Hi all,

In a recent thread on Value, there was some debate on the accuracy of the SP market as a predictor or chances. I've said in the past that on average, across all races, the market is the most accurate predictor of chances. Some disagreed with that statement and irrespective of that, the ensuing debate bought out some extremely valuable punting principles.

At the end of it all, I think everyone agree that our task as punters is really to outperform the accuracy of the market on the races we select to tackle. Further more, the key to success is only to tackle those races where we are confident we can outperform the accuracy of the market. As the saying goes..."you can beat a race, but you can't beat the races".

Yesterday I put together some numbers of the performance of the market as a predictor of chances based on all Metropolitan racing in NSW, QLD, VIC & SA for the past 44 months.

I have grouped the prices into ranges to simplify the presentation of information. The table below shows the expected and actual win rates based on the actual market SP prices, and then shows the expected adjusted win % which is calculated by normalising SP's back to a 100% market standard (instead of the average 120%). In other words, if our markets were set to 100%, what should the strike rate be for that odds range.

Price Range	Avg. Price	Exp Win %	Act. Win %	Gap	Adj. Exp Win%	Gap	Sample
$1.10-$1.90	$ 1.71	58.5%	52.3%	-10.6%	53.0%	-1.3%	834
$2.00-$3.90	$ 3.03	33.0%	30.4%	-7.9%	28.3%	7.5%	9851
$4.00-$5.90	$ 4.77	21.0%	18.7%	-10.8%	17.5%	6.8%	14686
$6.00-$7.40	$ 6.51	15.4%	13.3%	-13.4%	12.7%	4.9%	12007
$7.50-$8.90	$ 7.94	12.6%	9.7%	-23.0%	10.3%	-6.2%	8843
$ 9.00	$ 9.00	11.1%	8.7%	-21.7%	9.1%	-4.3%	5951
$ 10.00	$10.00	10.0%	7.8%	-22.0%	8.2%	-4.4%	5047
$ 11.00	$11.00	9.1%	6.9%	-24.1%	7.4%	-6.8%	5836

At SP prices there is clearly a gap between the actual win rate and the expected win rate, which is as we would expect in markets set to approx. 120%. The longer the price the larger the gap which indicates the long shot bias previously spoken about. However in the context of a 120% markets, SP prices are the single most powerful indicator of performance and finishing order.

When we look at the gap based on the 100% market expect strike rate (adj exp strike rate), we see that the market is quite accurate at predicting chances across thousands of races. In some price ranges, the strike rate is actually greater than what we would expect.

In saying that, this adjusted strike rate analysis isn't really of any use to us as our markets are never set to 100%, but nonetheless it shows how accurate the market is across all races.

More than anything, this confirms a few things successful punters already know:

* The market is generally very smart at predicting chances, especially at low odds ranges. You need to have your act together if you are going to out perform it.

* In reality there is no single punter that can predict chances more accurate than the market across all races (regardless of how good you think your rating numbers and pricing algorithm is)

* The key remains to have the discipline to wait and pick the spots where you are confident you can outperform the market....based on sound for assessment and reasoning.

In my view, the fundamental question when deciding to tackle a race is whether or not you believe you can outperform the market in its accuracy and why? Any punter who does not approach their activities in this way has in my view a slim chance of long term success.

Comments?


[ This Message was edited by: osulldj on 2002-12-13 15:26 ]

Quality post!

Have to agree here osulldj - quality post!

:wink:

You can't expect to win every race every day, it doesn't work like that, as you can see the odds are against you before you start.
If you can single out selections in races that have a better chance and are value - you might just keep your head above water to make a handy profit.
Backing races for the sake of it, obtaining poor value,not knowing enough about the horsebefore selecting it or chasing losses means you drown!
The odds are just too tight to outperform the market advantage every day.

Good work once again.

What a great read!

The notion of outperforming the market is an interesting concept.

If the market is more right than any single punter can establish, then can it be used as a tool to find value?

eg When the market selects two at $4 but they are priced on the tote just before the race at $7, what is happening there?

Otherwise by what you mean by outperforming the market, are you looking for gaps where it's not telling an accurate price and punters are willing to back it at the erroneous price?

There two ways to find the edge here.

does anyone know how this breaks down between country and city races...
i.e is the SP fav more likely to succeed in the country, or in the city????

just flicking through some numbers this week my gut feeling would favour country races, but????

see ya
Every Topic

HI Fryingpan,

You said:

"eg When the market selects two at $4 but they are priced on the tote just before the race at $7, what is happening there?"

I years gone by, tote crushers used to be quite common. Recognising that the SP market is the most accurate predictor of chances, they would wait until the last minute of betting and then back all the significant overs on the tote. However as happens with all "simple" strategies like this they don't last. These days, even if your selection is big overs on the tote, the last fluctuation, which often takes place after they jump, sees the horse backed into equal to or less than its SP. Of course there are isolted instances we could all sight where we have gained tote overs, but across a large number of races the two markets are similar. If anyhting the tote offers poorer value for shorter priced runners and better value for the long priced ones.

In outperforming the market, I always look for one or more key horses. Those key horses are either runners I believe the market has significantly underassessed, or on many occassions, runners the market has significantly overassessed.

If you have it assessed pretty much the same as the market then you are best off to leave the race alone. Sure your ratings might say the horse is $4 and the market has it at $4.50, but we know upfront that the market offers very poor value across all races and if you have this race generally assessed the same as the market, do you really think $4.50 against your $4 is value?

Every topic,

Following are favourite stats since Jan 01 this year:

Country: 1079/3678 (29.3% SR) loss 16.5%
Provincial: 1558/4745 (32.8% SR) loss 8.5%
Metro: 1019/3483 (29.3% SR) loss 14.3%

Total: 3656/11906 (30.7% SR) loss 12.7%

Remember Don Scott's research in Winning More. He found that in the 1/1 to 5/1 price range, bookmakers only worked to a 5% margin.

In the early 90's Superform looked at it too and found that when their top-rated selection was in the 1/1-7/2 range they made something like a 29% profit.

all the current literature suggests we should be looking for "overlays" i.e. a horse showing Bookmaker odds greater than our own assessed odds.
In this age of internet form everyone has access to form history. The question may be not "can I may make an overlay" but "can I make a profit on this race"?
Last saturday's Sandown meeting was a "Wake-up" call to me - I rated 7 races and could only find ONE race were I could "Dutch-book" three horses to make a profit.
I think this may be the the most important lesson - If you can't BY YOUR OWN ASSESSMENTS make a profit on this race - DON'T Bet.
One a more positive note, I have "stolen" the ideas from forum members about sectional times and the "pace" of races to create a different set of ratings.
The first weeks are too god to be true - time will well!!

thanks for the numbers osulldj,
do you also have the same breakdown for city/country/prov for the favourite running a place this year????

look forward to your reply,
thanks
Every Topic

zorro wrote...
"One a more positive note, I have "stolen" the ideas from forum members about sectional times and the "pace" of races to create a different set of ratings.
The first weeks are too god to be true - time will well!! "

now zorro, if you are going to generate success from the ideas handed out freely in this forum, at least you could not taunt us with the goodies and instead share your findings for us all to see.

whilst I respect the right of all of you to keep your own secret systems to yourselves, I think its only right that if you take ideas from the contributions of the members and turn them into something useful then you share the new method with all.

if we dont do this the contributions might dry up :smile:

have a successful day,
Every Topic

Every topic,

Place Stats for favourites this year:

Country: 2243/3678 (61% SR) loss 8.9% (NSW tab)
Provincial: 2949/4745 (62.1% SR) loss 8.6%
Metro: 2115/3483 (60.7% SR) loss 7.1%

Total: 7307/11906 (61.4% SR) loss 8.2%

In reply to Every Topic:
The approach I am developing is to-
1) Get a horse's actual last 600m time using the Sportscolour web-site, ignoring the occasionally crazy times they post.
2)Adjust this time for the Daily Track Variant i.e. the fastness/slowness of the track on that day.
3)Compare the adjusted time with the Par time for that track/distance and thus obtain a rating for the horse.
4) Adjust the rating for the pace of the race.e. g . if they crawl for the early sections of a race the last 600m will naturally be proportionally faster than normal so the rating should be down-graded.
Of course the integral part of the whole deal is determining the par times and the standard % that the last 600 represents of the total race time.
Hope this gives you some ideas.

hey Zorro,

Quite a few years ago now (between 1994-97) I used to do something similar. But the actual time they ran wasn't important. Actually, I didn't the study times they ran.

I used to buy the Sporting Globe each week and use their brilliant Race Photos for the 800m, 400m and wining post. I would track each horse for the last 400m and measure how many lengths it gained or lost in the business end of the race, and give each horse a rating (100 being the best for the race).

You could then compare the performances of each horse in the race to one another. I always took special notes of races where only 1 or 2 horses in a race ran the last 800m (or 400m) at least 2.5 lengths faster than the rest of the field. It was a brilliant pointer to a horse improving and running well in its next 2 starts.

I would look at the pace of a race, and make a judgement on whether it was fast early, slow etc, to further analyse the merit of these runs.

I really noticed some quite excelent performances with that method, and quite a lot of these horses went on to win at Group level soon after. So you can back them as they step up through the grades.

My advice is to keep it up, it's a good method, I found it very rewarding. A lot of value in my selections, as it wasn't a common method of analysis.

Good luck.

The true beauty in compiling SP and actual win results by probs breakdown is it gives a basis to do further studies on jockeys & barriers.
Like checking every ride of each jockey based on SP probabilities then compare to the Overal results and so on thru the barriers..
That is if you add the probs for ech ride say Jockey A has been lucky an has ridden $2.00 chances in 10 starts, you would add .5+.5+.5+.5+5 and so on for the 10 starts the total would then be his expected strike rate. It soon becomes clear when you add a second variable like Barrier the true abilities of each jockey and from different sections of the Barrier, this can be applied in their ability in Field size ..the list is endless..if you think about it..

[ This Message was edited by: woof43 on 2003-01-04 10:08 ]

The power of race times is in knowing the acceleration and deceleration of horses of every age and carrying every weight, with this knowledge one is able to make comparision between horses, based on each horses past performances.
Once you have the comparision it gives you the basis to being able to assign probabilities to each horse.
A hint to using Time & Weight linear regression isn't in looking at the distance of todays race, but the avg distances that all horses have competed over in Todays race.