You already know what a value bet is: a price above the fair price, where fair is what is left once the margin is taken out across the market. This page starts one step later, at the part that decides whether an edge on a screen ever reaches a bookmaker account. If the theory is not settled yet, read value betting and positive EV betting first, because none of it is repeated here.
From here the fair price is a given. The question is what has to be true before software can act on it, and why the honest answer is that the arithmetic is the easy part.
The four steps, and what breaks at each
Every automated value bet is the same four steps, whether a person runs them by hand or a session runs them to rules.
| Step | What it does | What goes wrong |
|---|---|---|
| 1. Build a fair price | Take the margin out across the market to get a fair chance for each outcome | A stale reference, a de-vig method that flatters longshots, a market too wide to price inside |
| 2. Find a price above it | Edge = offered price / fair price - 1 | A different line or market counted as the same bet, so the edge is manufactured |
| 3. Decide the stake | Turn the edge and the price into a dollar figure | Overbetting a noisy estimate, or a stake above what the bookmaker will accept |
| 4. Place it | Get the bet on at or above a price floor | The price moved, the bet was refused, or the same bet went on twice |
Steps 1 to 3 can be checked on paper and corrected at leisure. Step 4 happens once, against a price that is already moving, and it is the step that decides what the first three were worth.
Why the last step leaks the most edge
Take an illustrative selection. The margin has been taken out across the market and the fair price is $1.95, which is a fair chance of 1 / 1.95 = 51.28%. One bookmaker is showing $1.98.
Edge = 1.98 / 1.95 - 1 = 1.54%. Per $100 staked the average value is the edge times the stake, 0.0154 x 100 = $1.54.
Now move the bookmaker price a cent at a time and leave the fair price where it is.
| Bookmaker price | Edge | Average value per $100 |
|---|---|---|
| $1.98 | 1.98 / 1.95 - 1 = 1.54% | $1.54 |
| $1.97 | 1.97 / 1.95 - 1 = 1.03% | $1.03 |
| $1.96 | 1.96 / 1.95 - 1 = 0.51% | $0.51 |
| $1.95 | 1.95 / 1.95 - 1 = 0.00% | $0.00 |
Three cents is the entire bet. That is the arithmetic behind the claim that placement is where the edge goes.
Example: at $1.98 the bet is worth $1.54 per $100. Place it two cents late, at $1.96, and it is worth $0.51 per $100: the same selection, a third of the value, and nothing about the event changed.
There is a worse version of the same move, and it does not show up as a smaller number. If the whole market drifts while the bookmaker stays at $1.98, so that the fair price is really $2.00 by the time the attempt lands, the bet is 1.98 / 2.00 - 1 = minus 1.00%. The row still read 1.54%, because the fair price behind it was the old one: a stale reference does not shrink an edge, it flags bets that were never positive. A price that is out of line is usually out of line because one bookmaker has not updated yet, so the window is short by nature. Prices on any board are read on a repeating cycle rather than tick by tick, so the number on screen can trail the bookmaker's own.
What a price floor does, and what it costs
A floor is the rule that converts the leak into a refusal. Instead of placing whatever price is there when the attempt lands, the bet carries a minimum: at or above the floor it goes on at the bookmaker's current price, under it nothing is placed. Betting bot security covers the mechanics and what a refused attempt leaves in the record; the question here is where the number goes and what it costs.
Set the floor at the price that still leaves the edge you need. If 1% is your minimum and fair is $1.95, the floor is 1.01 x 1.95 = $1.9695, so $1.97 at the next cent a bookmaker quotes. At $1.96 the attempt is refused, and the refusal is correct: a 0.51% edge was not the bet that was modelled.
The cost of a floor is placement rate, not value. A tight floor refuses more attempts, and a floor set at the exact price you saw on screen refuses every attempt where the price has shortened at all, because the screen is behind the bookmaker by design. The useful habit is to read a refused attempt as information about the market rather than a fault: a strategy whose attempts are nearly all refused is a strategy whose edges were mostly gone already.
Writing the rules down before you automate them
A rule applied by judgement cannot be automated until it is a number. Before anything is switched on, each of these has to have a value, and a session is only as sensible as the worst of them.
- Minimum edge. The threshold under which an attempt is not worth making.
- Odds range. A minimum and a maximum price. Long prices make the edge estimate least reliable and the variance largest.
- Stake rule and caps. Fixed, banded by price, or a share of balance, with a minimum and a maximum stake in dollars.
- Which markets. Named exactly, including which lines count as the same bet, because a total at 2.5 is not a total at 3.5.
- Which accounts, at which bookmakers. Accounts in your own name, and nothing else.
- Run windows. Daily start and stop times, so an unattended session is not running at hours nobody meant it to.
- Stop conditions. A profit or loss figure at which new bets stop.
Note: these settings are ordinary software settings, not protections. A command can fail or act late, so the deposit limit set with each bookmaker is the firmer cap. See responsible gambling.
Sizing a small edge with a fraction of Kelly
The Kelly stake is edge divided by the price minus one, as a share of bankroll. At $1.98 with a 1.54% edge that is 0.0154 / 0.98 = 1.57% of bankroll, which on a $5,000 bankroll is about $79.
A fraction of that is the usual answer, and the reason is the estimate rather than nerve. Suppose the real edge was 0.54%, one percentage point lower, because the reference used to build the fair price was a minute old. Full Kelly on the true number is 0.0054 / 0.98 = 0.55% of bankroll, about $28. The $79 stake was nearly three times the right size. A quarter Kelly stake of 1.57% / 4 = 0.39%, about $20, would have been slightly under it instead.
That asymmetry is the whole argument: overbetting a noisy edge costs more growth than underbetting it does, and an automated session makes the same sizing mistake on every bet rather than one. Fractional Kelly works through the growth and drawdown side of the choice.
Measuring it: closing line value, not profit
Profit over a few hundred bets at edges near 1% is mostly noise, so it is the slowest possible way to find out whether the first three steps work. Closing line value asks a narrower question with far less variance in the answer: did the price you took beat the market's last price, once the margin is taken out of the close.
CLV = your price / the margin-free closing price - 1. Placed at $1.98 against a margin-free close of $1.92, that is 1.98 / 1.92 - 1 = 3.13%. Placed at $1.98 against a close of $2.05, it is 1.98 / 2.05 - 1 = minus 3.41%: a price the market later said was short, whether the bet won or lost.
Read per step, CLV is also a diagnostic. Good CLV with poor results is a sample size problem. Poor CLV on prices a screener rated highly points at step 1 or step 2. Good CLV on the attempts that landed, with most attempts refused, points at step 4 and nothing else. Closing line value covers the sum and its edge cases.
What automating value betting does not fix
Automation changes how fast the four steps run. It changes nothing about who decides whether a bet is accepted.
Bookmaker terms govern. Many bookmakers restrict automated betting in their terms, the bookmaker decides whether any single bet or stake is accepted, and no software prevents an account being restricted or closed. A restricted account is the quiet version of the same problem: small stakes on a good price are worth little, so a strategy can be right and have nowhere to run. Why bookmakers restrict accounts sets out what usually prompts it.
Automation also makes it possible to bet far more, far faster, than by hand, which is a reason to set the stake caps and run windows before the first session rather than after. Deposit limits are set with each bookmaker and cap how much new money any strategy can put at risk.
Where B337 fits
B337's +EV screener lists bookmaker prices across sports markets that sit above a fair price built by taking the margin out across the market, and ranks them by edge, with filters for sport, market and bookmaker. The de-vig is adjustable: change which prices count and how much margin comes off each, and every edge is worked out again to match.
Placing is a separate decision and stays with the account holder. For sports, on Terminal Pro or Full Automation you can place from the Terminal into your own bookmaker accounts, and code can send bets through the betting API, where target_odds is the floor described above. For racing, sessions you control can place +EV bets in your own accounts on the rules you set. The software runs on your own computer, so nothing is placed while it is off. Bets placed through B337 use credits, a per-bet usage charge, so count them beside the plan price. A free account shows the live count of prices above fair and the best edge, with the rows locked.
Risk: Betting involves risk. Bookmakers can restrict or close accounts and void bets, automation can fail, prices move, and a positive expected value bet can still lose. There is no guarantee of profit. Automating placement does not make a small edge dependable: the fair price behind every row is an estimate, and a price that has moved is a different bet from the one that was flagged. See responsible gambling for limits and support.
For free and confidential support call 1800 858 858 or visit gamblinghelponline.org.au.