SPEED RATINGS

[harry_rag]

4 minutes ago, Torque said:

Not aimed at you in particular @harry_rag rather anyone that has the numbers. Using the average odds you mention and a 5% yield, you'd need a sample of about 4,500 bets to suggest strongly that luck wasn't at play. That is to say that if you had a yield of 5% after about 4,500 bets at average odds of 3, there'd be a less than 1 in a 100 chance that those results were down to luck.

Cheers, I’ve probably got more than that overall but have split the data off since SPIN effectively ceased to exist. Luckily the returns seem to be holding up despite me only having one set of prices as input rather than two.

To continue the movie references, in terms of speed figures given the bigger average odds you’re going to need a bigger sample size!

[Zilzalian]

4 hours ago, Zilzalian said:

Best bet today both from a speed perspective and value is DRAMA in the 730 Kempton 9/2 is 3points over what it should be imho.

Backed into fav but ran poorly or wasn't off and a lot slower than (5 secs over 6f) its best figure on the Kempton aw. 

[Villa Chris]

What would you do if your odds for a horse are identical to the bookies . What that class as a bet if you’re looking for value ?

[Torque]

11 minutes ago, Villa Chris said:

What would you do if your odds for a horse are identical to the bookies . What that class as a bet if you’re looking for value ?

Move on to something else.

[harry_rag]

8 minutes ago, Villa Chris said:

What would you do if your odds for a horse are identical to the bookies . What that class as a bet if you’re looking for value ?

I’d say no. If you think the price is fair then you don’t bet. If you’re serious about calculating your own prices then you have to have enough faith in them to only bet when the price offers you an edge. If you find that you’d make a profit backing horses at your fair odds then you need to go back to the drawing board as you’ve not quite got it right yet.

[MCLARKE]

On 7/13/2024 at 4:44 PM, Torque said:

If you can post up number of selections so far, average odds and harmonic odds then I can check from a statistical perspective how much of what's happened so far might be attributed to luck.

Finally got round to this, the number of selections is 931, profit is 131 and average odds are 11.14.

You'll have to excuse my ignorance, but what are harmonic odds ?

[harry_rag]

12 minutes ago, MCLARKE said:

You'll have to excuse my ignorance, but what are harmonic odds ?

The ones you get at the 5th and 7th fret (as opposed to the 12th)? (One for all the guitar players there.)

[Torque]

2 minutes ago, MCLARKE said:

Finally got round to this, the number of selections is 931, profit is 131 and average odds are 11.14.

You'll have to excuse my ignorance, but what are harmonic odds ?

In a nutshell harmonic odds are a way to smooth out the effects of outliers in a sample. So for example, you've got a cluster of odds that are all very close to let's say 5.00 and then one or two odds in the sample that are let's say 20. Those fewer larger odds can disproportionately skew the average odds of the sample, effectively giving a higher weight to the outlier odds which are very much in the minority in the sample and giving you distorted average odds.

Harmonic odds are less useful if all of your odds are in the same range, say a touch of odds on to a touch of odds against for example, but that's unlikely to be the case in a sample of horse racing odds unless you have a system focussing just on favourites. In practical terms, where harmonic odds are useful is they can allow you to validate a system much earlier where you have a wide range of odds, although if average odds are used instead with a wide range of odds that means that once you've validated your system you've really validated it because if you used harmonic odds instead you'd have validated it earlier, if that makes sense.

If you can put up the yield from the 931 bets I can have a look at what chance the results so far could be the result of luck and off the top of my head I'd say there's a very good chance that so far you are riding a wave of positive variance - that's not to say that the positive results won't continue and they certainly will if you're identifying odds that are bigger than they should be.

[MCLARKE]

Is the yield not 14% (131 / 931) ?

[Torque]

I didn't realise you were using level stakes @MCLARKE I make it a 1 in 9 chance results so far are due to luck. Harmonic odds might reduce the luck factor if you've got a wide range of odds.

[Torque]

If the yield and odds are maintained, then luck will be more or less out of the picture once you've got about 3200 bets in your sample.

[MCLARKE]

So a 8 in 9 chance that the system is robust ?

This appears quite good

[Torque]

1 minute ago, MCLARKE said:

So a 8 in 9 chance that the system is robust ?

This appears quite good

Yes. It's a good start no doubt, but to put into betting odds it's like a ten to one on shot and they get beaten all the time. Also, and as you mentioned already, the yield is unlikely to be maintained and that will bring the 8 in 9 chance figure down.

[Torque]

1 minute ago, Torque said:

Yes. It's a good start no doubt, but to put into betting odds it's like a ten to one on shot and they get beaten all the time. Also, and as you mentioned already, the yield is unlikely to be maintained and that will bring the 8 in 9 chance figure down.

Having said that though, if you're finding on average a 14% edge then the yield will obviously be maintained.

[MCLARKE]

On 7/11/2024 at 12:52 PM, harry_rag said:

Might as well share my initial attempt at coming up with a value indicator for the races we've done collectively so far. If it's not too much effort to apply this to your data sample it would pretty much determine whether it adds any value.

1 - convert each runners' rating into an index number where the race average is 100 so if the top rated is 110 it means it's 10% better than the average for the race, 120 would be 20% better. (I do this anyway because I'm combining various ratings but it could still be useful for you to convert your ratings and see if backing all selections (say) >110 gives less selections and/or better results than your current approach.)

2 - For all horses that come out >100 subtract 100 from the adjusted rating (so 120 becomes 20 etc.) You can do this as a step or just include it in the formula to arrive at the next number.

3 - Calculate the value figure by multiplying the above by the horse's price, so a 120 horse at 5/1 = 120 (6 x 20) but one at 10/1 = 220. The horse with the highest number is potentially the best value selection in the race.

4 - (Optional) - I converted this to an index so that the highest value figure = 100 so in the example above the 220 would be 100 and the other horse would be 54.55. Might be useful for comparison across multiple races.

I think step one on it's own might be worth you considering, instead of just looking at a raw speed figure, index it based on the race average. If there are several qualifiers in a race you could discard those that are only just over 100 and favour those that have a sufficient margin above the average.

#	Horse	Odds	AvAv	V#
1	AMBIENTE FRIENDLY	6.5	121.6	71.2
2	ANCIENT WISDOM	6.5	113.8	45.4
3	CITY OF TROY	4	110.6	21.5
4	DANCING GEMINI	10	108.7	43.8
5	SAYEDATY SEDATY	41	104.8	100.0
6	MACDUFF	13	102.1	13.5
7	MR HAMPSTEAD	151	100.9	69.3

That was the take on the derby. 3 horses at 110+ and 4 below that but better than average. Sayedaty the top rated by the value indicator, AF the best if you set a minimum rating of 105 or 110.

I hope that makes sense! 

Results from option 1

Would have been nice to have been more linear but still quite promising

 

[MCLARKE]

These are the results if I include horses that have achieved a speed figure of 67

[harry_rag]

9 hours ago, MCLARKE said:

Results from option 1

Would have been nice to have been more linear but still quite promising

 

The headline take is really encouraging; all selections with a no better than average rating show a loss and those that are better than average are profitable. I think those bands are too wide, in particular you need a high band with greater volume and to break down the 100-109 band. I’d suggest maybe intervals of 4 so 100-104 then up to 108, 112 and >112. You could tweak the interval if you wanted to find the one that comes closest to having equal bet volume in each of the 4 bands.

A couple of questions; what are PS and PP and what time period does this data cover?

[harry_rag]

9 hours ago, MCLARKE said:

These are the results if I include horses that have achieved a speed figure of 67

So if I’ve understood this right the first table is all horses and this one filters out those that haven’t achieved a rating of 67+, meaning that almost 69% of all horses have achieved such a rating? What strikes me most here is the really low volume in the two higher bands. Only 27% of those horses with an indexed rating of 110+ have a speed figure of 67+ compared to 80% for the two lower bands. Somehow that seems counterintuitive. The majority of the best horses based on index score have a rating below 67. 

[MCLARKE]

Current profit 114.33 points

COURSE TIME  SELECTION ODDS  BOOKMAKER PLACES  Thirsk 1.52 Art Design 4.00 BETFAIR   Bath 2.30 Fregola 7.50 BETFAIR   Bath 3.30 Second Collection 9.00 BETFAIR   Thirsk 3.52 Vince Le Prince 6.00 LADBROKES   Bath 4.00 Malham Tarn Cove 4.00 BETFAIR   Thirsk 4.22 Lord Abama 8.00 BETFAIR   Bath 4.30 Beggarman 15.00 LADBROKES 4 Thirsk 4.55 Equity's Darling 13.20 BETMGM   Bath 5.00 Rita Rana 6.50 BET VICTOR   Yarmouth 5.07 Chola Empire 8.50 BETFAIR   Bath 5.30 My Ambition 2.50 BETFAIR   Yarmouth 8.10 Fletcher's Flight 4.00 BETFAIR   Ffos Las 8.30 Zambezi Magic 8.50 BETFAIR

[MCLARKE]

3 hours ago, harry_rag said:

A couple of questions; what are PS and PP and what time period does this data cover?

PS is proportional stake (to make 1 point profit), PP is proportional profit.

The data covers this flat season, I will need to go back over the last 5 seasons to give me confidence in the results.

[harry_rag]

21 minutes ago, MCLARKE said:

PS is proportional stake (to make 1 point profit), PP is proportional 

Still a bit bemused, can you give me an example based on how you would stake a selection at a particular price. I’m struggling with the concept of a staking plan that turned the level stakes loss of 353 points on the <100 category into a 439 point profit! 

[MCLARKE]

The 439 is the proportional stake, 36 is the proportional loss, i.e. a ROI of -8%. I prefer it as a measure of profitability as it takes out the impact of any big priced winners.

I normally show returns on here to level stakes as this is easier for people to understand. 

 

[MCLARKE]

5 hours ago, harry_rag said:

So if I’ve understood this right the first table is all horses and this one filters out those that haven’t achieved a rating of 67+, meaning that almost 69% of all horses have achieved such a rating? What strikes me most here is the really low volume in the two higher bands. Only 27% of those horses with an indexed rating of 110+ have a speed figure of 67+ compared to 80% for the two lower bands. Somehow that seems counterintuitive. The majority of the best horses based on index score have a rating below 67. 

For this analysis I am only using ratings achieved in the current year. There are therefore a lot of horses that are making their seasonal debut or have not run before.

Altogether there were 6,819 runners that did not record a speed figure and these are excluded from the analysis.

7,067 achieved a speed figure of 67 or above and 2,981 had a speed figure less than 67.

Not quite clear on why you believe it is counterintuitive. As you tighten the band from 1-100 to 67-100 the variances to the average are going to become lower. As an example I have a race where there are 3 runners with speed figures of 2, 70 and 76. In the first analysis there would be 2 horses with an index of 140+, in the second analysis they would be 96 and 104.

 

 

[harry_rag]

You’re taking my indexing approach into uncharted waters there. I’ve been looking at and thinking of races where every horse has a rating so the indexed figure relates to the whole field. I’ve no idea if it will work similar, better or worse where you are only indexing some of the runners. But you’re a lot better at doing this than me so I’m sure you’ll be able to extract some value from the concept.

ps - you might want to warn us all off that nag that posted a speed figure of 2! 

[MCLARKE]

This is another area where you can end up with paralysis by analysis !

There will be some races with only 1 horse that has a speed figure so that will end up with an index of 100.

I'll have a look at analysing by number / percentage of runners in a race that have a relevant speed figure.

[MCLARKE]

Step 3. Index value multiplied by forecast odds.

I have split this into 10 ranges.

So 0.10 is the highest numbers, from 765 to 97, 0.20 is the next 20%, from 96 to 66 etc.

I think the top 2 are encouraging, not sure what the bottom 4 are telling me !

 

[MCLARKE]

11 hours ago, MCLARKE said:

Current profit 114.33 points

COURSE TIME  SELECTION ODDS  BOOKMAKER PLACES  Thirsk 1.52 Art Design 4.00 BETFAIR   Bath 2.30 Fregola 7.50 BETFAIR   Bath 3.30 Second Collection 9.00 BETFAIR   Thirsk 3.52 Vince Le Prince 6.00 LADBROKES   Bath 4.00 Malham Tarn Cove 4.00 BETFAIR   Thirsk 4.22 Lord Abama 8.00 BETFAIR   Bath 4.30 Beggarman 15.00 LADBROKES 4 Thirsk 4.55 Equity's Darling 13.20 BETMGM   Bath 5.00 Rita Rana 6.50 BET VICTOR   Yarmouth 5.07 Chola Empire 8.50 BETFAIR   Bath 5.30 My Ambition 2.50 BETFAIR   Yarmouth 8.10 Fletcher's Flight 4.00 BETFAIR   Ffos Las 8.30 Zambezi Magic 8.50 BETFAIR

A run of 28 losses followed by 3 winners, it's a strange game sometimes. Managed a tiny profit for the day.

[harry_rag]

9 hours ago, MCLARKE said:

Step 3. Index value multiplied by forecast odds.

I have split this into 10 ranges.

So 0.10 is the highest numbers, from 765 to 97, 0.20 is the next 20%, from 96 to 66 etc.

I think the top 2 are encouraging, not sure what the bottom 4 are telling me !

 

Not sure how comparable this is to what I do. Firstly it’s (index value -100) x price and I only do it for those runners with an index value over 100. The idea is that it gives me some measure of value for all the runners that have a better than average rating.

[MCLARKE]

Yep, sorry, that's what I've done, so it is comparable to yours.

I'm now going back 5 years to see if I hopefully achieve similar results.

[MCLARKE]

Current profit 114.74 points

COURSE	TIME	SELECTION	ODDS	BOOKMAKER
Chepstow	3.35	Milliterries	3.50	LADBROKES
Leicester	3.45	Spring Chorus	7.00	BETFAIR
Chepstow	4.05	Howzak	2.25	BET VICTOR
Leicester	4.15	Admirable Lad	28.00	BETFAIR
Hamilton	4.25	Musical Diva	18.00	BETMGM
Chepstow	4.40	Penalty Shootout	5.50	BET VICTOR
Leicester	4.50	Carpathian	7.00	BETFAIR
Hamilton	4.55	Ledger	5.00	BET VICTOR
Leicester	5.25	Queens Road Revue	7.50	BETFAIR
Epsom Downs	5.55	Spell Master	3.50	BETFAIR
Epsom Downs	6.30	Charlie Mason	14.00	BETFAIR
Epsom Downs	7.00	Commander Crouch	4.75	LADBROKES
Epsom Downs	7.35	Roost	9.00	BETFAIR
Epsom Downs	8.05	Dream Of Mischief	11.00	BET VICTOR
Epsom Downs	8.40	Land Of Magic	6.00	BET VICTOR