From my limited reading the two most practiced forms of ratings are the Weight(Don Scott) ratings and the American Speed (Andrew Beyer) ratings.
Both types of ratings seem to rely on selecting a "base" run from which various factors are added/subtracted to develop a final rating.
Obviously if you don't select the horse's base run correctly all the extra adding and minusing are a waste of time.
Which brings me to the question to the ratings guys on the forum - which run do you use?
As an example I normally use the most recent run, except:
1. When the horse is first-up,
2. When the last run/s have been over unsuitable distances,
3. When the runs have been on rain-affected tracks,
etc,etc
In the end, is it sensible to be using just one run to base the ratings on, or should we be using some sort of weighted-average?
I doubt if there's a definitive answer but I'd love to know how other people tackle the problem.

Hi,
Depending on how you first formulate your ratings my way of thinking is you need a Life to Date average and a LTD Standard Deviation these would become your Base starting points.
Then slight adjustments are made to both of these figures for various Variables, ie a Variable maybe Barrier Position, you then find the average an Std dev for all starts from similiar positions if say you have 5 starts from Todays BP and there have been 20 LTD starts which equates to 25 % then you make that sort of adjustment to both the LTD average an LTD Std Dev.
You do this over an over for the various variables..hope this gives you some food for thought
Cheers

Some ratings use best of last 2 runs, while others extend back to the best of the last 5 runs with some weighting to recent runs. Of course there are exceptions for first up, first start, second up, etc.

Hi, the tough part is in deciding whether something you are observing is actually "form" or just a random series of events. If you are playing roulette and red comes up twice in a row, is red "in form" now? You know the answer is "no" because of statistics and probabilities. But with racing many people draw a different conclusion. A horse has two good races in a row and he's "in form." But how do you KNOW it's form and not just a statistical coincidence?

Well, the way that you determine that is by actually studying hundreds of horses. If such a thing as "form" exists, then you should be able to see it by using sequential comparisons. A performance in a given race should be closer to a performance in his previous race than to his career average. (You have to be VERY careful when doing this, or you will see CLASS ladder artifacts and mistake them for form, or you will see lifecycle effects at the beginning and end of careers.) But I've done study after study and never seen a statistically-significant relationship between adjacent performances. In every study, a horses adjacent performance look exactly like what you would expect from random performance around their average talent level, given their standard deviation. Thus my conclusion, at least with current evidence, that form doesn't exist as a predictive factor.
Cheers

It may not be predictive but i would rather back a horse that had 2 good runs than a horse that ran last at both it's last starts

Humans are pattern-finders. Evolution has put us in a position where we try to see patterns in everything, even where they do not exist. Look at a series of long-term past performance lines for nearly ANY horse. You'll see a 7th-place finish followed by a 2nd-place finish, followed by a 5th-place finish followed by a 1st-place finish. THAT is the most common kind of pattern. The usual thing you observe is that a horse's next line DOESN'T look like the previous line. So when two lines "match" it sticks out and people jump on it and say "form" (good or bad) because two in a row are identical. Of course when that "form" doesn't pan out in the next race (and it most often DOESN'T, of course) the failure of the form play is quickly forgotten, and the excuse of the day comes in. If people actually stopped and THOUGHT about it for a minute, if form actually existed, then favorites would probably win a lot more often than 25% of the time, right????

Actually Woof most studies show favourites win about 1 out of every 3 races = 33%.
I agree that using past finishing positions to establish a form pattern is dangerous. There is a world of difference between 5th place, beaten 1 length and 5th place, beaten 12 lengths. That's why I reckon finishing margins are much more useful as an indicator.

Woof....the thing you are missing is there is a lot more to form than just where a horse finished in its last race or even its last 3 races.....let me explain the way i would read a horses form.
First thing i look at the finished lenghts of its last 4 runs i don't care what position it finished in just how far back it was i look at the type of race it was in the class of race the weight it carried and the track condition....then i try to compare it with the rest of the field maybe the horse has had a couple of bad runs it may have been wet and the horse doesn't like the wet....it has been wet in Sydney over easter so you might see some form reversals when it dries up....other reasons for bad runs in racing out of class....some horses can only perform in thier class level this is another thing you need to look at.....maybe it carried to much weight last time in.....most horses that win at big odds if you look more closely at thier form after the race you can often see why they won.....you have to also remember that some horses gain condition more quickly than others meaning they can win first up or second up...others need time there are patterns in a horses form you need to look at maybe after say 2 runs they improve very quickly you can often see this if you look at there races....the reasons favs don't win more than 30% of races it that people over estimate the best horse in the race and underestimate the next 4 or 5 and you have to remember that all important factor the jockey.....why did the $1.30 fav Lohnro not win the Jock says he stuffed up....if he had gone earlier on him would he have beet Grand Armee who knows....but if you had looked at Gran Armees last couple of runs and the prices the horse had started at the horse was underestinated at all 3 starts the horese had a preety impressive record...but just came up against a better horse on the day....the truth is most people don't really read the form and the bookies have to set a market so they make money other wise they may have 5 horses under 5 dollars you won't see that the tab prices are just the opinion of illinformed people.....do you really think the that people that make money out of backing winners put thier money on at the tab....don't think so....to you question does form exist.....you bet it does....but how you understand the form that you are reading depends on the individual....every one reads form differently....who is right and who is wrong....well at the end of the day or year whe ever has the most money is the winner....thats how you know

To woof43,

I've seen a few of your posts talking about the monte carlo method. I've downloaded and setup the SimTools Excel add-in, also found one called MCsim. Trouble is, I don't know where to start! Could you possibly give me a brief step by step explanation on how to run SimTools?

Many thanks,

If you post your email address i will send a simple 1000 or so simulator to gain an understanding.
Cheers