luck vs chance calcs for b. system

[liero1]

Hey!

Does anyone use the luck versus chance calculations for any back testing? You can download the template here http://www.football-data.co.uk/blog/luck_skill_sports_betting.php Also a good read!

I find the results often very eye opening - if a system looks good just on ROI and profit, but the p-value is high and there's a chance of 1 in 20 that is was luck.. well, I bin the system!

This below system looks promising in that aspect, one of the best so far .. what do you think?

Edited October 24, 2018 by liero1

[chrisgas]

It would appear from the data you provide that you have a winning system so I hope it does go well for you. The number of bets needed to achieve 1% statistical significance level (p-value = 0.01) is only 83 and you already have placed 276 bets so the system appears to be validated.

I currently use Excel 2010 so if you have this or a later version you may like to add the following lines to the down loaded p-value-calculator:

Cell A9   'Number of bets required for significance = 

Cell B9   Type any positive number eg 500

Cell A10 'T.Dist.RT =

Cell B10 =T.DIST.RT(SQRT(B9)*B2/B4,B9 -1)

Now go to Data, What if Analysis, Goal Seek. This will highlight an input table as follows:

SET CELL to B10

TO VALUE to 0.0099

BY CHANGING CELL to B9

The resulting value in B9 will be the number of bets required for significance for any yield/betting odds pairing (for horse racing usually in the 1000's).

You may also be interested in another statistical method for evaluating systems using the Chi Squared statistic. If you wish Google "Archie a method of evaluating systems". You will find this on the roulette forum although, it was originally wrote for horse racing systems. 

Anyway, good luck with the above system.

 

[liero1]

On 5.11.2018 at 7:42 PM, chrisgas said:

It would appear from the data you provide that you have a winning system so I hope it does go well for you. The number of bets needed to achieve 1% statistical significance level (p-value = 0.01) is only 83 and you already have placed 276 bets so the system appears to be validated.

I currently use Excel 2010 so if you have this or a later version you may like to add the following lines to the down loaded p-value-calculator:

Cell A9   'Number of bets required for significance = 

Cell B9   Type any positive number eg 500

Cell A10 'T.Dist.RT =

Cell B10 =T.DIST.RT(SQRT(B9)*B2/B4,B9 -1)

Now go to Data, What if Analysis, Goal Seek. This will highlight an input table as follows:

SET CELL to B10

TO VALUE to 0.0099

BY CHANGING CELL to B9

The resulting value in B9 will be the number of bets required for significance for any yield/betting odds pairing (for horse racing usually in the 1000's).

You may also be interested in another statistical method for evaluating systems using the Chi Squared statistic. If you wish Google "Archie a method of evaluating systems". You will find this on the roulette forum although, it was originally wrote for horse racing systems. 

Anyway, good luck with the above system.

 

thanks so much, much appreciated will add all of that!! cheers!

[liero1]

Hey @chrisgas not sure if you still follow this thread. however - I struggle using the p-value test for laying systems.. I think I did a mistake above and used the ROI as return / staked rather than return / liabilities. If I use liabilities the 1-in-x value gets really low though.. I am currently working on the below system. And I just CAN'T believe this is a 1-in-5 chance?! Am I wrong using return/liab.. If I would use return/stake it's a 0.16 ROI and 1-in-x would go up to 930,000 ... 

 

 

Oh and thanks for the goal seek formula, works like a charm

Thanks

liero1

Edited November 19, 2018 by liero1

[chrisgas]

Litero1 if I have read your last post correctly then a p-value of 0.18 does not reject the null hypothesis which would suggest the system is based on luck. To reject the null hypothesis you would need a p-value <=0.05 (95% confidence) or a p-value <=0.01 (99% confidence). This would then suggest the results are derived from something other than luck.

The standard deviation calculation in the p-calculator depends on a relationship between the yield and the odds. Strictly speaking this will only hold true if for each bet the stake is the same and the odds are the same although, it wont be that far out for small adjustments to either. However, for lay betting you need to discard this standard deviation calculation as your lay stake/liability may be variable.

To overcome the above problem with lay betting I suggest you set up a column in your spread sheet headed 'profit and loss'. Then, for each bet enter either the positive return (less stake) if you win or the negative liability if you lost. Once all your bets are recorded you need to find the average (x-bar) and the sample standard deviation (s) of the column. With these values you can then use the following formula to calculate your t-value:

t-value = SQRT(number of bets in column)*(x-bar - 0)/(s)

and finally calculate your T-Dist function. Remember the degrees of freedom is the number of bets in the column -1. Hope this helps resolve your problem with lay betting.