What is a "z-value"? Is it different from a "z-score"? How is a
"z-value" calculated as it pertains to sports winning/losing records?
Here are some examples for some situations:(win-loss-tie)
10-0-0; z-value 3.14
4-0-0: z-value 1.41
22-6-0: z-value 3.02
54-23-2: z-value 3.53
37-10-4: z-value 3.94
source: snoopdata.com

I would like to know what's the formula in excel for z-value
calculation and could you please show me how I might use the built in
formula to calculate this? Please show at least one example explained
in as simple terms as is reasonable. Thank you.

---

Clarification of Question by pizzamik-ga on 15 Mar 2005 21:04 PST

Here are some details of situations discovered by the program: (each
game is evaluated based on historical results and looks for trends
sorted by z-value for the purpose of exploiting mathematically
incorrect betting lines. Since the book maker's primary interest is
achieving equal wagering on both sides and therefore capturing a
riskless profit of approx 5% on each dollar bet, the theory is sound -
bet on the math, not the line, which is designed to produce 50/50
wagering and not to predict the game outcome)
When:
1) Cleveland is playing, either home or away
2) Date range - all regular season games 1997 to 2003
3) Over/Under range from 177 to 199
4) Cleveland last game played was at home
5) Cleveland last game played went "under"
Then:
Bet the "over", which in this case was 187. The game went under and
the proposed bet lost. Z-value was 3.49 based on historical data
stating that all games that qualified under these conditions 49 went
over, 20 went under, and there were 3 tied games (bet is returned).

This example applies to the game played today, 3/15, Utah at Cleveland.

Here is another: (This applies to tonight's game, 3/15, Minnesota at Dallas)
When:
1)Dallas is playing at home
2)Date range is all regular season games from 1997 to 2003
3)The spread on Dallas' last game is -3.5 to +7.5
4)The opponent they are playing's last spread range is -5.5 to +7.5
Then: Bet "under". In this case, the over/under was 200.5, and the
final score total was 191, so the bet won. Z-value 2.92, with 10
losses, 28 wins, and no ties.

One more: Tonight's game Orlando at Sacramento (over/under 209.5)
When:
1) Orlando plays home or away
2) season range 1997 to 2003
3) Win to date % for the team Orlando is playing is from 30 to 60%
4) Orlando's opponent went "over" in their last game
Then: Play the "Over". This game is still being played as I write this
so we don't know who will win yet. It's the 3rd quarter with Sac ahead
65-47. Z-value 3.84 with 14 losses, 43 wins, 1 tie.

What I want to do is understand how the program calculates the
z-value. The program is written by Mike Orkin, Ph.D of statistics at
UC Berkeley who is a big wheel in the arena of Gaming.

You can download a copy of his program which is fully functional for
30 days for free to test this for yourself (snoopdata.com).

Let me know if you need more data. If the research is going to be more
extensive, I will offer more money for the solution.

Thanks for the help!

Hi,

As far as I know z-value and z-score is same thing.
Your data is not clear can u tell little bit more about the data.  I
can  not get the data from given source.

10-0-0; z-value 3.14
4-0-0: z-value 1.41
22-6-0: z-value 3.02
54-23-2: z-value 3.53
37-10-4: z-value 3.94
source: snoopdata.com

Tell in detail what you want to do.  May be z-value is website specific.
In statistics z-score or z-value whatever you call it is always
related to normal distribution.

z-value and z-score can be used interchangably in effect they mean the same thing.

Your other problem needs more information

here is an example with hypothesis testing...

suppose we have a bunch of data that is approximately normal and we
wish to test whether the mean (mu) is 0 or not... suppose our sample
mean (xbar) =0.5
and our sample standard deviation (s) is 2.5 with a sample size of 21

the we calculate the z score (or value) as

z = (xbar-mu)/(s/sqrt(n)) = (0.5-0)/(2.5/sqrt(21)) ~ 0.91 

now... this z is a z-value which you compare to a table... 

in order to do that i need to have a test... so... are you assuming
the teams will win x% lose y% and tie z%?  because if you do that then
basically you have a trinomial problem... and you can calculate
z-scores based on total points... but i need more info

So your betting over under on total scores?

Assume thats the case... then the line is mu

the actual value observed is (xbar)

and somehow hes calculating the variablility in the total scores from
the past data... that i cant say more without the program...

then he calculates the zvalue as

(xbar-mu)/sigma = (actual total score - line )/(historical variability)

if you had the code for the program i could tell you what the formula
is... but otherwise i hope this helps