Quote:
|
Originally Posted by ubetido
Could someone explain what this ARCHIE calc is and how do you apply it?
|
From a posting of mine on another forum.
===
The Punters Chi-Square Test.
Quote:
Originally Posted by Whisperin 
What is a Chi Test? I've seen this expression before but do not know what it measures
|
In simple terms it's a test used by statisticians to determine the probability that a set of observed results were obtained by chance. I'm not a statistician, and simple is always good for me
For punting it can be used like this. I'll use Peanuts numbers for the example.
1) Determine the OBSERVED number of winners and losers. In this example Peanuts tells us there has been 9 winners from 49 selections, therefore 9 winners and 40 losers.
2) Determine the EXPECTED number of winners and losers. This is a bit more complicated, and I'm assuming that the SP is an accurate indication of the selections chance. Peanuts tells us that the average divi is $2.99. I'm assuming this is for all selections, not just the winners. We need to convert this to a percentage.
1/2.99 = 0.3344
Next we need to "normalise" that percentage to a 100% market. I'm guessing that the original market was around 115%.
0.3344/1.15 = 0.2907
There where 49 selections so we multiply 49 by 0.2907 to give us the EXPECTED number of WINNERS
49*0.2907 = 14.24 winners
From there we simply subtract 14.24 from 49 to give us the EXPECTED LOSERS.
49-14.24 = 34.76 losers
Ok. So now we have the following
OBSERVED WINNERS = 9
OBSERVED LOSERS = 40
EXPECTED WINNERS = 14.24
EXPECTED LOSERS = 34.76
3) Let's put these figures into Excel and use the built in CHITEST function to work out whether the tipster is any good. Enter the following into the listed cells without the quotes.
Cell A1: "9"
Cell B1: "40"
Cell A2: "14.24"
Cell B2: "34.76"
4) In a blank cell enter the following.
=CHITEST(A1:B1,A2:B2)
You should get 0.099215 returned. This is the likelihood expressed as a percentage (9.92%) of the observed results being due to luck. From this we can determine that if the selections are losing, and there is a greater than 90% chance of this being due to something else besides luck, then it's probably time to get a new tipster.
This is the quick and dirty method, but is good enough for punting purposes. The result will jump around alot until you get a fair few observed results in there.