Draw no bet is a football market with two outcomes instead of three. You back a team to win, and if the match finishes level your stake is returned in full. The draw is removed from the equation, and in exchange the odds are shorter than a straight win bet on the same team. It is the market to understand if you have ever watched a correct call end 1–1 and lose anyway, which is exactly what happened to nine of our 14 losing predictions this season.
How draw no bet works
A standard football match has three results: home win, draw, away win. That is the 1X2 market. Draw no bet, usually written DNB, collapses it to two:
- Your team wins: the bet wins and pays at the DNB price.
- The match is drawn: the bet is void and your stake comes back. Not a win, not a loss.
- Your team loses: the bet loses in full.
Bookmakers price it by stripping the draw out and redistributing its probability across the two remaining outcomes. That is why a team priced around 2.20 to win outright will normally be a good deal shorter on DNB. You are buying insurance against the draw, and the premium is taken out of your potential return. The whole question is whether that premium is worth paying, and that is an empirical question, not a matter of opinion.
What our own record says about draws
We publish every prediction with a confidence score and grade all of them in public against the official result, wins and losses alike. Under our rules a pick wins only if the named side wins the match, so a draw counts as a loss. That grading choice turns out to be expensive.
Across 33 graded picks between 4 and 23 August 2026:
| Outcome | Picks | Share of all picks |
|---|---|---|
| We named the winning side | 19 | 58% |
| Match finished level (graded as a loss) | 9 | 27% |
| We named the wrong winner | 5 | 15% |
| Total graded | 33 | 100% |
Read the middle row again. Nine of our 14 losses, 64% of everything that went wrong, were draws. Only five picks in the whole sample actually backed the losing team. The model was not often wrong about which side was better. It was repeatedly caught by matches that refused to produce a winner at all.
The nine that finished level
| Competition | Match | Our pick | Conf. | 1X2 | On DNB |
|---|---|---|---|---|---|
| Champions League | Olympiacos 0–0 NEC Nijmegen | Olympiacos | 62% | LOSS | REFUND |
| MLS | Real Salt Lake 1–1 Minnesota | Real Salt Lake | 57% | LOSS | REFUND |
| MLS | Sporting KC 1–1 St. Louis City | St. Louis City | 53% | LOSS | REFUND |
| Serie A | Udinese 1–1 Como | Como | 52% | LOSS | REFUND |
| La Liga | Atletico Madrid 2–2 Villarreal | Atletico Madrid | 50% | LOSS | REFUND |
| Premier League | Newcastle 2–2 Liverpool | Liverpool | 49% | LOSS | REFUND |
| MLS | Austin 1–1 Philadelphia | Philadelphia | 45% | LOSS | REFUND |
| La Liga | Rayo Vallecano 1–1 Alaves | Rayo Vallecano | 42% | LOSS | REFUND |
| MLS | Philadelphia 2–2 Inter Miami | Philadelphia | 41% | LOSS | REFUND |
The draw is not spread evenly. It clusters on close calls
This is the finding that changed how we read our own board. Sorting the same 33 picks by the confidence band we published them in shows the draw is overwhelmingly a low-confidence problem.
| Confidence band | Picks | Ended in a draw | Draw rate | 1X2 strike rate |
|---|---|---|---|---|
| Strong, 70% and above | 7 | 0 | 0% | 86% |
| Good, 55% to 69% | 13 | 2 | 15% | 62% |
| Close call, under 55% | 13 | 7 | 54% | 38% |
That makes intuitive sense once you see it. A match our model rates as close to a coin flip is a match between evenly matched sides, and evenly matched sides draw. A match we rate at 75% or 80% is usually a mismatch, and mismatches resolve. The confidence score is not just telling you how likely we are to be right. It is also telling you how likely the match is to produce a winner at all.
What the draw actually costs, in numbers
Take the same 33 picks and settle them three different ways. The break-even price is the average odds you would have needed just to get your money back.
| Market | How draws settle | Effective strike rate | Break-even odds |
|---|---|---|---|
| 1X2 (straight win) | Loss | 19 from 33, 58% | 1.74 |
| Draw no bet | Stake refunded | 19 from 24 at risk, 79% | 1.26 |
| Double chance | Wins | 28 from 33, 85% | 1.18 |
Those break-even figures are not the same thing as profit. They are the hurdle. A 1X2 bet on our picks needed average odds above 1.74 to be worth making. The same picks on draw no bet needed only 1.26, because nine of the 33 came back rather than losing.
The rule that falls out of it
Compare the two directly. Call the 1X2 price o and the draw no bet price d for the same selection. Draw no bet returns more than 1X2 whenever:
On our record that is 9 draws divided by 19 wins, which is 0.47. So over this sample, draw no bet beat the straight win bet whenever the DNB price sat within 0.47 of the 1X2 price in decimal odds.
The practical consequence is the opposite of what most people assume. On short-priced favourites the gap between the 1X2 and DNB price is usually small, well inside 0.47, so the insurance is cheap. On near coin-flip matches the gap is usually much wider, which is precisely where the draw risk is highest and the cover is most expensive. Bookmakers are not unaware of this.
Two honest caveats. First, 33 picks is a small sample and a 27% draw rate is on the high side of what football normally produces over a full season. Second, the 0.47 figure is descriptive of our record to date, not a constant. We will republish it as the sample grows, and it will move.
Draw no bet vs double chance
They solve the same problem in different ways and are frequently confused.
| Scenario | Draw no bet | Double chance |
|---|---|---|
| Team wins | Pays at DNB price | Pays at DC price |
| Match drawn | Stake refunded | Pays at DC price |
| Team loses | Stake lost | Stake lost |
| Odds | Shorter than 1X2 | Shortest of the three |
| Best used when | You want the winner but fear a stalemate | You mainly want the other side not to win |
| On our 33 picks | 19 paid, 9 refunded, 5 lost | 28 paid, 5 lost |
Double chance covered every one of our draws and needed only 1.18 to break even, but at those prices you are staking a lot to win a little, and a single wrong call wipes out several correct ones. Draw no bet sits between the two.
How we use this on our own board
We do not publish DNB or double chance selections, and we are not going to start dressing our 1X2 picks up as safer markets to flatter the record. What we do now is label the risk properly:
- Picks under 55% confidence carry a close call badge. Read that as a real warning: 54% of them have finished level.
- Bet of the day is only issued when the model has a clear edge. On days when it does not, we say so rather than manufacture one.
- Every pick stays on the record after the result, including the ones that ended 1–1.
Keep reading
- Weekend predictions results, 22 to 23 August 2026, the round where four of six losses were draws
- How accurate are football predictions? Our real, graded numbers
- Correct score predictions, and why the hit rate on that market is 10 to 15%
- What makes a good football prediction site? The 7-point checklist
- Today’s football betting tips
Common questions
Is draw no bet the same as a handicap?
No. An Asian handicap of 0.0 settles almost identically to draw no bet, but the standard handicap markets apply a goal head start or deficit before settlement. Draw no bet applies none.
Does draw no bet cover extra time or penalties?
In knockout competitions, DNB normally settles on 90 minutes plus stoppage time only. A tie decided in extra time or on penalties is still a refund if the 90 minutes finished level. Check your operator’s rules, because this varies.
Is draw no bet a safer bet?
It reduces one specific risk, the draw, and charges you for it in shorter odds. It does nothing about the risk of backing the wrong team, which accounted for five of our 14 losses. No market removes that.
Which leagues draw most often?
Our sample is too small to answer that responsibly. Seven of our nine draws came from MLS, La Liga and the Champions League, but with only 33 picks that is not evidence of a league effect. We will revisit it once the record is several hundred picks deep.