Rated price accuracy

[AngryPixie]

As you know I'm not much of a ratings guy. One of the issues I have with the various methods used to convert ratings to prices is their long term accuracy. For example do your Evens horses win 50% of the time? Do your 2/1 priced horses win one in three? etc etc.

My view is that ratings are of most use in ranking the selections, not pricing them

Though it may be an interesting discussion point.

[moeee]

do your Evens horses win 50% of the time? Do your 2/1 priced horses win one in three? etc etc.

This is probably the best and most useful test a Rater could perform , but because I am frightened for some reason to do it with My Markets , I can't help at all.

I was almost to do it one time , but then I figured that supposing my animals weren't performing as they should? - what now?
I figure there would be errors in all brackets of animals.
the percentage error in each bracket should be identical.

Woof43 knows a little about this stuff.
Hopefully he notices this thread , and if he don't , I will alert him at my next opportunity.
The only problem is that he explain his thoughts so's other members can understand.
Although Angry Pixie , you seem a little better versed than most in WoofSpeak so you should be fine :)

[moeee]

This is probably the best and most useful test a Rater could perform , but because I am frightened for some reason to do it with My Markets , I can't help at all.

Well thats the most innacurate and most misleading post I have ever posted.

I just realized that getting the animals to win in proportion to their assessed Chances is totally useless.
If your system did that , then you would be ending up with the General Publics Market.
The worthiness of a Ratings Market is in the profitability of the Overlays - NOTHING ELSE IS OF RELEVANCE.

[AngryPixie]

I just realized that getting the animals to win in proportion to their assessed Chances is totally useless.
If your system did that , then you would be ending up with the General Publics Market.

You want them to win equal to or better than your assessed chances don't you?


The worthiness of a Ratings Market is in the profitability of the Overlays - NOTHING ELSE IS OF RELEVANCE.

Indeed. The overlay comes when your assessed chance differs to the positive from the chance arrived at by the market. To make a long term profit your assessed chance would therefore need to be more accurate than the market. This may be possible on select races but I doubt anybody has achieved this across all races.

[woof43]

First of all you need to make sure your ratings/points have been compiled with a statistically valid method.

I have attached a sheet that shows how many points or weighting would be assigned for a Factor or each individual rank within a Factor.
I think I may have posted this a long time ago, it shows the weight of the Factor and the ROE just check the ROE +/- against your selected confidence level

Remember you need to measure your handicapping with Strike rate and Wagering with ROI.

The general public odds need to be converted to Natural Odds.

Your rated 20% chance indeed needs to win 20% of the time, this in itself is very hard to measure, as it is very rare that you will get a large enough sample of exactly 20% rated chances to test a small deviation makes a large difference.

Also your rated 20% chances if you had a large enough sample would win 20% of races right thru all the odds ranges. ie if I had 100 of these .20 chances and they started at even money i'd expect them to win 20 races with a normal variance.
hopefully this is understood.

[AngryPixie]

Your rated 20% chance indeed needs to win 20% of the time, this in itself is very hard to measure, as it is very rare that you will get a large enough sample of exactly 20% rated chances to test a small deviation makes a large difference.


Woof you could group them +/- 2.5%. It would be close enough to be indicative. Some of the long time ratings bod's would have plenty of past results at their disposal.

I guess I'm not suprised that this thread has had little interest. As moeee suggests it's probably a frightening prospect for many. Variance is nice when it's going your way ;)

[moeee]

I guess I'm not suprised that this thread has had little interest.
Forums are like Radio Stations.
They attract people with different qualities.
It seems this Forum is for those that like to have fun.
Prospective Winners will have to look elsewhere.

[AngryPixie]

It's all easy listening for me these days. :)

Another useful exercise would be to look at your top rated selections that went off under your price and therefore weren't bet. Are these winning more than your prices would suggest they should?

[mattio]

Actually I have been trialling the use of a set of rated prices to find overlays that I plan to lay, testing so far has yielded around 1600 selections for a 95% lay strike rate with average winner (or loser in this case) around $9. I have been testing since December 2011 and will start live betting at the end of the month.

My thoughts were like most peoples in that overlays were the key to value and therefore profits when backing but I soon found that the market is a much better guide and I found that horses starting as "unders" had a better winning strike rate and lost less than those that would have been considered "overs" or value.

Of course not all overlays are bad value, I have introduced a few filters to maximise the layability of selections but it has made me re-think the notion of "always bet the value selections".

Cheers,

Mat.

[beton]

Mattio
Interesting finding. About 6 mths testing. Is it all days, all venues. all goings?
Joe Public is reasonably good picking the toppies and as such these are usually the unders. They shun the potential also rans. These become the overs. You are using ratings to determine the overs, meaning that rating system has determined these horses better than what they actually prove to be. There are two types of overs, the first being the rated assessment of each of the horses compared with each other, and the second being the straight out numerical breakeven point. One is used when you are betting one race at a time and the other when betting en masse. The first is only an opinion. The second is useless if you are only betting an occassional race. What filters are you using? Beton

[Chrome Prince]

My thoughts were like most peoples in that overlays were the key to value and therefore profits when backing but I soon found that the market is a much better guide and I found that horses starting as "unders" had a better winning strike rate and lost less than those that would have been considered "overs" or value.


Matt,

And this is the inherent flaw with overs unders betting / laying.
Ratings tell us what we know, fluctuations tell us what we don't in general.
One can definitely profit from overs betting, there's no argument about that, but perhaps there needs to be something incorporated into ratings which recalculates them by a factor given market information.

Because horses are not teams nor sportsmen, they are animals.

One can rate AFL or Tennis or Soccer and come up with definite odds and overlays, because in general, it is what it is.
With horses sometimes it is what it isn't!
All the stats in the world can't tell you stable opinion, or closed trial form, or negative medical issues.

[AngryPixie]

My thoughts were like most peoples in that overlays were the key to value and therefore profits when backing but I soon found that the market is a much better guide and I found that horses starting as "unders" had a better winning strike rate and lost less than those that would have been considered "overs" or value.


Matt I've had success in the past backing the top selection of those "Crowd Ratings" at unders. If I could get under the rated price I'd back the selection, anything over I'd leave. Worked very well on the top selection, the second top roughly broke even and from third selection and higher you did better backing the overs and leaving the unders. Clearly my pricing needed to give more emphasis to the top two, therefore evening out the gaps. But there's not much use pricing your selections so they match the market ;)

[Chrome Prince]

Agreed AP,

In fact as a method, many people note down the prices 30 mins to the off on UK racing and back anything under the established market price just before the off.
If ever there was proof that this works, it has been my experience years ago of losing heavily when I was laying early at the given price. The horses that drifted did not win as often as the horses that were nibbled at or plunged on.
So I would always get matched on the money horses and not matched on the unwanted. The firmers were profit backing, the drifters were a huge loss.

But these are long term results, you can have really good or bad days or weeks.

[mattio]

Pixie, I have found that focusing on the first 3 favourites primarily has been where the majority of the success is in the testing of this idea although there have been some big profits in the higher end of the scale as well.

The key to what makes this idea a success is by using a multiplier applied to the stake based on the overlay, the higher the overlay the higher the stake but I have a buffer built in by laying to liability. I have been testing with a $30 base liability and applying the multiplier to that, the biggest loss has been $146 with the applied multiplier.

I am also looking at a backing system for unders and so far I have found that of the selctions I am looking at (favourites only with a few filters) the most profitable band has been when the price falls to 60% or below the rated price I use. Same type of multiplier used to maximise the returns, so far there has been around 200 selections (since December) for a 45% S/R and 20% POT at TAB prices.

Chrome, I am yet to look at how fluctuations affect these systems but it is something that I will definitely look into.

Cheers,

Mat.

[mattio]

Mattio
Interesting finding. About 6 mths testing. Is it all days, all venues. all goings?
Joe Public is reasonably good picking the toppies and as such these are usually the unders. They shun the potential also rans. These become the overs. You are using ratings to determine the overs, meaning that rating system has determined these horses better than what they actually prove to be. There are two types of overs, the first being the rated assessment of each of the horses compared with each other, and the second being the straight out numerical breakeven point. One is used when you are betting one race at a time and the other when betting en masse. The first is only an opinion. The second is useless if you are only betting an occassional race. What filters are you using? BetonBeton the testing has been on all days and venues and goings, the filters are nothing out of the ordinary, just things like weight, barrier, distance etc. The key I believe is the overlay based on the rated price I use which so far has proven to be successful but only when I start to place real bets is when I will know for sure.

Cheers,

Mat.

[Vortech]

Your rated 20% chance indeed needs to win 20% of the time, this in itself is very hard to measure, as it is very rare that you will get a large enough sample of exactly 20% rated chances to test a small deviation makes a large difference.

Also your rated 20% chances if you had a large enough sample would win 20% of races right thru all the odds ranges. ie if I had 100 of these .20 chances and they started at even money i'd expect them to win 20 races with a normal variance.
hopefully this is understood.So basically, a ratings system that has a predicted price of $5.00 should long-term have a strike rate of 20%. If it doesn't its impossible to get a correct overlay.

This is all new to me this area of punting, so if I understand the hard thing to predict would be the strike rate is 20% is on all bets if correct.

But if a punter was to bet only on the overlays, this might only have a 10% strike rate because the underlay bets are at 30% making an average of 20%.

In an ideal world if you could obtain $6.00 on every bet rated at $5.00 but often there will be a short $2.00 horse in the mix.

[mattio]

Vortech I would be very surprised if any ratings system was accurate enough to have the $5 chances win 20% of the time etc because as Chrome Prince has pointed out that ratings are based off information we know whereas the price fluctuations are based on information we don't know (basically) which is why a horse might firm or drift late in the betting.

As far as overs winning 10% and unders winning 30% of the time and evening out to 20% it isn't really that simple but in the testing I have done so far with the prices I use the unders have a better strike rate at the pointy end of the market.

Cheers,

Mat.

[Lord Greystoke]

Mattio,
Would the general conclusion be that we don't necessarily get 'value' with unders, but they win up to 3x more often(than overs) due to positive factors we are unaware of, so we have a better chance of a POT over the long run?

This would make price fluctuations the ultimate filter to be applied?

LG

[Vortech]

Do you have to incorporate both your ratings value and the starting price and then somehow factor price movements.

[mattio]

Mattio,
Would the general conclusion be that we don't necessarily get 'value' with unders, but they win up to 3x more often(than overs) due to positive factors we are unaware of, so we have a better chance of a POT over the long run?

This would make price fluctuations the ultimate filter to be applied?

LGLG, I can only speak from my experience and the testing I am currently doing so I don't want to draw any general conclusions but relative to what I am doing I would say yes. I am still to incorporate price movements as a filter.

Cheers,

Mat.

[mattio]

Do you have to incorporate both your ratings value and the starting price and then somehow factor price movements.All I currently look at is rated price and starting price, no price movements at the moment.

[Vortech]

I'm just trying to get my head around overs and unders. At the moment I don't see the logic.

If a horse has $2.00 fixed odds in Race 1

On racing and sports the don scott rating might be $1.50, hence a overlay.
But another set of ratings might have this at 2.50 which is an underlay.

What set of ratings are correct? I would assume by being able to see over history - comparing both the strike rate of all $2.00 rated horses on don scott and the $2.00 rated chances on the other ratings.

Should you not see a 50% strike long term give or take.

If the strike rate over 50,000 bets (example) in racing and sports for $1.50 horses is 40%, then the price is not $1.50 but it should be set at $2.50.

What is the point to finding value in ratings when you are really underbetting in thinking its an overbet.

I suppose it is really difficult to measure as we don't know the true value of a horse and its chances. We can assume on history only.

[Lord Greystoke]

I suppose it is really difficult to measure as we don't know the true value of a horse and its chances. We can assume on history only.

Wouldn't the 'true value' of a horse be what the market prices it at the jump, this being a combination of facts about a horses chances which are known and unknown to Joe P?

A consistent ratings machine being just the starting point for a shortlist of runners which can win, the ultimate filter for those that are most likely to win being the movement of price from PP to open to jump?

LG

[mattio]

I'm just trying to get my head around overs and unders. At the moment I don't see the logic.

If a horse has $2.00 fixed odds in Race 1

On racing and sports the don scott rating might be $1.50, hence a overlay.
But another set of ratings might have this at 2.50 which is an underlay.

What set of ratings are correct? I would assume by being able to see over history - comparing both the strike rate of all $2.00 rated horses on don scott and the $2.00 rated chances on the other ratings.

Should you not see a 50% strike long term give or take.

If the strike rate over 50,000 bets (example) in racing and sports for $1.50 horses is 40%, then the price is not $1.50 but it should be set at $2.50.

What is the point to finding value in ratings when you are really underbetting in thinking its an overbet.

I suppose it is really difficult to measure as we don't know the true value of a horse and its chances. We can assume on history only.I can understand your point mate, what I am doing goes against everything I have ever been taught but it has found a very solid profit during the testing period so I am going to run with it. As far as what ratings do you follow only you can decide that, what works in one scenario won't work in another as ratings are very subjective depending on the method.

[moeee]

in the testing I have done so far with the prices I use the unders have a better strike rate at the pointy end of the market.

Perhaps you need to take a closer look at your method of converting your ratings into Prices.
Could you post your method , or would you prefer to keep it a secret?

[mattio]

I'll keep the ratings method to myself (although its really nothing special) but the pricing method I use is one that was posted on here a while back from a Betfair article that uses a power number. I think you mentioned that you use it for your greyhound ratings?

I found the whole idea quite by accident when I was looking at ways to filter the ratings, I just put all selections in a spreadsheet with the rated price and SP (TAB not track) and then started a few "what if" scenarios.

[moeee]

Okay.
Then you have my full approval Mattio.
I would suggest that you increase the Power number so that your Odds a stretched a little more.
It seems your top selections are being undervalued consistantly.
Increase the power Number an amount that shortens your top selections a little more.

Woof43 wrote a good post very recently.
Can't remeber exactly what it was , but ain't hard to find.
Something about the frequency of your selections winning is about your handicapping skills.
And the strike rate of your selections is about your Pricing Methods.

Nah - best I find the post.
I'll chuck it in the EDIT.

EDIT : Its actually in this thread
"Remember you need to measure your handicapping with Strike rate and Wagering with ROI."

[AngryPixie]

I'll keep the ratings method to myself (although its really nothing special) but the pricing method I use is one that was posted on here a while back from a Betfair article that uses a power number. I think you mentioned that you use it for your greyhound ratings?

I found the whole idea quite by accident when I was looking at ways to filter the ratings, I just put all selections in a spreadsheet with the rated price and SP (TAB not track) and then started a few "what if" scenarios.

Matt

Yes I mentioned that method in a "Crowd Ratings" posting yesterday. The POWER figure is all important. 0.9 is about right for those ratings.

"Crowd Ratings" will be travelling to the west today. Just got out of bed so it's a little late to do anywhere else. :)

[mattio]

I use 0.9 for my ratings but sometimes I make adjustments anywhere up to 0.94 if the result seems too far out of whack.

[mattio]

Chrome, are you able to email me as I am working on something that I think you can help with - weststigers4life at hotmail dot com

Cheers,

Mat.