I had that info I reported on descending profitability by race type and trainer and dslr profile code.

I’ve done so for just for a dozen or so trainers initially because working out all the stuff on the fly is a 'bit' resource intensive. Plus I limited to 1 to 5 prep runs for that reason

Thoughts on this top end of the list for Mark Johnston to £1 level stakes ISP since 2001?



Oh yeah for clarity if you look at the top two rows UUJDC is a subset of UUUDC

 

[pawras]

And here is the top end for Mark Johnston by race type, hcap/stks

 

[mick]

Good work pawras i have read the theory that these days training facilities have improved ,and as a result a horse running after a long break may not be the negative which many backers used to think.With this in mind have you looked at your G - J sections in this respect.Are they showing progressively better via recent years.?

 

[pawras]

Here's Mark Johnston just looking at G - J codes for the most recent runs

 

[pawras]

John Gosden has few runners out today at Ascot and Newmarket , therefore here's the top end of his list re total profit for his DSLR profiles by race type and hcapstks

 

[pawras]

A quick update, I set about generating the DSLR code for all trainers in all races since 2001, it was something that I left to run over night.
Anyway so I've started playing around generating the racetype , trainer, dslr code report for all trainers (3.3 million row report). Funnily enough even in the full list Johnston and Gosden feature in the top end.

 

[Tomasteros]

I have just registered as I wanted to look more closely at the DSLR stats. My own very simplistic system is to pick horses running after 21 days. I paper traded for a while across all DSLR and 21 seemed to work best, closely followed by 24 days, next best was 42 days funnily enough. I also only bet on the price range 2/1 to 12/1. Outside of this range the strike rates lose money. My main interest is to find out if there is any way of filtering horses by DSLR to save me trawling through RP every morning! I don’t keep stats so I can’t tell you an ROI but I only bet very small amounts to keep it fun as losing streaks often occur but it’s a long time since I’ve had to fund my B365 account. I also find if you apply too many filters to can go days without pacing a bet which defeats the object of having a bit of fun.