This is the less discussed key to being a successful value bettor, and gambler in general. As already discussed, value betting is focused around identifying betting opportunities where the statistical likelihood of something occuring is greater than the implied likelihood in the odds offered. Inherent in this, is a condition that if you are going to take advantage of this, you need to be able to cope with bets that do not win.
Determining what your bankroll is is the easy part. It's the amount of money you have available that you're willing to lose. We do not condone adopting a value betting strategy with an imaginary bankroll. You need to have access to the money, and be willing to ride the emotions of the results.
The stake size appropriate for a given bet depends on the 2 key items you need to have an idea about in order to determine whether the bet has value in the first place: 1) the true likelihood of the event occurring, and 2) the odds available.
The Kelly criterion is frequently used by punters and investors to help identify staking sizes. Alternatively, these days, with the power available from simple applications such as Excel, you can easily create your own Monte Carlo simulations (as is the approach used by the Betting Bible Syndicate) to help determine stake size.
Kelly's formula is:
F = BP - Q
F = Percentage of current bankroll to bet (ie stake size)
B = the Odds offered in fractions (ie subtract 1 to decimal odds, so if something is 2 to 1, it is 3.00 in decimals, therefore B = 2)
P = Real probability of event occurring
Q = Real probability of even NOT occurring
In practice, the F derived from Kelly's criterion is halved or even divded by 4, when used by value bettors.
The key thing to remember when coming up with stake sizes is that you need to be able to weather a run of losses that is statiscally likely given the probabilities of the events you're betting on.
So as an extreme example, if you find an event that is priced at 100 to 1, but that you believe should be priced at 50 to 1, even though if you had a 100 £1 bets on this, you'd expect to win twice, and lose 98, therefore implying an edge of over 100%, you shouldn't be having a big proportion of your bankroll on an individual bet. In fact, if you use Kelly's criterion, it suggests a 1% stake. And as mentioned above, people often scale Kelly down.