Stugots,

I ran those numbers and came up with a 10.1906% chance that the result is due to luck. This is an almost 90% chance that the result is due to an edge. These numbers are only relevant if the numbers you have provided are on real bets and not on a back fitted system. Back fitting will always tend to give a lower result.

Do you mind if I ask if this is betfair or tote prices ? If they are betfair prices have you taken into account commission at your level (2%-5%).

Another 6000 bets and a similar 1/price and actual winners would reduce it to a 2% chance that luck is the reason you are showing a profit.

Another way to analyse these figures is to check the standard deviation of the results lumped into 100 bets. You have enough selections there. You would have 60 data points and this should produce a nice distriubution. This is very useful in analysing your risk. If its not showing a nice bell curve (or close to one) then I would be worried and not betting the selections.

For others who do not have 6000 bets I would suggest using 20 or 50 bets. This can easily be plotted and you should see a bell curve as a result. You will be able to see your edge pretty quickly doing this as the peak of the curve should be to the right of 0.

Hope that gives a little bit of insight on one way to anaalyse a system.

There are another few tests I would consider still but I need access to your data to do these and thats not possible. But they include a number of advanced tests.

I think the information above on plotting the distribution is pure gold. How many people actuall start doing that is probably 1% of the people who read it. If you are the 1% then i wish you luck, by doing this you will learn more about your system/rating, probability and how to be a winner then you will ever imagine.

Dale,

I see what your saying and fully agree. Its an analysis tool and it requires data to be useful. Without enough data it will always bias towards chance playing a larger role then it should. Also if you input biased data into the calculation you end up with it leaning towards having an edge when realistically its the biased data influencing the results.

As always bad data in means bad data out.

I would query anyone on why they are betting (professionally) without enough selections for a proper analysis though. This is a mistake 99% of punters make.

thanks for your advice wesmip1, re your points raised -

- the figures provided are not backfitted, just the results of my of ratings. I have backed about 60% of them as I have use the raw ratings to create a couple of systems.

- the prices are betfair & applying commission @ 5% on all selections changes the sum figure to 1002.59.

- will have a go at the bell curve & upload a spreadsheet

just a question - at what point would you consider taking a selection method live based around the luck/edge percentages?

cheers

I usually wait for half a percent chance due to luck. This means that there is 1 in 200 chance that the system/rating is performing due to luck alone.

For instance yours was at 10% ... So there is still a 1 in 10 chance the whole thing could come crashing down in the next 6000 bets. Thats too high for me but might be acceptable to you.

attached a spreadsheet, had a go at the bell curve using the guide at - http://support.microsoft.com/kb/213930

interested in any thoughts

Attached Files

BELL CURVE.xlsm (84.1 KB, 406 views)

increased the random number sample & used the correct chart with this one

Attached Files

BELL CURVE.xlsm (254.3 KB, 368 views)

Looking at the chart not sure what you are plotting.

I will assume the blue line is the random sample (perfect bell curve).

What are the purple and green lines ?

I wasn't sure what you were plotting so a bit hard to work out. I normally plot the return as a percentage over X bets. So over 100 bets if it returns 110 then that is ploted as +0.1, and if it returned 90 that is plotted as -0.1

In th csae that you have done If I assume yours is the green line then it looks like a good curve but I can't make any other comments without knowing the type of data you plotted.

what i plotted was the sum of 1/price in 100 bet blocks

the attached chart is based on net returns on 100 bet block, seems now the peak is fractionally to the left of centre.

Attached Files

Book1.xlsx (183.4 KB, 352 views)

try to get it right one last time

Attached Files

Book1.xlsx (183.6 KB, 357 views)