Betting Experts Answer: At What Point Has Variance Played Out in Betting? | Part 1

Welcome to the 9th question in our series: “Betting Experts Answer The Industries Biggest Questions.”

This week we are answering a question that is regularly asked by Trademate Sports users and value bettors around the world, as we asked nine betting experts:

Q9: Assuming you have an edge, at what point can you start accurately evaluating your results and say that variance has played out? Is there an amount of turnover or number of bets?

Check out Part 2 of our answers here, where we get the opinions of former odds compiler - Matthew Trenhaile, betting analyst’s Mark O’Haire & Joseph Buchdahl and current odds compiler - Jeevan Jeyaratnam.


The generic answer is the more bets you have to analyse the more indicative they are of your true edge. The short answer is about 2000. The Long answer involves a little math. The magic formula is:

N = Z^2/4E^2

N is the required number of bets. Z is your z-score for a one sided confidence interval. It basically tells you within certain confidence that your data point will be within Z standard deviations of the mean. We use a one sided instead of a two sided z score as only a lower limit is required (Thanks to Professor Harry Crane for correcting me on this).

spanky's workspace

For 95% confidence z-score would be 1.65 99% confidence z-score would be 2.33. 99.999% confidence z-score would be 3.09. These values can be found in the back of most intro statistics books. Finally E is your standard error. This tells you how much off you are willing to accept the observed win rate to be.

Quick example. How many bets are required to be 99% confident that your observed win rate of 55% would be profitable. Before we plug and chug into the formula, we need to set a value for E. I like to use 2.5%. Why? Because betting at -110 the breakeven rate is 52.38%. If I choose 2.5%, then I create a lower bound of 52.5% (55%-2.5%) still resulting in a profit. Let’s plug and chug into the formula:

N = Z^2/4E^2

N = 2.33^2/(4*.025^2) = 2172

If you used a 95% confidence interval (a z-score of 1.65), the result would be 1089.

If you’re hitting 56% or 57% the number of required bets would be less since you can be more lenient with your standard error. I’d stick with 99% confidence (2172 bets). Don’t wanna over gamble on your gambling estimates. Most important disaster to avoid is thinking you have an edge when you really don’t.

Follow@Spanky on Twitter & check out the article The Ringer wrote about him here.


I'm a handicapper, not a line grinder, so I'll speak for us: there is no magic number. As your sample of bets increases, the relevance of the earliest bets decreases, because the markets are always changing. You never, ever know for sure than you haven't just been lucky. The info comes on a continuum, without absolutes. The larger the sample and success rate, the greater the likelihood of edge, but the answer is never definite. There is no X number of bets and Y rate of success equals Z (real) edge. The bigger sample, the higher the win rate, the greater the likelihood of edge. This is not a biz you can ever go to sleep on; you can always lose your edge, because the market can always pass you by.

a guide to modern sports betting

Follow @TruePokerJoe on Twitter & check out his book ‘Sharper: A Guide to Modern Sports Betting’.


Again this very much varies depending on the profile of bets you are taking on and there is no 'one-size-fits-all' solution here like most things with betting. You need to dig into the data to get the answers.

For example if you are betting at average odds of say 40/1, a sample of 500 bets would be inadequate to make any judgement calls. There is just too much opportunity for luck or variance to play a part in such a small sample at big odds. Yet if you were betting at odds of 2.0 (evens in fractional prices) and had a 500 sample size, that would be more informative, although certainly not conclusive.

The best approach is to calculate your p-value rating, which is a test to establish the likelihood that a series of bets were achieved through luck or chance. The closer your score comes to a p-value of 0.0 then this is an indication that the results were obtained by skill alone and not luck. At this point the variance will have played out.

For example, one tipster we recently evaluated had a p-value score of 0.01 - yet they have a data set of just under 4000 bets with a 10.10% ROI so we are dealing with a large sample size here. Another tipster with a much smaller record scored higher with 0.313 in the p-value test, which represented a 1 in 3.2 chance that their results were obtained by luck alone. In this case, whilst the tipster in question has impressive pedigree and record to date, the sample size of data we had for them was not the largest.

Follow @SBCinfo on Twitter & check out the Smart Betting Club website here.


Obviously the more data you have the better. But it is also dependent on the average odds that you are betting at and also your staking strategy. I like to use a p-value estimation calculator that I downloaded from Joseph Buchdahl's Football data site to give me an idea of how likely results could have been achieved by pure luck.

There are also other mathematical models that you can use. But I imagine you will be getting answers from the likes of Joseph Buchdahl and Pete Ling, who are much experienced and knowledgeable about these models then me. So I will let those guys do a better job of explaining them lol.

Follow @SmSportstrader on Twitter & check out the Smart Sports Trader blog here.


It’s a bit of both. If your bets have very different stakes it is surely important to consider that when analysing your betting record. As a rule of thumb, if you have positive closing line value, a few hundred bets would be enough. If you are looking at ROI alone, it would be more like a few thousand.

Follow @ChurchOfBetting on Twitter & check out The Church of Betting website here.

Check out Part 2 of our answers here, where we get the opinions of former odds compiler - Matthew Trenhaile, betting analyst’s Mark O’Haire & Joseph Buchdahl and current odds compiler - Jeevan Jeyaratnam.

Love getting the opinions from experts in the betting industry? Then subscribe to the Trademate Sports Podcast, where we interview the most important people in the sports betting industry.

Here are the other questions we have got our industry experts to answer so far:

  • Q1: Top 3 tips for betting beginners? Part 1 & Part 2.
  • Q2: How do you define “finding value” in betting markets? Part 1 & Part 2.
  • Q3: How do you determine whether your betting results are based on luck or skill? Part 1, Part 2 & Part 3.
  • Q4: Kelly criterion or flat staking: Which stake sizing strategy do you consider to be the best and why? Part 1, Part 2 & Part 3.
  • Q5: What is the best method to use to make money from sports betting? Part 1 & Part 2.
  • Q6: How difficult is it to beat the sports betting markets? How efficient are the odds? Part 1, Part 2 & Part 3.
  • Q7: How to manage risk when betting? Part 1 & Part 2.
  • Q8: Is there a best sport to bet on? If so, what is it? Part 1, Part 2 & Part 3.
  • Q10: What is the one thing you would like to see change in the gambling industry? Part 1 & Part 2.
  • Q11: Do you think emotions play a part in people's sports betting results? If so, how should they overcome this? Part 1 & Part 2.
  • Q12: What are the top 3 mistakes people make when betting? Part 1, Part 2 & Part 3.

Get The Best Articles and News To Your Inbox

Subscribe to our newsletter and stay updated.