How do you know a system works?

I've been testing a new system of mine for the last two weeks. How do I know if I've just been lucky or if there is some merit in the system. This maybe a question for the maths and stats geniuses:S

test it for longer than 2 weeks!

JustinV Wrote:

---

> test it for longer than 2 weeks!


Wise words

If it works for a life time you could be on a winner but then a bit late to pass on

Lol

When testing new systems i test them over a 3month period....than judge afterwards....but each month i expect to see so kind of growth....

When you withdraw instead of deposit.

spot on JD

Play the system, don't dabble on the side -

Don't share your system, remain a winner

I was hoping to get answers to some of the following questions....

What should be a big enough sample size?
What minimum average profit per bet should I get to eradicate the effect of one lucky longshots effect of the stats?
What longest loosing run is acceptable?
What minimum strike rate is acceptable?
Should I include or exclude outliers?

Lucky_Dude Wrote:

---

> I was hoping to get answers to some of the
> following questions....
>
> What should be a big enough sample size?
> What minimum average profit per bet should I get
> to eradicate the effect of one lucky longshots
> effect of the stats?
> What longest loosing run is acceptable?
> What minimum strike rate is acceptable?
> Should I include or exclude outliers?

1 whole month aught to give you a really good snap shot
Play day by day, one day total losses can accumulate alot of money, if in doubt play with your head not heart, bail if the kitchen is too hot
24/5% win rate, i would not gauge on less than that
Outliers should be ignored, they not indicative of your system, like any stats you need to ignore the occasional flier.

Hope this help LD

I hope your system does not consist of following Zietsman Oosthuizen.

Lucky_Dude Wrote:

---

> I was hoping to get answers to some of the
> following questions....
>
> What should be a big enough sample size?
> What minimum average profit per bet should I get
> to eradicate the effect of one lucky longshots
> effect of the stats?
> What longest loosing run is acceptable?
> What minimum strike rate is acceptable?
> Should I include or exclude outliers?

My opinion, if you have a basic knowledge of statistics:

Your minimum sample size must be that size when all of the mean (average), median and the mode approximate each other. (I'm referring to the average of whatever variable you are testing).

Once you have a sample size, instead of using only averages use an average plus one standard deviation.

IMO, a Return on Investment for your punts is a better key performance indicator than a Strike Rate. The ROI must be at least better than your personal interest rate to your creditors.

Include the outliers - your correct sample size and the standard deviation accommodate it.

One hint - rather than playing R1 for all bets irrespective of the price make the stake in proportion to its price ie a set win amount divided by its price. This is your risk mitigation.

Another hint - published SP's bear no relation to reality. Increase it by your own demonstrated average.