We're moving house


Please check your email for details on the move and PM Pete if you have any issues.
See more
See less

Population statistics

  • Filter
  • Time
  • Show
Clear All
new posts

  • Population statistics

    In my view, this is an area where the vast majority of punters operate.

    And when I read this elsewhere, "The reference to 3yos is entirely endorsed by the well proven wfa scale. As the season progresses the 3yos carry more weight as they mature and strengthen.", it made me think there is another Thomass out there prepared to put that in writing.

    The issue with population statistics is that as per statistical theory, they potentially apply to a population. They are not an event probability statistic.

    And one of the other very big issues with population statistics is that if indeed they were somewhat valid, they need to be taken in the context of the population they are based on - and importantly, what is know about that population.

    You can't take a population statistic based on a truly random selection of NZers suggesting that 51% of the population is female and apply that in any useful way.

    You ceratinly can't walk into an all girls college and start laying down bets that 49% of the students are going to be male. Information known about the population that varies from the information known in regards the population statistic, effectively debunk the statistic.

    As hard as it is for punters to not think about population statistics when betting, that is what you should do imo (largely because the population statistic is flawed from an event probability point of view).
    Even if you have little or no way to work out something about a horse, why would you assume it follows a statistic (which the majority of other punters have assumed - and therefore the price is also reflecting that belief).

    Population statistics have zero correlation with the chance of a horse in a race.

    Let's take an overall Australasian population statistic. 20% of the population of horses that won their last start, win their next start.

    But what does that mean for the chances of horse A. Does it have a 20% chance of winning. No. Does it have an improved chance of winning compared to if it had not won? Maybe, but not because it won as such.

    Winx doesn't have a 20% chance of winning her next start because she won her last start. And a horse that races her that won its last start is unlikely to have a 20% chance of winning either.

    It's like barrier draws. People often want barrier draw stats for track/distance combinations. Apart from they are often represented in such a poor way, they get interpreted to mean something they don't. But even if it appeared that barriers 1 -3 were far more favourable for the 'population' to have raced there, than other barriers, does that mean that is the case for the horse being assessed for the current race.

    Things such as what Barry writes about where the individual is assessed against race tempo and its expected running position might assist with determining the usefulness or detriment of a barrier position. But human behaviour will be to subjectively be thinking that horse A is much less of a betting proposition due to its supposedly poor barrier etc - all because of the supposed notion around the benefits of barrier position.

    And I'll add add one extra key piece since people are going to continue to use such things. Be mindful of the ones that show a straight percentage. The chances (if even for all) of a horse drawn 15 in a 15 horse field is 1/15th. The chances (if even for all) of a horse drawn 1 in a field of 3 is 1/3. Most sites don't represent stats apportioned as 1/starters for each start the barrier has had. So inside barriers will often have higher % winning since some of the races have less starters when compared to barriers that are wider.

    And also, be mindful of the fact that the statistic has zero input from the perceived chances of the horses. If there have been 100 races from barrier 1 and 10 races from barrier 15. The outer barrier sample sizes is getting increasingly smaller. And the volatility around the makeup of the horses in those barriers therefore becoming increasingly important as to its 'winning' chance.

    On the quote at the top. Horses mature over time. The WFA scale endorses that. Which is true for the population and the WFA scale is based on the population. Nowhere does it say that an individual follows any such progression.

    So anyone betting on a horse improving because it is a 3yo after Christmas is NOT betting on a horse, they are betting on a population and prepared to risk their money on the horse following a population trend.

    It's where other punters are profiting. So many punters assess races in this type of fashion, all believing that all the horses in race X are following the population trends and statistics. Which is skewing the prices allowing others to profit for not adhering to such rules.

    Such as 3yo after Christmas
    Down in grade
    Apprentices on in the wet
    Blinkers on
    1kg = 3/4L

    and many more. None of them consider the individual horse and whether they are going to benefit and by what degree.

    I tend to bet more on horse that go against convention simply because the price is favourably impacted by people following conventions around weight and barrier. So horses go up in weight (or up the top of the handicap), or supposedly disadvantaged by its draw, purely because their price is higher due to punter behaviour.

  • #2
    I take your point Mardi and as you know have often commented on the utility (perhaps that should be futility) of using population stats in the way that media and many punters do and I generally concur with your comments above.

    However, I have begun to wonder whether they may actually be quite useful. A few months ago, I did a little hypothetical study. 100 consecutive win bets, so not very robust I should note. I had the 100 consecutive bets on all horses with barrier draws >7; carrying more weight than they did in their last start; and where they finished within 5l of the winner in their last start.

    ROI 127%. Much better than I can otherwise achieve. Obviously, this scheme needs more work but it appears to be loosely associated with backing horses which the markets are overpricing relative to chance because of the way markets generally use population stats.


    • #3
      Originally posted by curious View Post

      ROI 127%. Much better than I can otherwise achieve. Obviously, this scheme needs more work but it appears to be loosely associated with backing horses which the markets are overpricing relative to chance because of the way markets generally use population stats.
      Nice curious. And I definitely agree with the concept of taking horses that the market is under-assessing purely due to those types of things. And such an approach may actually even work long term or for a decent period anyway.

      But, if you used the same approach to further identifying individuals that were assessed at better than market on top of the impact those attributes have likely had on that market, then the results should be better.

      Each of those horses have probably had their price positively impacted (for the punter) by the negative views of those attributes. So on that basis alone, you may be looking at value in a high number of cases if the assessment pre those impacts was somewhat reasonable.

      It's no doubt why I back a lot of horses that are drawn wide and carrying higher weights. I'm betting against the other punters and many of them are the reason for the value. But I would still to veer away from doing it on a 'totality' basis.


      • #4
        I have been sick in bed but when I wasn't sleeping, I was thinking as I tend to do. Anyhoo, I came up with a little play that combined the race pace of barryb, the value of mardigras and the population stat of days from last start.

        So far:

        88 bets
        $117 return
        $29 profit
        32.95% ROI

        Ascot Park 31 Mar.
        Morena Roc $12
        Mac Murdoch $26

        Phar Lap Raceway 03 Apr.
        Midsummer Magic $10
        Sure Is $11

        Pukekohe 07 Apr.
        Hammer King (best longshot) $13
        Highland Dancer $11

        Wingatui 11 Apr.
        Double Down $21

        Arawa Park 12 Apr.
        Suffice To Say $11

        Sometimes there are up to 4 bets in a race, sometimes just one. I did have my own ratings based upon class but I didn't have the energy to keep them up. They also showed a profit. Perhaps combining the 2 would refine it even further. Will have a go. At the very least a bit of fun.


        • #5
          Originally posted by LookingForValue View Post

          Ascot Park 31 Mar.
          Morena Roc $12
          Mac Murdoch $26
          Good one. Backed Mac Murdoch myself. I wasn't really sure there was much to the protest.


          • #6
            Had Le Cailloux at Otaki in three, paid $41 but I stop at $30 as I haven't the bank to bet waiting to hit such a beast. Did also have another that won at $100. Those bets build up when the favourites are winning. One day perhaps, if my little play has much more stats behind it.