• Hi Guest Just in case you were not aware I wanted to highlight that you can now get a free 7 day trial of Horseracebase here.
    We have a lot of members who are existing users of Horseracebase so help is always available if needed, as well as dedicated section of the fourm here.
    Best Wishes
    AR

AD's daily selections

We are delayed today and I know we've had some races but I am still putting them up regardless, I'm not doing 600 odd bets to pull a stunt with anything that's already run. If I need to put an asterisk next to the results, I will do.
 
So here we are a little late for today. Because of this, it is reviewing the correlation with the market based on one or two races that have already happened but for the record its 83%. York attached too.
 

Attachments

  • Live_Analysis_2026-05-13.xlsx
    32 KB · Views: 6
  • Live_Racecard_Landscape_York_2026-05-13.pdf
    1.6 MB · Views: 0
1778680385172.png
These are today's value picks. I've gone back over them and in the daily live analysis they are highlighted in yellow but on occasion are top rated altogether - I've only accepted those where each way / place terms are applicable.

The thread has raised 132 of these in total (average BSP 22.56) and delivered 15 winners for 47.58 BSP profit, and 2.75 BSP profit for places. It is volatile and we've only won with 3 of the previous 60 selections so it will give lots of long price losing runs but that's the way of it, the primary model is going to grind out the winners and this is take or leave stuff.
 
Again here we are with lots of winners but it's getting on my wick that we are not really moving the profit along much, will keep being patient but today 40 winners, 13 winners and 13 second places gave a BSP loss of 2.51 pts. The last 4 races were winners to reduce the daily damage.

TOP RATEDS AFTER DAY 16
680 RUNS
220 WINS
32.5% STRIKE RATE
60.74 PTS TO BSP
8.9% RETURN ON INVESTMENT

I am scrapping the value bets at least in terms of reporting, as it's too much time for now to keep an eye on. Looking at the last 7 days, where the BSP is 5 or less the A/E is 1.12, for those where BSP is over 5, the A/E drops to 0.90. We will keep ticking to day 31 and it's either proper money or a re-think of approach.

Also noting that if we'd had £2 per selection our initial £200 bank would be £321.48 and a daily 1% of bank would be £305.31 meaning stakes tomorrow of £3.05. Not really much between those totals.
 
As an aside, it's been apparent over this test period that the model is actually much more effective with flat races than it is with the national hunt. The racing on the level is running around 34.4% and 75pts or so with the National Hunt around 27% and a loss. I am going to continue for now but see if there are any missing links to the NH racing that we can try and address.

Anything we do change is tested privately along the public selections before we put forward for replacement. However the selection model remains unchanged so far since day 1.
 
Today will be a big hit I fear, but I may have stumbled across an issue with the National Hunt selections and even yesterday a re running of the selections upped the winners from 13 to 16 so I'll re run the models from day 1 and see if its done better.

If it has a new model will be worked off and its only fair for transparency to close the current one.

I'll probably keep this going until the weekend and run the new one as of Monday next week with a new scoresheet.

I will also monitor by race type and see if larger rating margins impact on either strike rate or on profit. I still want an all race output. The performance on flat handicaps up to now is nevertheless quite mindblowing.
 
7 winners from 35 and a trashing of 16pts lost. I have got a bit of a cob on.

TOP RATEDS AFTER DAY 17
715 RUNS
227 WINS
31.7% STRIKE RATE
47.29 PTS TO BSP
6.6% RETURN ON INVESTMENT
 
Today's selections, and the card for my home track. I am trying two other models at present - we'll see how they perform but they've not been put to any use for these choices. 81.5% market correlation today.
 

Attachments

  • Live_Analysis_2026-05-15.xlsx
    39.9 KB · Views: 6
  • Live_Racecard_Landscape_Aintree_2026-05-15.pdf
    1.4 MB · Views: 1
17 winners from 50 today and a successful day with 11.24 pts. It threatened to be a bit better at one point but it is what it is.

TOP RATEDS AFTER DAY 17
765 RUNS
244 WINS
31.9% STRIKE RATE
58.53 PTS TO BSP
7.7% RETURN ON INVESTMENT

It's notable that the handicaps, novice and maiden races are absolutely hammering it, but the jumps is generally tanking.

1778880637922.png

We will carry on to see if it turns but I am giving some thought to whether the model should stop looking at National Hunt and we construct something a little different. I think the feature engineering documents could require some review. Altogether though if you'd said we'd get through 750+ bets and be in strong overall profit, you'd be taking it.
 
AustinDillon75 said:
17 winners from 50 today and a successful day with 11.24 pts. It threatened to be a bit better at one point but it is what it is.

TOP RATEDS AFTER DAY 17
765 RUNS
244 WINS
31.9% STRIKE RATE
58.53 PTS TO BSP
7.7% RETURN ON INVESTMENT

It's notable that the handicaps, novice and maiden races are absolutely hammering it, but the jumps is generally tanking.

View attachment 167226

We will carry on to see if it turns but I am giving some thought to whether the model should stop looking at National Hunt and we construct something a little different. I think the feature engineering documents could require some review. Altogether though if you'd said we'd get through 750+ bets and be in strong overall profit, you'd be taking it.
Click to expand...
I'd be getting rid of the NH stuff for sure. (y)
 
Sullybomb said:
I'd be getting rid of the NH stuff for sure. (y)
Click to expand...
Its getting that way. However Punchestown aside its summer jumping, and the model has tested fine over past winters. I'm going to keep it for complete transparency but perhaps post split results and if at months end the trajectory remains I'll bin it.
 
Today's racing, with Newbury and Newmarket. 78.7% market correlation with cards downloaded in the past half hour.
 

Attachments

  • Live_Racecard_Landscape_Newmarket_(Rowley)_2026-05-16.pdf
    1.2 MB · Views: 2
  • Live_Racecard_Landscape_Newbury_2026-05-16.pdf
    1.6 MB · Views: 3
  • Live_Analysis_2026-05-16.xlsx
    44.1 KB · Views: 6
Yesterday was day 19, not day 18. In terms of it being a Saturday, 59 races and 17 winners, so 28.8% strike rate (it was feeling more successful as the day went on but obviously not) and a total profit of 1.98pts. A profit is a profit.

TOP RATEDS AFTER DAY 19
824 RUNS
261 WINS
31.7% STRIKE RATE
60.51 PTS TO BSP
7.3% RETURN ON INVESTMENT

And guess what, when I've amended my results file I've found that it hasn't actually attached any results to my master record since May 3, so it's been guessing for effectively a fortnight. So hopefully we'll see this pick up a little now that's corrected.
 
I didn't put up today's selections but we got 11 winners from 30 and it generated a loss of -2.61 pts. The model is still on 57.9 pts, but I have been working on the fix above and it is now profitable across all three disciplines, and

BETS WINS 2ND 3RD STRIKE RATE PROFIT AND LOSS RETURN ON INV.
85427215711231.9%57.96.8%

So I re-ran a new script that looked at the saved cards and ratings and it delivered the following:

BETS WINS 2ND 3RD STRIKE RATE PROFIT AND LOSS RETURN ON INV.
85027415511232.2%112.5513.2%

And the breakdown:
1779060055044.png

1779060084734.png

So the handicap hurdles are still tanking but I am sure the strike rate is due to improve on those. I am seperately analysing the bets now too.
 
Here are today's selections. We are going to tally these up from today, we've made some adjustments to reflect the chase and hurdle models and we'll keep the total selection total running as well so we'll have two running totals, one applying since we started on April 27, the other total fresh with the new model.
 

Attachments

  • Live_Analysis_2026-05-18.xlsx
    34.4 KB · Views: 8
I am going to keep these going as an exercise but now take a break from issuing them, yesterday was a bit of a nightmare which does happen but it's feeling like it's missing something with the choices, we lost over 18pts yesterday and was 9 of 42 (expected 10.84 winners). I am sure this is an expected slide but the testing mixed very good says with small losses, now in real time we see small gains and some catastrophic losses. I will need to re-run. The model is either not being responsive enough or conversely is adjusting more than it really needs to.
 
AustinDillon75 said:
I am going to keep these going as an exercise but now take a break from issuing them, yesterday was a bit of a nightmare which does happen but it's feeling like it's missing something with the choices, we lost over 18pts yesterday and was 9 of 42 (expected 10.84 winners). I am sure this is an expected slide but the testing mixed very good says with small losses, now in real time we see small gains and some catastrophic losses. I will need to re-run. The model is either not being responsive enough or conversely is adjusting more than it really needs to.
Click to expand...
Its just variance I'd imagine, with the amounts of bets your model generates daily I'd be expecting fairly substantial swings.
 
Sullybomb said:
Its just variance I'd imagine, with the amounts of bets your model generates daily I'd be expecting fairly substantial swings.
Click to expand...
I'm sure you are right and I am just being a bit spooked. As you said earlier in the thread these runs are to be expected but the work isnt stopping. The model isnt changing fundamentally but may be suffering a bit too much noise and distraction from the sheer number of features so I am going to try a bit of pruning to see if I can get it a little more focused.
 
AustinDillon75 said:
Some of you may know that I've been working away with speed ratings, ELO and other related things, and in time I will maybe take these to another more detailed thread when there is more to say but for now I'd just like to put up horses suggested by my own personal system.

I've been able to run backtesting with quite stunning results on these using a range of statistical techniques. Over a long period of time, they tend to rustle up 6 or 7 selections a day on average, but for now I'll just put in the two selections it's found today.

Yesterday I tried it and it was 2/2 with Kazar Forez (5/6) in the 1440 at Sedgefield and Argento Boy (5/2) in the 1525 at Naas. It also identified Largy Train (11/2) at Naas and it got the cross country winner at Cheltenham over the weekend as well (noting that whilst top rated, it did not suggest a bet based on available confidence).

So here you go for now, we'll keep track to odds on the HRB racecards at the point of posting, and record to both those and to SP.

1400 Hereford - Draco Malfoy (1/4) !!!!!
1730 Wolverhampton - Mighty Vega (5/1).

Imagine if my first tip on this results in a 1/4 shot being turned over in a 3 runner race. It does give a top rated for each race but I wont blind the thread with those for now.
Click to expand...
Stick with it AD. You are approaching the experiment in the right way IMO. Sometimes, things will be frustratingly unexplainable.
 
You must log in or register to reply here.
Back
Top