Significance of Sample

The easiest way to test any system is put real money down.

One of the differences between paper trading and real money is how you cope with the inevitable losing run. At these odds you must expect an enormous losing run. Will you be prepared to keep going?

This may be back fitting. Are you coming up with your picks / bets before the matches take place ?

put your bets up on the thread and give it a dry run.

The bets have had average odds of 68.04.

There have been 58 winning bets at average odds of 27.85.

At first glance it looks like you could improve your strike-rate and reduce losing runs an awful lot by ignoring the bigger priced outsiders

And if your profit is down to 2-3 of the very big outsiders coming up it could be that the system is not actually worthy and that within your next 1000 bets those 2-3 big odds win will not happen.

At this stage the sample seems too small. Whichever, good luck with it

Shezter , i strongly suggest you go to at least 100 winning bets before making any 'conclusions' -
the bigger the sample size the more accurate your conclusions will be.
GL.

Yeah, the bigger the sample the better. 700 is probably pretty reliable but you'll never be able to guarentee a system will work from past results. If you wanna do a hypothesis test using binomial, poisson or chi squared this will tell you what the chance is of your system being successful by luck. Also, the guys above are right, if your success is down to a couple of big odds winners then this is less reliable as they are harder to predict. If you are willing to go ages without a winner then try it out but I would suggest you might be better off leaving the bigger odds out and improving your strike rate but you would have to examine your data first. Good luck, it sounds like you could have something good there

http://www.surveysystem.com/sscalc.htm will help you i suspect (Just a sample size calculator to do with opinion polls or something but it works for gambling.)

Turn your results odds into percentages (so if you've won 20 times in 200 races, it's 10% winners) put the number of bets into sample size on the second calculator, set confidence level to 99%.
The calculator returns the confidence interval.

The confidence interval is the plus-or-minus figure usually reported in newspaper or television opinion poll results. For example, if you use a confidence interval of 4 and 47% percent of your sample picks an answer you can be "sure" that if you had asked the question of the entire relevant population between 43% (47-4) and 51% (47+4) would have picked that answer.

That will tell you what % of bets can be winning. Compare that to the odds and take a view from there. The further from 50% you are (and you're miles away) the more silly these numbers will appear. They won't be wrong though, just annoyingly useless until you put in a big sample size.

You need to be careful with sample size calculators because there are several different types according to how the sample is derived.

Its questionable whether this is a sample anyway because you can only sample something that currently exists.

What you have here is actual performance data not really a sample at all. The extent to which it is represtentative of future performance can only be measured with new data.

This is why most shrewd heads will tell you to carry on until you start losing and then take a view.

Many Thanks for your replies to the initial question.

Regarding leaing out the large priced outsiders, how would i determine what price I start to leave out the bets?

I understand that the profit margin might be town to a couple of big priced winners but there has only been 2 with odds of 100,

Fulham 3 - 0 Man United (Odds 170) and Middleboro 0 - 3 Blackpool (Odds 130)

These obviously to level stakes increases the variance on profit. Level stakes isn't what I'm hoping to use though, I have been using the Kelly Staking Method in order to decide on stake size in relation to how much 'value' I believe a price has.
I just posted up the level stakes results because from what I have read, no staking system can turn a level stakes loss into a profit.

So, would you still start leaving out the large prices even though they are considered to be value, and if so, at what price do you start leaving them out?

FatalStatistics, I have been keeping up with your Correct Score thread even though I havent posted on it, are you just concentrating on level stakes or have you tried other staking methods? If so, which have you had most success with?

Thanks again in advance of any replies.

Shezter

You're actually 300 points up, so you're exactly breaking even on everything else without those 2 bets, which isnt too bad at all .
Someone advised to press on and thats fair enough just dont count on a 24% return. Good luck!

PS - So has Kelly enhanced your returns in reality? If it's worked for you and is based on statistical reasoning for the 'true' chance there's nothing that wrong with it. It's other 'stepping-stone' or doubling-up methods that are invalid.

Kelly hasn't enhanced ROI, it has given about a 16% ROI, which I believe is better than level stakes as the high odds winners don't affect the variance as much because a lot less is staked on a 100/1 shot rather than a 3/1 shot.
This is why I was asking if the longer odds should still be eliminated even if they are perceived to be value?

Thanks

without knowing exactly what you are doing it may be difficult to get a proper assessment but if, as shown in your example of the two big odds wins you are betting Correct Score, then unless the odds on offer are too high (which they may be) you may have had more than your share of good fortune in your sample and may well get a poor run later.

The average odds you quoted are quite large and someone else pointed out they felt a losing run may be out there for you.

Hope it works for you, but from experience thousands of bets doesn't always give a proper picture of what you have.

Backing at big double and treble figures requires not 700 matches, but more like 7000 before you can draw conclusions IMO. I would need 700 for evens chances.

Your expected number of winners at those avrage odds is 18 which gives you an enourmous ARCHIE score of around 90. Get your betting boots on.

Shezter , i bet only level stakes = 0.25% of bank ,therefore increasing as bank increases.
I would not consider any 'staking plans' until you are 110% sure your method works at level stakes. GL.

Have I got this right?

An average of just over 2 bets per match and just 2 of the 58 winning bets account for almost all of the profit?

Wouldn't get those betting boots on just yet.