what does a prediction formula look like?Prev TopicNext Topic

• hypersoniq

  

  Pennsylvania
  United States
  Member #1,340
  April 6, 2003
  2,581 Posts
  Offline
  Jan 19, 2006, 8:49 pm

  ▲▼

  this was just my spin on one component of neural nets (weighted inputs)...

  this is what I am looking for.. the concepts behind prediction in general,then I can figure out if (and how) theycould blend into excel or maybe one of the other office tools or programming languages I have available (access/C#.net/VB.net/C++/java)

  I'm first trying to nail down a concrete definition and series of steps for general prediction (doesn't need to be lottery, could come from stock analysis, risk management, weather prediction, image recognition etc...)

  I'm dropping all further development on all of my systems until I get the basics down.

  in a nutshell

  2-4-6-8-10-12-?

  I don't care yet about the 14, I am far more interested in the steps required to arrive at ANY answer.

  there are a great deal of posts where people advise to spot trends... they even throw in the occasional example but not one has ever been able to define a trend... to say, "this is what you need to have in your data to constitute a genuine trend"... or "this is a pattern, here is what you need to identify a pattern".... I think everyone would benefit from having such a resource, regardless of their system(s). As developers, we're all looking for answers but we're not asking the proper fundamental question(s)...yet.

   

  Have an EXCELlent day!
• paurths

  

  Switching between Fairfax, VA and Belgium
  Belgium
  Member #19,286
  July 29, 2005
  2,254 Posts
  Offline
  Jan 19, 2006, 8:54 pm

  ▲▼

  Without a fingerprint, one can not identify.
  Patterns all over the place, but only when they have been, never when they are to arrive... a fingerprint...
• hypersoniq

  

  Pennsylvania
  United States
  Member #1,340
  April 6, 2003
  2,581 Posts
  Offline
  Jan 19, 2006, 9:17 pm

  ▲▼

  a fingerprint after the fact... a profile beforehand...

  fingerprinting requires matching a number of points based on some geometric pattern-matching algorithm against pre-existing patterns on file

  profiling (or predicting) requires the use of a series of indicators (the more reliable the indicators, the more accurate the profile)

  question is, what is a reliable indicator? how many are needed?

  S-?-S-MT

  pattern? sure it is, the ?=P

  the pattern expanded would be Salt,Pepper,Sugar,Meat Tenderizer

  makes sense to me since that is the arrangement on my kitchen countertop... if someone else doesn't recognize it, is it any less of a pattern? Known patterns identify a fingerprint match, all fingerprints have points of interest, but if not on file... no match

  Have an EXCELlent day!
• time*treat

  

  United States
  Member #13,130
  March 30, 2005
  2,171 Posts
  Offline
  Jan 19, 2006, 10:11 pm

  ▲▼

  "Profiles" are based on indicators that have proven to be reliable in the past. Patterns that are not recognized by others are no less valid, only less recognized and therefore provide more opportunity for profit. Whether it's lotteries or stock markets (and there are many similarities), there are plenty of patterns occurring all at once. People see these and develop systems based on them.

  If people all use the same pattern/system, there is much less profit. If one person tries to use them all, there is much less profit for him. There is a saying in the stock market ~ "Bulls get rich & bears get rich, but pigs get slaughtered."

  As far as how many reliable indicators are needed, there is no "magic" number. And too many (filters) can be worse than too few. Bleach, ammonia, and lemon juice are great for cleaning. They should not be mixed in any combination. Pick one.
  

  In neo-conned Amerika, bank robs you.
  Alcohol, Tobacco, and Firearms should be the name of a convenience store, not a govnoment agency.
• RJOh

  

  100

  mid-Ohio
  United States
  Member #9
  March 24, 2001
  20,272 Posts
  Offline
  Jan 20, 2006, 11:36 am

  ▲▼

  Seems like the general answer to the question "What does a prediction formula look like?" is "I can't describe it, but I'll know it if I ever see it."  Good luck on your search.

   * you don't need to buy every combination, just the winning ones * 

         
  
• powerplayer

  

  United States
  Member #17,833
  June 28, 2005
  2,087 Posts
  Offline
  Jan 20, 2006, 12:04 pm

  ▲▼
  Quote: Originally posted by hypersoniq on January 15, 2006

  back to the core of development, what is a prediction algorithm? what does it look like? what are the steps needed to look at a list of numbers and try to extrapolate the next draw?

  there must be something outside of the realm of lottery that can be tried...

  links to sourcecode/math texts appreciated

  I'm on the look out for one also there buddy.

  Here I found this today:

• You can create a formula that generates a pseudo-random number. When designing the formula, the idea is for it to produce a string of numbers that would look random to anyone who did not know what the formula is. Characteristics of a good formula include:
  • No repetition: The sequence does not cycle around and repeat itself.
  • Good numeric distribution: If the formula is producing random numbers between 0 and 9, the number of zeros, ones, twos, etc. that it produces should be roughly equal over a long period of time.
  • Lack of predictability: You have no way to predict what the next number will be unless you know the formula and the seed (the initial value).

int rand()  {   random_seed = random_seed * 1103515245 +12345;    return (unsigned int)(random_seed / 65536) % 32768;  } 

it's in C prgramming 

This formula assumes the existence of a variable called random_seed, which is initially set to some number. The random_seed variable is multiplied by 1,103,515,245 and then 12,345 gets added to the product; random_seed is then replaced by this new value. This is actually a pretty good pseudo-random number generator. It has a good distribution and it is non-repeating. If you use it to produce random numbers between 0 and 9, here are the first 20 values that it produces if the seed is 10: 
4 4 6 0 7 4 2 3 5 0 5 6 6 4 5 6 7 6 7 4 
 
If you have it produce 10,000 values between 0 and 9, here's the distribution: 
0 - 1015 1 - 1024 2 - 1048 3 - 996 4 - 988 5 - 1001 6 - 996 7 - 1006 8 - 965 9 - 961 
 
Any pseudo-random number formula depends on the seed value to start the sequence. If you start with the same seed, you will get the same sequence of values from the formula. So if you give the rand() function shown above the seed of 10 on one computer and look at the stream of numbers it produces, it will be identical to the stream of numbers produced on any computer that runs it with a seed of 10. In the case of the GPS this reproducibility is used as a way to give each satellite a predictable but different pattern of values that the GPS receiver can track. 
To create a random and unpredictable sequence, the seed must be a truly random number. To get this truly random number for the seed, most programs use the current date and time, converted to an integer value (for example, converted to the number of seconds that have elapsed since January 1, 1970). Since this is a different number every time you start the program, it makes a good seed. 
 
 
The code is on the site. PM me if anyone wants it.
 
Hope this helps
 
PP