• 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

Bad News For Bower ?

OK here are the figures for 2011 to 2025

using OR we have 43,937 bets in the short list backing all of them this produced to BFSP a ROI of -1.37% and for top ranked +1.26%

Now using weight we have -2.27% on all short list bet and +3.88% on top ranked

I balanced the mid range criteria to make sure there was pretty much an equal number of bets in the weight option

I would be interested in your thoughts on this. broadly speaking OR performs better on the short list but worse on the top ranked
 
MySportsAI said:
OK here are the figures for 2011 to 2025

using OR we have 43,937 bets in the short list backing all of them this produced to BFSP a ROI of -1.37% and for top ranked +1.26%

Now using weight we have -2.27% on all short list bet and +3.88% on top ranked

I balanced the mid range criteria to make sure there was pretty much an equal number of bets in the weight option

I would be interested in your thoughts on this. broadly speaking OR performs better on the short list but worse on the top ranked
Click to expand...

Nice work :hi:

A good data size to work with. Based on what you have, I'd lean more towards the top ranked area for comparisons. FWIW, my view is not to back all short-listed runners and refine my decisions towards the top rated end anyway. It's interesting that the weight (lbs) option, profits more on the top ranked.
 
Can you break it down more, to check position finish the last race or may be last 3 or is that too many equations just a thought
 
If it does better betting top ranked then possibly it would be even better cutting those races down to top rank races also and even possibly top rank trainers and jockeys.
So once you have your selection the race before hand had to be say minimum 8k races and then looking at trainer and jockey recent form.
 
Michael231 said:
Can you break it down more, to check position finish the last race or may be last 3 or is that too many equations just a thought
Click to expand...
OK before I answer that what do you think will happen to the top rated ROI% if I focus say only on those beaten less than 3lbs LTO so that is winners and close finishers
 
gerry said:
If it does better betting top ranked then possibly it would be even better cutting those races down to top rank races also and even possibly top rank trainers and jockeys.
So once you have your selection the race before hand had to be say minimum 8k races and then looking at trainer and jockey recent form.
Click to expand...
There is no restriction on race value in my figures even though Bower suggests higher value races, do you think he is right ?
 
MySportsAI said:
OK before I answer that what do you think will happen to the top rated ROI% if I focus say only on those beaten less than 3lbs LTO so that is winners and close finishers
Click to expand...
With regards the distance beaten if you go to 5 lengths then you would also look at the class up drop or down then going but I was thinking of the IV for the figures
 
This is the sort of thing my computer program goes through to evaluate the field as well as Bowers.

Strongly backed 6.5→5/2 ⭐
📊 IV 2.41 (Strong) — medium fields ✓
🌦 Good going match (Good to firm)
📋 Won last run
🎯 Class rise (Cls 4→2) ⚠️
📅 20 days — good preparation
🏆 Course winner
Weight down 7lbs ✓
👤 Trainer 10% at Goodwood (2/20)
+4.9
 
OK if you only consider > 8k races you get -2.93% and +7.7% so a small reduction in all pruned list but an increase in ROI for top ranked. Personally I would trust the larger sample size of all the list after all Bower suggests this is where you use conventional form analysis and IMO you need to reduce this loss in this are as much as possible in order to lower the incline you are trying to climb up
 
I only checked the top ranked horses for winning and getting beaten less than 3lbs and not surprisingly to me it reduced the ROI. If you allow yourself to be channelled into the same thinking as the crowd you will move towards mirroring their performance
 
MySportsAI said:
OK before I answer that what do you think will happen to the top rated ROI% if I focus say only on those beaten less than 3lbs LTO so that is winners and close finishers
Click to expand...

Probably will increase. due to less qualifying selections, but could give a false positive. *Holy horseshoes, just read your last comment*
 
MySportsAI said:
The improvement in better class races may be due to something Nick Mordin touched upon namely that class is like a pyramid and a somewhat increasing gradient as you go up the pyramid. In other words it gets harder to break into the class above whereas a class 5 race can have all number of horse potentially breaking into it
Click to expand...
MySportsAI MySportsAI , I have a question that somewhat touches on your points here .

Over the last few years or so i have tried to incorporate times/sectional times into my logic of race reading and had expected some common sense within the numbers but that's not always the case.
For instance yesterday i was looking at two races over 5f, one was a cl2 and the other a cl5, roughly speaking there was about 30lbs between the two standards and while the class 2 race was a little faster, not enough to my mind and so the reason for my question.



Did LITTLE MI MI simply run faster than her 65 rating suggested she could achieve ?
 

Front-End Pace Map​

Mapping the front end of both fields reveals that neither race featured a Lone Speed Anomaly or a leader getting a free pass on the lead. Both races were characterized by contested early fractions that resulted in late pace collapses, perfectly setting up efficient stalkers/closers to take the wins.

  • Race 1 (Al Shabab Storm): Canon's House pushed a very aggressive early pace (1f in 13.08s), contested closely by Vantheman and Paddy's Day. This early exertion led to a severe late deceleration for the front-runners, evidenced by Canon's House returning a Finishing Speed % (FSP) of 97.48%.
  • Race 2 (Little Mi Mi): A similar dynamic unfolded. Miss Rainbow contested the early lead (1f in 13.62s) alongside Tanjen and Azuinthejungle. The front end completely collapsed late, with Miss Rainbow returning a drastically inefficient FSP of 96.28%.

Class vs. Final Time Comparison​

There is a massive class gulf between these two fields, yet the final times are separated by only a fraction of a second.

Race WinnerOfficial Rating (OR)Implied Race ClassFinal TimeFSPVs. Par
Al Shabab Storm98Class 257.86s99.63%-1.11s
Little Mi Mi65Class 658.08s100.61%-1.65s

The final time difference between the Class 2 winner and the Class 6 winner is just 0.22 seconds (approximately 1.3 lengths).

Key Reasons for the Time Proximity​

The convergence in final times despite the wide gap in class is driven by three primary overlapping factors in the data:

1. The Weight Discrepancy (The Great Equalizer) Higher-class horses generally run faster, but weight anchors speed. Al Shabab Storm (Class 2) carried a massive top weight of 9-12 (138 lbs). In contrast, Little Mi Mi (Class 6) carried just 8-13 (125 lbs). A 13-pound weight pull is highly significant in a 5-furlong sprint and acts as a massive equalizer in bridging the raw time gap between a Class 2 and Class 6 animal.

2. Optimal Energy Distribution While Al Shabab Storm ran an excellent race, Little Mi Mi ran a nearly flawless tactical race. Little Mi Mi's FSP of 100.61% indicates a perfectly even distribution of energy from wire to wire, capitalizing exactly as the leaders (like Miss Rainbow) fell apart. Al Shabab Storm's FSP of 99.63% was slightly less efficient. The Class 6 winner squeezed every ounce of potential speed out of her physical ability through perfect pacing.

3. Performance Relative to Par Race 2 was simply a statistical outlier for its grade. When measuring against the time pars for their respective classes, Little Mi Mi overperformed heavily.

  • Race 2 generated a Time Index of 9.5/10 and ran -1.65s faster than par.
  • Race 1 generated a Time Index of 8.9/10 and ran -1.11s faster than par.
The Class 6 field ran a remarkably fast race for their level, while the Class 2 field ran a solid, but slightly less exceptional time for theirs. The combination of the 13lb weight swing and Little Mi Mi's peak tactical efficiency perfectly explains why the final times are nearly identical.
 
I copied this into GPT within a project area I'm already working on. It's reply took into account my previous conversations, but the outcome makes sense in relation to T tacker original question.

1785659153783.png

GPT Response.

This is actually an excellent example of why Official Ratings (ORs) measure ability, not necessarily the performance on one particular day.
The difference in winning time is 58.08 − 57.86 = 0.22 seconds (At 5 furlongs, that's roughly 1.2 lengths)

First observation​

That immediately tells me one (or more) of the following likely occurred in separate races.

Little Mi Mi produced a career-best. (Yes)
Al Shabab Storm didn't run to his OR. (Unsure)
Track conditions altered slightly. (Unlikely)
The races were run at different early tempos. (Yes)
The Class 5 race was genuinely strongly run. (Yes)

The sectional data suggests, the last explanation deserves serious consideration.

Also, Little Mi Mi finished the race more strongly.
--
Did Little Mi Mi outperform her OR?
On this day:

carried only 8 lb less
finished just 0.22 sec slower
finished stronger
won comfortably

That suggests she almost certainly ran above 65. Perhaps somewhere around 72–78, depending on the strength of the opposition she beat.
---

My Response.

FWIW, my focus would be on how A.I. tries to project an OR for future races, based on the data provided. So what happened next?

The OR awarded for the next race was 70 (+5lbs) for it's next race @Hamilton. If A.I was to be believed (and why not?), it still has 2lbs - 8lbs) in hand if running in similar class over a similar distance. (If you trust Little Mi Mi can transfer AWR form to turf)

8.00 Hamilton: Class 5, 5 furlongs Good Ground
Prominent, switched right and pressed leaders over 1f out, led inside final furlong, kept on well inside final 110yds, readily

Market: 4/1f (previous run 9/1)

1785660028006.png


1785660133665.png
---
 
DuckandDive DuckandDive
Sometimes AI can confuse itself with this type of question a needs to be taken with a pinch of salt but reading your interaction with chat gpt it appears your earlier previous conversations have been a good help.
I should say that this certainly isn't the first time that i've witness such apparent anomaly and in my opinion there is something fundamental happening when we see the numbers which on the face of it distort the official ratings.
There's a delicate balance to be had when trying to analyse sectionals and if you go back and compare all the horses in both of those races it becomes even more complicated and that's before you need to take into account of beaten horses being eased down.
I have said before final time percentages are worthless without understanding the pace of the race, mainly the early stages but also other parts of the race, distances are obviously a factor to throw into the mix.

This conversation might have worked better yesterday morning ?
 
Hi The pace of the race will effect the overall time even in a 5f race I would wait until the sectionals are available before deciding whether Mi Mi ran better than class. With regard to overall times and RP speed figures the latter do show a linear increase in value of average SF as you rise in class so yes class 2 horses tend to run higher SFs than class 6. If I was you I would inspect the sectionals compare them to pars if you have them and do the same for final SFs. Mi Mi has 2 entries on August 7th
tacker said:
MySportsAI MySportsAI , I have a question that somewhat touches on your points here .

Over the last few years or so i have tried to incorporate times/sectional times into my logic of race reading and had expected some common sense within the numbers but that's not always the case.
For instance yesterday i was looking at two races over 5f, one was a cl2 and the other a cl5, roughly speaking there was about 30lbs between the two standards and while the class 2 race was a little faster, not enough to my mind and so the reason for my question.



Did LITTLE MI MI simply run faster than her 65 rating suggested she could achieve ?
Click to expand...
 
You must log in or register to reply here.
Back
Top