• Hi Guest Just in case you were not aware I wanted to highlight that you can now get a 20% discount on Inform Racing.
    Simply enter the coupon code ukbettingform when subscribing here.
    We have a lot of members who are existing users of Inform Racing so help is always available if needed.
    Best Wishes
    AR
  • Hi Guest Just in case you were not aware I wanted to highlight that you can now get a free 7 day trial of Horseracebase here.
    We have a lot of members who are existing users of Horseracebase so help is always available if needed, as well as dedicated section of the fourm here.
    Best Wishes
    AR

anyone remember ?

  • Thread starter I wanted to be deleted
  • Start date
A question for you guys,

as you might know, i am experimenting with Genetic programming/Alg's to find better ways to look at data ( as vdw himself advised )

basically, you can think of a genetic algorithm as population of "creatures", each creature is essentially a phenotype, which has been produced
by a specific Genotype, consisting of genomes, which in turn consist of individual genes.

Now i realise that to some, the above will illicit the reaction of "WTF does all that mean"... bare with me ;-)

to simplify, think of each Genotype as a set of genomes that produce a Phenotype which to all intents & purposes
(in simple terms) you can think of as rules of a system.

eg :- one genotype may produce a "creature" that simply says "if the horse is FAV-today...AND the horses PM rating LTO is greater than the PM Rating 2nd-LTO.....then BET"

i must stress, my programming all of this, is not an attempt to "systemize" vdw's methods, but rather to dissect the various aspects of "data" and find which is
better than which.

As a result of all of this, each creature is evaluated against a set of test race results to see how well (or badly) it performs.

Each Creatures is then given what is called a fitness score based on how well it performed.

Here is the problem, this fitness evaluation of each creature, has to take into account the following...

win %
Profit(or loss)
Profit as a % of total bets placed
Number of bets placed

using the above 4 factors, i ask you guys, what would YOUR ideal results actually be?

an example, we could go all silly, and simply say..... oh.....100% winners, with 3 million % profit.....jobs a good 'en !

but lets be realistic,

obviously, it would be no good having 1 winner from 1 bet placed @ 5/1 therefore 500% profit, hence my including Number of bets in the above 4 factors.

So... in simple terms, what would your ideal set of results be, in terms of win %, profit, profit %, number of bets (per week/ per month?)

Cheers.
 
I would think that you would be looking at a Bayes type prediction of probability and then backing when the odds on Offer are greater than the model predicts.

Statistics from the 2016-17 Season

Bayes Theorem

The Final Link will give you a System with the number of bets, Strike Rate and ROI from the Hoof Computer Model and perhaps give you a target to aim for if using a AI Strategy
System B
 
As the stats indicate there Chesham non handicap races as a rule of thumb provide the best returns. There is no surprise there.

Racing can be extremely simple or can be made very complicated. I understand some people like to experiment with different ideas and there is nothing wrong with that. Rave you appear to be going down the complicated route,but if you get enjoyment from it, then that is great.

Lee I believe is a busy man, with his own business,who worked from lists, which is easier.

I myself work in a similar way ie from lists, the last race at wolves on saturday provided the obvious winner, and the runner up Ocean of Love has gone on to my list.

Good quote by the way Chesham.
 
alright,

which of the following two would you guys prefer,

40-100 (40% win).... with +20 profit (20%) and obviously 100 bets

28-100(28% win)...with +27 profit (27%) and again 100 bets

cast your votes
 
I agree Ark,

so would you say then, that on a % basis, profit % has greater "stature" so to speak, than win%?

How about the number of bets (over a given period).. which of the folliwng would you prefer

28-100, +27pts (in say, 3 months)

14-50, +13.5pts (in the same 3 months)
 
Profit is surely the bottom line , rather than 'stature' ?
The win % will obviously influence the unit stake in order to cater for expected losing runs.

As regards the number of bets over a particular time then for both monitoring purposes and profit there again it seems obvious that the more bets the better.

Or am I missing something here ?
 
I think you'll struggle to find anyone who would rather win less money overall! Why would you? Having a good win strike rate is a bonus, maximising profits is the goal.
In my opinion.
 
If you have a Win Probabilty = 50% you can expect a losing run of 10 after 2048 bets

If your win probability is 30% you can expect a losing run of 10 after 118 bets


If your win Probbility is 20% you can expect a losing run of 10 every 47 bets and a losing run of 15 every 92 bets

It is all down to ones betting psychology. If your sole income is from betting then being able to withstand the losing runs from a psychological perspective is as important as the ability to select winners. Alan Potts stated that he did not need to find anymore winning bets than he already did at that time, but to back less losing bets.
 
Last edited:
which of the folliwng would you prefer
28-100, +27pts (in say, 3 months)
14-50, +13.5pts (in the same 3 months)

Like the others I am struggling to see why you are asking these questions?
There is surely only one answer to the question asked. Who would choose to have less profit ?
If you have an edge then you would want to exploit it as often as possible. As dave58 dave58 said if , on average, you are making 27% profit on every bet then the more bets you have the better.

Like Dave I feel we are missing something from what you are trying to find out:confused:
 
Ark ( & others).. i appreciate your comments & questions,
basically, in order code the genetic algorithm, it requires that the fitness function (the valuation structure of each phenotype) is coded correctly,

or at least, the more correctly the better.

in this, i see 3 parts to how well (or badly) a "system" (and i again repeat, i am not attempting to systemize vdw, simpy using the rules of a system
as a nice simple example of a phenotype) performs, in terms of, win %, profit(%), number of bets,
basically, there are 3 variables there...

1) win%
2) profit (as a %?)
3) number of bets


these 3 variables(values there of ) need to be combined into a single "value" in order to determine that
"this system" performed better than "that system"....... hence........ combining those 3 factors, is not simply a case of adding up 3 values,

they need weighting, ie.....which is more important........win % or profit %?......obviuosly......profit seems to be the one to choose.......but......having said that,
suppose we have a system/methoid that produces 10,000% profit !!........fantastic !!........but if it only produces 1.5% win %....err......... whos going to back those then?

so....... the reason for asking these questions ( simplistic as they may seem ) is to try and establish the "relative" importance
of those 3 factors, win %, profit(%?), number of bets... so that i can code them into a singlew value, in order to score or evaluate
each phenotype in order to be able to select the top 10% or 12 % .....or whatever %.....so that these top few...can be used to "mate" in order
to produce the next "generation"

hopefully, that makes sense

phew
 
What AI software are you using @rave154 . A friend of mine used Tiberius on the Ability Ratings combined with other data to come up with an algorithm
 
No sorry, you are asking questions to which there is no definitive answer, I think that's where the problem lies.
Nobody can put empirical figures to the 'relative importance' of the factors mentioned. You could ask 100 gamblers for their opinions and get 100 different answers.
Could you maybe aim for the objective rather than the subjective, then you might be able to get some useful information rather than just opinions ?
In the real world, I think most here work out profitable ways of betting first, and then decide whether the profit:strike rate ratio suits our own betting profile.
 
Last edited:
Consider using A/E, Chi and ROI as indicators of how well a system performs.
Strike rate is useless as a measurement of "goodness" you could have a 50% S/R but if all of the selections are odds on you will not make any money.
 
Last edited:
The Tiberius Model would output a positive or a negative. (Strength Column)The highest Positive Number best for backing and the lowest best for laying

Example

CCE7B2B7-030A-4D92-A791-A96BFAD4FCF0.jpeg

E3483D58-81BB-49E6-9DFF-61AF4D771F22.jpeg 8BCCD31E-3D0A-41A5-85DB-BBC4A5478CE4.jpeg
 
Chesham, ncLad, dave & Ark,

thanks for the comments :) stuff for me to ponder :)

Chesham, the software i am using....is handwritten by myself, it is however not AI ( in the strictest sense)
rather it mimicks the process of Evolution in the natural world.......in the software world.

Just as you have "animals" in the natural world, competing for food, resources, females !....etc.....in my
software world you can think of the animals as Racing systems ( to put it very simply ) and
they are being measured against each other not in terms of how much food they aquired or how many females
they mated with, but rather.....how much profit they made, what % of winners, how many bets.
Once this is measured for all the "animals"... the top 10% ( or so ) are selected for mating to produce
another generation and the procedure repeats.

Hope this makes sense
 
while pondering all your comments, i have decided for the time being, to use the following way of combining
win % and profit % into a single number in order to rate/rank the successfullnes of each "animal"

consider 2 "animals" producing

A) 20% winners, 40% profit

B) 50% winners, 10% profit

same number of bets in both cases,

now, both "animals" ( systems, phrasing it in a simple way ) have performed well,

however, it is undeniable, that A) produced 4 x as much profit as B)

however, if we simply add win% to profit % for both cases, we get...

A) 60 (%)

B) 60 (%)

ie..both seem to be equal, therefore i suggest that i will "weight" the profit (or loss) figure by a factor of 2 to start with,

thus it would give

A) 20 + (40 x 2) = 100

B) 50 + (10 x 2) = 70

thus now A) shows up as "better" than B)

we will see how it goes, and i am still very much open to comments etc

Thanks again
 
To be honest I think you are wasting time using S/R and Profit % as means of comparing system performance.
The number of bets is far more important than the S/R.
The S/R for a system that has produced 3,000 selections is for example considerably more reliable than the S/R of a system that has produced 100 bets.

Can I ask why you have dismissed A/E and Chi as a means of measuring performance?

Are these 2 systems equal?
A) 20 + (25 x 2) = 70

B) 50 + (10 x 2) = 70
 
Back
Top