UFC value betting means backing a fight price that sits above the fair price, and in MMA the hard part is not the idea, it is the height of the bar. One bout is a thinner market than a round of football fixtures, the book is wider to match, and the improvement you need before a bet exists is exactly that extra margin.
The method does not change: turn each price into a chance, add them up, scale back to 100%, compare. What belongs to this sport is how much margin sits in the book, how few bets a card offers, and how easily a fight changes shape before the cage door shuts.
Thin liquidity means a wider book
Removing margin evenly hides a useful identity. If the fair price comes from scaling every implied chance down by the same factor, then fair price = quoted price x book total. So the improvement you need over the quote you de-vigged is the overround, nothing more and nothing less. Three illustrative book totals against the one quoted price:
| Two-way book total | Fair price of a $1.53 favourite | Improvement needed on $1.53 |
|---|---|---|
| 102.51% | $1.57 | 2.51% |
| 105.00% | $1.61 | 5.00% |
| 107.03% | $1.64 | 7.03% |
The same opinion about the same favourite needs a 2.51% better price in a tight market and a 7.03% better price in a wide one, and the wide shape is what a thinly traded bout tends to be quoted in: fewer operators price a prelim, less money goes through it, and margin covers the uncertainty money would otherwise resolve. So the first number to work out on a fight is the book total, not an edge, because a big total leaves no fair price precise enough to hold a small edge. Bookmaker margin covers the total.
Worked example: a two-way fight market
Illustrative prices on one bout, chosen to show the arithmetic rather than to describe a real fight.
| Fighter | Priced at | Implies | Fair price |
|---|---|---|---|
| A, the favourite | $1.53 | 65.36% | $1.64 |
| B, the underdog | $2.40 | 41.67% | $2.57 |
One over 1.53 is 65.36% and one over 2.40 is 41.67%, so the chances add to 107.03% and the book carries 7.03% of margin. Scaling both down by that factor leaves fair chances of 61.07% and 38.93%, the fair prices in the last column. Now say the best price anywhere is $1.65 on A and $2.50 on B.
Example: A at $1.65 against a fair $1.64 is an edge of 0.76%, because 1.65 / 1.6375 - 1 = 0.0076. B at $2.50 against a fair $2.57 is an edge of minus 2.67%. Each was the highest price quoted on its fighter, and only one of them was a bet.
The underdog is the instructive half: $2.50 is 4.2% better than the $2.40 the fair price came from, which feels like real improvement and still falls short of the 7.03% the book demanded.
Sizing an edge this thin
The Kelly stake is the edge divided by the decimal price minus one, as a share of bankroll. On A at $1.65 with a 0.76% edge that is 0.0076 / 0.65 = 1.17% of bankroll, and a quarter Kelly on a $3,000 bankroll is about $8.80.
What belongs to MMA is not the size of that stake but when it is exposed. A card is one evening: a dozen bouts, every bet struck inside a few hours and every one settled before the night ends, so a month of fight betting can sit on the board at the same time. Eight stakes of $8.80 is $70 of a $3,000 bankroll live at once, which is why the number to decide first is the card's, not the bout's. The outcomes are independent of each other, but the pricing is not: one wrong assumption about a thinly traded card is wrong on every bout of it, so a bad fair price does not average out across the night the way a bad read on one fighter would. Cap what may be out on one event before the first bout, and let that cap the per-bet figure rather than the other way round.
Risk: Betting involves risk. A fair price is an estimate rather than a probability, so a bet above it loses often and there is no guarantee of profit; at this size of edge a long losing run is ordinary rather than evidence of anything, and bookmakers can restrict or close accounts and void bets. See responsible gambling for limits and support.
Why one fight tells you nothing
At a true 61.07% and a price of $1.65, a $100 bet wins $65 about 61 times in 100 and loses $100 about 39 times, so it expects 0.6107 x 65 - 0.3893 x 100 = $0.76. The standard deviation of that one bet is the square root of 0.6107 x 0.3893 x 165 x 165, about $80.45.
Expected profit grows with the number of bets while the swing grows with the square root of it, so the two are equal when 0.7634n = 80.45 times the square root of n. That gives n = (80.45 / 0.7634) squared, about 11,100 bets, or more than 900 cards at a dozen bouts each. That is only the point where the edge matches one standard deviation; the two standard errors usually asked of a record sit four times further out, near 44,000 bets. A fight is the purest version of the problem too, often close to a coin flip settled by one punch, so the result says almost nothing about whether the price was wrong. Measure closing line value instead, whether your price beat the market's last price, and see sample size for why profit waits so long.
Method and round markets are a different order of margin
UFC betting in Australia sets out what method of victory, round betting and the round over or under each settle on, and that the book across them runs wider than the head to head. The value question is the one after that: a wide market does not merely price worse, it disagrees with the tight market about what the fair price is, and the disagreement is larger than anything you were looking for.
Illustrative prices on the same bout. No draw is priced on this market, so the six outcomes de-vig onto the same winner-only basis as the two-way and the two readings are comparable:
| Outcome | Priced at | Implies |
|---|---|---|
| A by KO or TKO | $3.00 | 33.33% |
| A by submission | $5.50 | 18.18% |
| A by decision | $4.50 | 22.22% |
| B by KO or TKO | $6.00 | 16.67% |
| B by submission | $10.00 | 10.00% |
| B by decision | $7.50 | 13.33% |
Those add to 113.74%. A's three outcomes sum to 73.74% and B's to 40.00%, so an even de-vig puts A on 64.83% and B on 35.17%, fair prices of $1.54 and $2.84. The two-way market said A was 61.07% and $1.64, a disagreement of 3.76 percentage points. Put that in the units an edge is measured in, because percentage points of probability are not comparable with it: $1.54 against $1.6375 is 6.2%, about eight times the 0.76% the earlier example was chasing.
Note: when a wide market and a tight one on the same fight disagree, trust the tight one and suspect the de-vig, which assumed a $10.00 longshot carries margin in proportion to a $3.00 favourite. A six-outcome market is not a second opinion about the fight; it is the same opinion read through more margin.
A replacement opponent voids the fair price too
MMA changes its fixtures at a day's notice in a way a football round does not. A withdrawal, an injury at the weigh-in or a failed medical can put a new opponent in the cage, and what happens to the bet is set out in what happens to a UFC bet if the fight is cancelled.
The second casualty is the one people miss: the fair price was built for a matchup that is not happening. Carrying it across is the same error as comparing two different handicaps, except that nothing on the screen looks different, because the event name, the date and the favourite's name can all survive the change. Rebuild from the reopened market and measure its own book total.
Tip: the same applies to a change in the scheduled rounds. A bout moved from three rounds to five is a different proposition for every round, distance and decision market on it, even where the two-way price barely moves.
Settlement rules can be bigger than the edge
Take the draw. A two-way fight market with no draw price either refunds the stake or loses both sides under a dead heat or similar rule. On a refund the bet is really on A winning given that somebody wins, so the conditional fair price of $1.64 stands. If the draw loses, the unconditional chance is the one that counts: with an illustrative 1% draw chance that is 61.07% x 0.99 = 60.46%, a fair price of $1.65.
Example: at $1.65 the edge is 0.76% if a draw refunds and minus 0.24% if a draw loses. The settlement rule moved the number by 1.01 percentage points, more than the whole edge.
Three more to read before the bet rather than after it: a no contest, which an accidental foul can produce and which usually voids the two-way market; a missed weight, which commonly leaves the same two fighters in a catchweight bout with the bets standing; and a technical decision, scored on the completed rounds. UFC betting in Australia defines each market.
Where B337 fits
The +EV screener lists bookmaker prices that sit above a fair price and ranks them by edge, filtered by sport, market and bookmaker, and MMA is one of the sports B337 sessions cover. The fair price comes from taking the margin out of prices across the market and combining them, and the de-vig is adjustable: you choose which prices count and how much margin comes off each, and every edge is worked out again to match. On a six-outcome method market, that adjustment is the difference between the two readings above.
Four limits, plainly. The fair price is an estimate rather than a known probability, so a price above it can still lose. Prices are read on a repeating cycle rather than tick by tick, so confirm one in your own account before you bet. B337 places no MMA bet by itself: the screener finds the price, and on Terminal Pro or Full Automation you place it from the Terminal into bookmaker accounts you hold yourself. And the bot that places a bet runs on your own computer, so nothing goes on while that computer is off. A free account shows the live count of prices above fair and the best edge, with the rows locked, and bets placed through B337 use credits on top of the plan price.
For free and confidential support call 1800 858 858 or visit gamblinghelponline.org.au.
For the same arithmetic in deeper markets see AFL value betting, and for the method in general see value betting.