What Is Micro Betting and How Micro-Markets Are Priced Live

Micro betting is wagering on the next small moment of a live match, such as the next point or a goal in the next five minutes. Each micro-market lasts seconds or minutes, is priced from live data and settles fast.

This article explains what a micro-market is, how its time window works, and how a price for it is built from a statistical model and a margin. It walks through two worked cases with real arithmetic, then covers the two operations that make or break a micro-market product: suspension and settlement. If you run a licensed sportsbook and want the module itself, the micro betting page describes what RealTimeBets Platform generates and how it is delivered.

What a micro-market is

A traditional market asks a question about the whole match: who wins, how many goals, how many points. A micro-market asks a question about the next slice of play. The answer arrives quickly, often in less than a minute, and a new question replaces it right away.

Typical micro-markets by sport look like this:

• Soccer. Goal in the next 5 minutes, next corner, next card in the next 10 minutes.
• Tennis. Winner of the next point, ace on the next point, server holds the current game.
• Basketball. Result of the next possession, next team to score, points in the next two minutes.
• American football. Result of the next play, next drive outcome (touchdown, field goal, punt, turnover).
• Baseball. Outcome of the next pitch, result of the current plate appearance.
• Ice hockey. Goal in the next 5 minutes, next team to shoot on goal.
• CS2. Winner of the next round, first kill of the round.

The RealTimeBets engine generates a defined set of these: goal in a 5-minute window, shot on target in 1 minute and corner in 2 minutes in soccer; next point, ace and deuce in tennis; next possession, three-pointer in 60 seconds and points in 2 minutes in basketball; next drive and points in 3 minutes in American football; on base, strikeout and run in the half inning in baseball; goal in 3 minutes and shot in 1 minute in ice hockey; next round and bomb plant in CS2.

Because every market is short, a single match can produce hundreds of them. That volume is the commercial point: a customer who places one pre-match bet may place many small in-play bets if the markets are clear, fast and always available.

Windows define every micro-market

Every micro-market has a window. The window says when the question starts and when it ends, and it must be defined in a way that both the operator and the customer can check after the fact.

There are two kinds of windows:

1. Time windows. "Goal between 61:00 and 65:59 of the match clock." These suit continuous sports where scoring is rare: soccer and ice hockey.
2. Event windows. "Winner of the next point," "result of the next pitch." These suit sports that are already divided into discrete units: tennis, baseball, American football, and rounds in esports.

The window has to use the official match clock from the incident feed, not the wall clock of a server. Stoppage time, video reviews and clock stops would otherwise make two systems disagree about whether a goal fell inside the window. In a well-built engine the window boundaries are stored with the market (a start and an end value), so settlement later reads the same numbers that were shown to the customer.

From odds to probability and back

Decimal odds and probability are two views of the same number. The implied probability of a price is one divided by the decimal odds.

A bookmaker never offers the fair price. It adds a margin, also called the overround. The overround is the sum of the implied probabilities of every outcome in the market, minus one:

Overround = (1 / odds of outcome 1) + (1 / odds of outcome 2) + ... − 1

A worked case for the margin

Take a "winner of the next point" market in tennis where both sides are offered at 1.91.

• 1 / 1.91 = 0.5236 for each side.
• Sum: 0.5236 + 0.5236 = 1.0471.
• Overround: 1.0471 − 1 = 0.0471, so the book holds a margin of about 4.7%.

To read the fair probabilities back out of the offered prices, divide each implied probability by the sum: 0.5236 / 1.0471 = 0.50. Both sides are judged equally likely, and the 4.7% is the operator's price for taking the risk.

On a micro-market the margin is usually set higher than on a full-match market. The windows are short, the model has less information, and the operator faces more exposure to customers who are physically at the venue or watching a faster stream. A configurable margin per market type lets the trading team decide that trade-off instead of hard-coding it.

Pricing "goal in the next 5 minutes" with a Poisson model

Goals in soccer are rare and roughly independent of each other over short spans, which is the textbook setting for a Poisson process. The model needs one input: the expected number of goals per unit of time.

Suppose the pre-match and in-play data put the total expected goals for the full 90 minutes at 2.7. The step-by-step calculation is:

1. Rate for the window. λ for 5 minutes = 2.7 × 5 / 90 = 0.15 expected goals.
2. Probability of no goal. P(0) = e^−0.15 = 0.8607.
3. Probability of at least one goal. P(at least 1) = 1 − e^−0.15 = 1 − 0.8607 = 0.1393, so 13.9%.
4. Fair odds. "Yes" at 1 / 0.1393 = 7.18. "No" at 1 / 0.8607 = 1.16.
5. Apply a 6% overround. Scale both probabilities by 1.06: "Yes" 0.1477, "No" 0.9124. Offered odds: "Yes" 6.77, "No" 1.10 (rounded). Check: 1 / 6.77 + 1 / 1.10 is close to 1.06.

In a live engine the rate is never a flat 2.7 / 90. It moves with the score, the clock, red cards and the state of play. A team chasing a goal late in the match raises the rate, a red card can raise it for one side and lower it for the other, and the final minutes of a half often carry a higher scoring rate than the middle. That is why prices are recomputed every second in RealTimeBets: the model is re-run as the inputs change, and each selection receives a new price even when nothing visible has happened on the pitch.

Other sports use other models. Tennis micro-markets follow a serve-hold Markov chain, where the probability of winning the next point on serve feeds the probabilities for the game, set and match. Basketball and American football use possession models that describe what each possession or play tends to produce. The structure is the same everywhere: a probability from a model, a margin on top, a price out.

Suspension after dangerous incidents

A price is only as good as the information behind it. When something important happens on the field, every market that depends on it is out of date until the model has seen the new state. Taking bets in that gap hands money to whoever saw the incident first.

Engines handle this with suspension. The incident feed marks certain incidents as dangerous: a goal, a penalty award, a red card, a break point in tennis, a score in basketball or football. When such an incident arrives, the affected micro-markets are suspended for a few seconds. During suspension no bets are accepted. The model reprices, the new prices are published, and the market reopens or closes if its window has passed.

Three design rules keep suspension honest:

• Suspend first, compute second. The suspend flag must reach the customer before the new price does.
• Suspend by dependency. A goal suspends goal and result markets, but a "next corner" market may only need a short pause.
• Log every state change. Open, suspended and settled states with timestamps let the operator answer disputes with evidence.

The risk side of suspension (liability limits, customer limits, flags on accounts that keep beating the line) is covered on the risk engine page.

Settlement from the incident feed

A micro-market is only useful if it settles fast. A customer who called the next point wants the result before the following point is played. Settlement therefore cannot wait for a human. It reads the same incident feed that drove pricing.

For a time window, the settlement rule compares incident timestamps on the match clock with the window boundaries stored on the market. A goal at 63:12 in a 61:00 to 65:59 window settles "Yes". If the window closes with no goal, "No" wins. For an event window, the rule waits for the next incident of the defining type (the next point, the next pitch result) and settles from it.

Two edge cases matter. First, corrections: a goal can be ruled out after a review. The engine must be able to reverse a settlement and record why. Second, abandonment: if a match stops inside a window, the market is voided and stakes returned according to the operator's rules. Both need to be written into the market rules before the market is offered, not decided afterwards.

On RealTimeBets Platform, each bet the operator recorded fires a signed bet.settled webhook when its market resolves, so the wallet and bet history update without polling.

Try micro-markets without money

The quickest way to understand the rhythm of micro-markets is to play them. The RealTimeBets prediction game runs on the same engine: you call the next goal window, point or round for points, watch the multiplier move every second, and see your call settle from the incident feed. It is a game, no money is wagered and points have no cash value. Operators who want the same markets for their licensed book can compare plans on the pricing page.

Questions about micro betting

How is micro betting different from in-play betting?

In-play betting covers any market offered while the match is live, including full-match markets like the final result. Micro betting is the subset that asks about the next short moment, with windows of seconds or minutes.

Why do micro-market prices change when nothing seems to happen?

The model uses the clock as an input. As time passes without a goal, the expected goals left in the window fall, so the "Yes" price drifts upward and the "No" price downward, every second.

What margin is normal on a micro-market?

There is no single number. Operators usually set a wider margin than on full-match markets because the windows are short and information moves fast. RealTimeBets prices carry a margin of about 6 percent over the fair probability.

What happens if a goal is ruled out after a review?

The market should have been suspended when the goal was reported. If it was already settled, the engine reverses the settlement, resettles from the corrected incident and records the change.

Is the Poisson model accurate enough for live markets?

It is a sound base for rare, roughly independent events like soccer and hockey goals. Live engines adjust the rate with score, clock and match state rather than using a single flat rate.