Ditto,

Yes the last start winners ae overbet, the last start 2nds are underbet.

BUT look at the class of the opposition in regards to these horses.

CLUE.

CP

I use the stat "raw" in my method (one of 9 criterion) and the class of opposition last start doesn't come into my equation at any point. ALL of my selections MUST have finished 1st or 2nd last start to qualify.

G'day to Baggy! Nice to hear from you old mate!

Cheers

Privateer

Horses who run 3rd <=0.5L are significantly better than 2nds > 3L.

ROT: 87.8% versus 84.0%

S/R: 15.2% versus 14.4%

Runs: 8,326 versus 11,421

That's an interesting stat because in my computerised system I take no account of finishing position in past races merely lengths from the winner. This would mean that a horse running 3rd at 0.5L would get a much better past rating than a horse running second at 3L. Your figures show there isn't a huge amount of difference.
The comments made by other posters was prompting me to do a little research on this one and this reinforces it.

KV

Actually I thought there was a significant difference in favour of the one that ran 3rd.

Anyway I'd hate anyone to read too much into those instances and totally revise their ratings calculations.

For a start you have to ponder what it really means when a horse runs a close 3rd or a distant 2nd.

Please can you justify your use of "significantly better". A 3.8 percentage point increaes in ROT and a 0.8 percentage point increase in S/R does not qualify as significant in my book. the dictionary defines significant as 1. having or likely to have a major effect; or 2. fairly large in amount or quantity. i don't think either increase qualifys defined like this and could easy be result of chance.
Thank you. Winston.

"significant" in this context is a statistical term and can only be defined mathematically.

Those 2 samples of ~10,000 each were extracted from a database of over 1 million runs. They are far bigger than any comparable ones here, therefore more significant.

As a check I have just examined the performance of all 110,487 runners with a previous start 2nd.

The strike rate is 16.2%. That is 12.5% better than the 14.4% distant 2nds in the original sample.

I also measure excess wins and places over expected:

-W- .. -P-
56.7% 40.3% _ All prev 2nd
31.5% 25.0% _ prev 2nd > 3L
52.4% 38.3% _ prev 3rd <= 0.5L

Again there are significant differences.

G'day all!
Once again everyone is falling for the stats and ratings approach. All this handwringing about beaten lengths vs last start placing etc etc is a waste of time for the following reason: RACING IS NOT A SPORT. It is a business. Comparing human athletes/sportsmen to one another on recent (or not so recent) performance makes sense BUT NOT IN HORSERACING. We should not be interested in picking winners but, rather, in making money. Get that fixed in your mind and then it will all become much clearer.
Last Saturday Barbara Joseph took a horse from Canberra to Sydney for a First-Up Start. This should have immediately started alarm bells ringing. She was CLEARLY intending to win the race despite "the ratings" and the odds. At 40s SHE DID. It's a business; get that straight. Everything has a purpose (ie to make money) The previous week Diane Poideven-Lane put two "donkeys" in the same race (also Sydney) They ran 1st (at 40s) and 3rd (at 60s). how could they BOTH come good at once?? Answer; It's a business!!
On 13th Dec 2003 Ms Poideven-Lane took Terrific Taurus and Terrible Taurus to Rosehill (last race) to get 1st at 70s and 2nd at 100-1. Ratings, last start performance,race class etc etc meant (and mean) NOTHING when it comes to winning on the punt. For the trainers (and US) it is a BUSINESS. We are not in the game to get it "right" (ie rate the horses). We are in it to make money when the others get it wrong!!! When you buy a house or a share OR bet on a horse, you make extraordinary profits by "seeing" what the others have missed. Since everyone CAN see the last start results and everything else in the form guide OUR JOB is to understand what is IMPLIED in the guide. we have to understand what the trainer is up to AND what makes the majority of punters back THE WRONG HORSE.
The "wrong horse" is not "the loser"; it's the overbet horse which wins more than it's share BUT AT POOR ODDS. The "wrong horse" is almost always the obvious horse, the close up finisher who was fancied etc (yes even the "beaten fav") or the LSW now expected to do it again. OR the horse UP IN WEIGHT (ie down in class or "improving"). All too obvious and sure-fired losing propositions long-term. In the horseracing business like any other, the "product" is less important than what "management" DOES with the product!!!! That's it for now. Cheers.

Sorry but just because the stats comes from 1 million runs does not make them "signficant.". you said "significantly better" and I ask you again to justify how you can make that claim.
And the stats at the bottom of your most recent post are useless. please define how you "measure excess wins and places over expected:". until you have a valid measurement of "expected" then those stats are worthless. and you should know better than to ask us to accept them without such explanation. this is the sort of thing you berate others for so please dont fall into their trap.
Thank you. Winston.

oh and by the way you originally posted 15.2%. which is it, 15.2% or 16.2%. and that is not 12.5% better. "%" means per 100 you are getting 0.8 winners per hundred better which means 0.8% better. not terribly "significant.".
Thank you. Winston.

Winston,

As you appear to be a newcomer perhaps you may care to check out some of my earlier material where I tried to discuss proportional staking and related topics.

As far as I can tell there are no errors in my posted figures. The strike rate of 16.2% refers to the new control sample of 110,487 runners mentioned in the preceding sentence.

I believe the concept of expected wins and places is reasonably well known. I notice that ignorance of Actual/Expected happens to be a favourite bugbear of Anomaly Nick.

If you accept places as 1st, 2nd or 3rd, then for a field of N:

Expected Win = 1/N

Expected place = 3/N

just so I got this correct. in a field of 10 then
expected win = 1/10 = 10%
expected place = 3/10 = 30%
so i have 10% chance to win and 30% chance to place?
Thank you. Winston.

Very true. I've found a lot of work can go in to improving POT by a few measely % so the difference could be more significant than I think.
Be assured I won't be throwing out the baby with the bathwater, I'll just put in a few more tweaks based on position finished along with the other tweaks and see if it improves things. Then more likely than not abandon them as I do with the majority of things I try.

KV

Mr jfc.
you have had ample time to correct my mistake. you havent so I must be right. now to the consequences.
you are suggesting that every horse in a race of N has a 1/N chance of winning and a 3/N chance of placing.
please may i pick any horse and you give me N/1 to win and N/3 to place? you will be broke very quickly.
what you are doing is very similar to work of Roger Biggs and impact values. while i respect much of Rogers work i do not believe impact values are as good as the actual vs expected figures that Anomaly Nick talks much of.
if you were to show Nick the figures you suggest he will laugh his head off.

please do not make mistake that just becos i am new to this forum that i am uneducated in these matters. i feel that the above statement from you is very condescending.
Thank you. Winston.

Winston,

I did use Actual versus Expected expressed as a percentage difference.

Having just looked up Impact Values I note that they do not take field size into account whereas I base my Expected on field size.

Having again reviewed my material I can find no errors.

I don't understand your logic that if I don't correct your mistake then you are right. Presumably you meant to use some other word like "offering" rather than "mistake".

Typically I correct material when I believe there is a serious mistake likely to steer others in the wrong direction. I have noticed a number of mistakes you have made but did not comment because I assumed that most would be able to spot those errors, and it was of little consequence if they did not.

Mr jfc,
you convenient leave out my first part of post which asks if I pick a horse will you give N/1 odds to win and N/3 odds to place? this is at hart of the discussion so please to answer this part.

your figures are not "Actual versus Expected" under any dealing of this theory that i have seen and certainly not according to Anomaly Nicks interpretation of this figure. your "expected" is certainly not a true expected figure but more of a participation percentage. and this is why your figures are more like Rogers Impact Value figures.

by your own admission you have little knowledge of Impact Values therefore I respectfully submit you should not be making rash statement like this. Impact Values most certainly DO take into account field size.
knowing very much both impact values and actual versus expected theory i can tell you very much that your figures are more like impact values.

perhaps therein lie the problem.

now please mr jfc you have before berated others for calling mistake without showing proof or offering correction. please apply same expectation to you as to others.
Thank you. Winston.

Winston,


http://www.flatstats.co.uk/articles/impact_values.html

This free article defines and describes Impact Values, as well as providing a history of that methodology which preceded the computer commodity age.

Note that there is no mention of field size in that article.



You certainly are adamant that Impact Values do take field size into account, and that you are very knowledgeable about them. Unless you can prove that here in the face of the above contradictory material, I doubt that many will concur with your self-assessment.

Also do not misquote me. I have not admitted that I have little knowledge of Impact Value. Just because I just looked them up, does not imply that I gave them inadequate consideration, or that I made a rash statement.

Mr jfc,
hello. yes it is good article. have you actually read it?
i quote now from the article
if the 69234 runners mentioned were not the sum of the field sizes of the 6678 races then can you please tell me what they are? random number perhaps? i quote more.
IV=% of winners / % of runners
surely "% of runners" mean you must have field sizes for all races you consider?
is that enough proof for you?
Thank you. Winston.

This is precisely where one can go wrong with race modelling, or the information supplied can be misleading.

There are two glaring omission in these figures.

There is no research done on the form or class of those last time out winners.
At least half those runners had little or no chance of winning because of the massive class jump.

The conclusions drawn would be wrong.

Try the same statistics with top 3 API or career prizemoney and you'd see a massive jump in percentages.

Statistically, the last time out winners do win a lot of races, but are overbet in the market.

There is also no data on horses coming back from a lengthy spell, they could be last start winners also, but a great percentage could not win first up.

If one concludes they have a poor strike rate, one is not looking properly at the data.

I can get 48% win strike rate from last start winners with the correct filters compared with the 16.50% illustrated.

The author of the article has fallen into the greatest retrofitting pitfall of all time....

"But you should not look at the impact value figure in isolation. You should examine both the impact value and the ROI% figure.

A strong positive value for both of them is what you need to find as this indicates a group of horses who are winning more races than they should, and are going of at prices higher than they should."

This is the path to an empty wallet - guaranteed. He wants to use the impact values only which improve strike rate AND create profit, which is putting the cart before the horse.

The impact values should not be used on their own to "select" horses, as each horse has a different combination of factors and no two are the same, therefore the impact values, should be used to assign a rating to the horse based on combinations of POSITIVE impact ratios.

The final rating compared to the rest of the field will determine the fair price for the horse.

The punter will seek out the value runners for the race, and obtain value or pass the race.

Winston,

Why did you omit the telling remainder of that example?



The percentage of runners is # of LTO winners/ Total # of runners.

That does not tell you what field sizes those 5,548 runners competed in.

The average field size for them could have been much less than 69,234/6,678 (~=10) or it could have been much greater.

If the average field size those 5,548 ran in was relatively small then their # of wins should be flattered. Along with the IV.

Yet you ask:



Of course it doesn't mean that. What were the field sizes for those 5,548 runners? Even just the average or total. You do not have any of that information in that formula.

You claim to be

yet you don't seem to realise what erratic effects field sizes could have on them.

mr jfc.
one word answer please.
do you now concede that impact values take field size into account?
Thank you. Winston.

Winston,

I've already given you the answer. I've shown you that Impact Values take no account of field size.

Impact Values try to measure the effectiveness of a group with a certain characteristic.

But the process to determine an Impact Value ignores the field sizes of the races that group participated in.

Amazingly you suggested "% of runners" (for that group) somehow took account of field size. It doesn't.

Here is an example intended to demonstrate how Impact Values can draw the wrong conclusion.

Group A has a certain characteristic that Group B lacks.

Assume that an A and a B compete in races with field size 2.
And that 1 A and 9 ** run in races with field size 10.
There are 100 races with each of the field sizes.

Now if both A and B have the same chance of winning, then As should win 60 (=50+10) of the 200 races.

The 4 totals used to calculate the Impact Value are:

60 = A Wins
200 = Total Wins

200 = A runs
1200 = Total runs

The Impact Value is 30% / ~17%

= 1.80


This wrongly suggests that As should win 80% more than their fair share of races.

In fact both A and B have the same chance of winning, it's just As were lucky is running in a disproportionally high number of piddly races.

Winston, you claim to be very knowledgeable about Impact Values yet never once has it dawned on you that they can have a fatal flaw in disregarding field sizes.

Even after I told you about that flaw, you did not bother to check for yourself but instead clung on to your cherished belief of your self-assessment.

Mr jfc,
your contrived example makes no sense and is irrelevant to this discussion. we are not argue whether impact values good or bad. yet.
those 5548 runners ran in 6678 races. in those 6678 races there were 69234 runners. blind freddy and his dog can see that the 69234 runners is the total field size for those runners. and the average is 69234 / 6678 or approx 10.37
please how can the average field size be anything but 69234 / 6678?
there is old saying which goes "there are none so blind as those who will not see"

anyway, we get off the point. you offered your flawed statistics to try to justify your statement of "significance". nothing you do here has justified the use of that word. your premise that a horse has 3/N chance of running a place in a field size of N is flawed. as mr chrome prince (did not prince become "the artist formerly known as ..... "?) has rightfully pointed out this not consider the relative merit of each horse in each race. you have assign equal chance of running place to every single horse in the race. this is very dangerous assumption.
Thank you. Winston.

mr chrome prince.
you are correct that there is serious flaw with impact value calculation. you are right that it take no account of the actual chance of winning or placing for each horse. mr jfcs figures are equally flawed as he gives 1/N chance of winning a race to every horse in the race and 3/N chance of placing for every horse in the race. he knows this which is why he refuse to answer my question on this point.
Thank you. Winston.

Hi!
I don't want to buy into this argument on either side, but your assumption in this case about the field size being taken into calculation is just wrong.
True the average field size maybe 10.37 but it can be made up of countless variations in field sizes between 2 and 24. (unless you want to include a one horse race as well, they DO happen).

Good luck

Winston,

If my example makes no sense then why can't you identify which figures are nonsense.

My example makes perfect sense and exposes the flaw in Impact Values. And that happens to be the most important part of this discussion.

Once you understand the principles in that example you can then come up with more realistic scenarios where the Impact Value gives misleading results.

It's just that you've been sprung about your self-assessed expertise about this topic and have to resort to one more "value judgement" to try and wriggle out.



You are making the false assumption that every one of those 6678 races has to include at least 1 runner with the characteristics under consideration.

If you look at the other examples on that page you will see:


Note that there are fewer runners with the characteristic (=2,264) than the 3,127 races!

I remind you that my example include As and Beez (for the censor) in every race, but the IV still makes the wrong conclusion.

Now you, of all people, claim that this is getting off the point. Exactly what is the point of your sudden uninvited arrival here, where you pick up on insignificant matters. Then when I trouble to elaborate, you resort to misquoting me. For example there is a massive difference between my "more significant" phrase and your subsequent distortion into "significant" (the absolute).

mr jfc
please do not misrepresent my position. you berate others for doing this and now you do it youself. i have never said that impact values are not flawed. in fact on one occasion i actually state that i believe they are flawed and on another occasion i say that they are not as good as Nicks calculations. i think everyone here who read my statements will know this to be true.

you are trying to argue that your calculation are better than impact values. your calculation is probably better and for the sake of this argument i shall concede this point. it is after all irrelevant.

all you have done so far is throw up smokescreen to hide the fact that three times now you have ignored my assertion that your figures are flawed. they are fatally flawed. you argue all other points other than this becos you know that you cannot argue this point. to suggest every horse has 1/N chance of winning and 3/N chance of placing is just plain silly. your calculation are based on this premise and it is here that your calculation are flawed. ignore fact that impact value are flawed. ignore question of whether your calculation better than impact value. your calculation still based on flawed logic therefore calculation are flawed. even mr chrome prince agree with this.

now let us go back to original question i ask. you did not say "more significant" you said and i quote
"Horses who run 3rd <=0.5L are significantly better than 2nds > 3L."
you said "significantly better". i ask you to justify this and you post flawed figures. when i challenge you on your flawed figures you branch off into "insignificant matter" try to prove your figures better than impact values when the real issue is your figures are flawed.

i think this forum is open forum for all to discuss. i think discussion is good to try to understand nature of things.
why you not want me here mr jfc?
is it becos i do not blindly accept what you say as gospel truth and that i seek to find evidence of such truths?
i think anyone who blindly accept anything that is said on forum such as this is fool.
Thank you. Winston.

mr jfc
in post number 45 did you or did you not say
"Actually I thought there was a significant difference in favour of the one that ran 3rd." (my emphasis)
who is distorting what now?
Thank you. Winston.

Winston,

You are distorting now Winston, just as you did before.

Rather than refer to my actual quote, you found 2 similar ones of mine which you now want to labour.

The actual exchange in question was:





You quoted me then immediately distorted my "more significant" phrase into "significant" which has a distinctly different connotation.


As to the other material regarding "more significant", I already answered you.

For 2 samples of ~10,000. The wins/fair share indexes were 52.4% versus 31.5%.

I consider those figures significantly different.

If you don't agree with my conclusion that doesn't concern me. Feel free to bet that way.

mr jfc
you still have not disputed the fact that your figures are invalid. who cares what the numbers say. if the numbers are invalid then so are your conclusions.
thats 4.
Thank you. Winston.

winston,

Everyone who bothers to look back through my actual words as opposed to your distortions of them can verify that I have never claimed nor suggested that every horse in a field of N has a 1/N chance of winning.

In fact my precise words were:



An intelligent and knowledgeable person would know what I meant by that. That person would realise that this was one of many contexts where Expected does not mean Chance.

For a start here Expected = Chance is an absurd notion so therefore Expected is intended to have some other meaning than Chance.

Here the context was as in Actual versus Expected.

Now if I really believed that Expected was Chance why would I bother going to the trouble of counting Actual and computing A/E?

And why do I have to go to the trouble of demonstrating as I just have, that I did not suggest that Expected = Chance, when anyone who was concerned about that could have easily figured out my meaning for themselves?

Long ago I assumed that everyone else was well and truly bored with this discussion so I let your obvious misquotations and errors go, on the basis that most would see them.

But you insist on labouring your distortions, probably in the belief that if you drag your false accusations out long enough others will begin to believe you.

On the other hand I hope that more people will instead see you for what you really are.

Most words have a number of different meanings, so it is very easy going through others' work looking for something to misinterpret. If that fails, then misquote them, and start quoting the misquotes. But if even that backfires on you, merely change your identity again.

mr jfc
personally i do not think you can say that you expect horse to win 1/N without giving it 1/N chance. yes i paraphrase your words here but i think intelligent people will understand my point. in the context we talk "expect" and "chance" are related. i agree not exactly the same but related. however let us agree to disagree on this point. it is irrelevant to discussion.

you say that
this is the crux of the problem with your figures and why they are flawed. let us say we pick horse A that has 10 races and each race has field of 10. you calculate you expect it to win 1/10 for each race and therefore you expect it to have won 1 race by the end of those 10.

now. what if horse A is favorite in each race. your calculation still expect it to have won 1 race of the 10.
now. what if horse A is donkey and start at 200/1 in each race. your calculation still expect it to have won 1 race of the 10.
your calculation put Lonhro in 10 maiden races at taree and expect him to win only 1.

this is where your expected figures are flawed. using 1/N takes no account of ability of horse. extend this to a group of horses and the figure take no account of ability of that group of horses. your figure of 1/N whether you like it or no is just like a "participation rate" figure - very similar (but better calculated) than impact value.

if you really want to analyse "actual versus expected" then you need to calculate expected taking into account ability and actual chance of horse to win each race. this is what Nick do. when you do this then you will find some very interesting statistics.
Thank you. Winston.

Expected Win = 1/N
 

"Expected Win = 1/N"
Means exactly that!
In every race there is a winner!!
jfc!
While I disagree with some of the stuff you wrote in the past, in this case I think you are right, but I still would let it go if I were you, because by the looks of it you are pushing it uphill trying to convince your opponent.
Good luck.

Winston,

In recent Saturdays I receive e-mails with:

"results are now based on proportional staking according to the SP of each favuorite(sic)"


Pleasingly this outfit has now started to do what I've been doing for decades (even millennia).

Note I'm not suggesting anything untoward. Others have also independently settled on such a method, including that UK article on Impact Values.

And I often disclosed here that my ROT is calculated that way. Earlier I suggested you look back through my earlier posts for a discussion on that very topic. Incidentally I was trying to discuss that with an obvious acquaintance of yours.

So why are you trying to tell me something I evidently already know?

There is no flaw in my indicator because I don't use that for estimating probabilities.

It simply happens to be a more meaningful indicator than strike rate, because it usually copes with strike rate anomalies caused by field sizes.

I used it here late in the discussion merely to highlight what I consider a significant difference between 2 large samples.

But considering the abundant evidence I doubt whether any of this is really news to you or to any regular.

huh? how did we go from trying to show one group of horses is a better proposition than another to staking? i think this is more camoflage.

whether you use it for estimating probabilities or not is irrelevant. there is a flaw in your indicator. i can see it. mr chrome prince can see it. i think many intelligent people can see it. much more accurate indicator is easy to calculate. if you wish to continue to use flawed indicator to "highlight what i consider a significant difference" then so be it. it brings to mind famous quote by Mr Mark Twain

i go now. i can see there is no point to continue this discussion.