Interesting discussion on ratings and betting approach

[mardigras]

I found these articles by Greg Polson/Racenet.

EDUCATION: Is it profitable to be backing favourites? (racenet.com.au)

Expert profile: How Greg Polson continues to be a winning punter - Racenet

For me they are both somewhat flawed. 

Suggests that the betting public provides the best assessment of chance, and that horses $10 or under in the market are the ones he would tend to focus on.

Then somewhat contradicts all that when he then uses his own assessment to determine stake in a staking plan, and uses an example of a runner priced at greater than $10. If his assessment is the one to trust afterall, then the public assessment is irrelevant as you are not trusting it to be the most accurate. So you can ignore whether the price is shorter than or longer than $10. And if the public assessment is the accurate one, why would you place a stake based on your own assessment knowing it is not as accurate as the public one?

In his tables he talks about the loss on turnover from different price points.

The main issue with his tables is that he is using a poor assessment of chance. The bookies are not an accurate assessment of chance as they are not even close to a 100% market, they will take their fat largely from the lower price end of the market, and equally take ample gains from the higher price end of the market due to the longshot bias that still exists today in the less sophisticated punting group.

Less sophisticated punters are likely to bet with the TAB/corporates. They are also likely to take bets on longshots that are not reflective of chance, allowing those operators, long term easy money on long priced horses priced well below chance - hence what you see in the tables.

If you took something like betfair as an indicator of chance, you would find a different story to the supposed proof that the best chance of winning (profiting) is at the shorter end of the market. That table is not proof of anything, except an inability to understand why his table reads that way.

Here is a table of betfair returns using the same prices in the tables. (and for a much larger number of runners)

Odds	Runners	Winners	Odds Probability	Resulting Probability	Return	Return%	Profit/(Loss) on Turnover
$1.50	515	338	66.7%	65.6%	$507.00	0.984	-1.6%
$1.80	1075	592	55.6%	55.1%	$1,065.60	0.991	-0.9%
$2.00	2423	1162	50.0%	48.0%	$2,324.00	0.959	-4.1%
$2.50	3758	1542	40.0%	41.0%	$3,855.00	1.026	2.6%
$3.00	8225	2805	33.3%	34.1%	$8,415.00	1.023	2.3%
$4.00	20585	5177	25.0%	25.1%	$20,708.00	1.006	0.6%
$5.00	27694	5579	20.0%	20.1%	$27,895.00	1.007	0.7%
$10.00	59503	5803	10.0%	9.8%	$58,030.00	0.975	-2.5%
$20.00	57374	2842	5.0%	5.0%	$56,840.00	0.991	-0.9%
$50.00	57603	1159	2.0%	2.0%	$57,950.00	1.006	0.6%
$100.00	37848	402	1.0%	1.1%	$40,200.00	1.062	6.2%

 

This highlights that when looking at a better representation of chance, there is little variation around which price point is likely to give the best chance of winning (profiting). Actually, the highest price shown gives the best returns anyway. The issue being that he is relating a table that has no bearing on probability and suggesting you can therefore use this information to help you make better returns. It simply is not the case.

 

Another of the issues with the information is that by suggesting the shorter end of the market is the best way to give you the best chance of winning, then that is demonstrating a lack of understanding around statistics. You can't apply what is a population statistic to an individual race. So to assert that $2 runners at the bookies have a 45% chance of winning across the population (as per the table in the links) - is not the same as saying that a runner at $2 with the bookies in an upcoming event, has a 45% chance of winning. Statistics do not work that way.

 

If the punters provided the best assessment of probability in each race, then it would be difficult for a single punter to win given there is always margin. Punters may provide a decent assessment of probability across a population of races, but what a punter needs to do is determine where the punters have it wrong on an individual race level. That is where you are wanting the punters assessment to be wrong in that case, so that you are able to take advantage of that.

 

[curious]

Very good assessment in my view and I don't really see the point of the argument that if you concentrate on lower priced horses, you lose less. Most of us actually want to win, not lose less. My own data from multiple jurisdictions, like the BF data you cite suggests that longer priced horses tend to be more overpriced and more profitable. The problem with that from my perspective is that although they might represent better value, they also create the longest losing streaks, certainly beyond what I can tolerate. So I'm inclined to concentrate my betting on horses that I assess as better than a 10/1 chance.