A Study of Odds Movement in Value Betting
Is it effective to focus on static odds in Value Betting?
A: At present, most Value Betting articles, papers, and open-source projects roughly follow this pattern.
One point in time（Closing Odds、Current Odds、Opening Odds）→ estimate true probability → determine whether it is Value.
For markets where odds fluctuate significantly, this kind of sampling is no different from a random walk. It runs against the real judgment of Value, because it hands the core of the value judgment to one specific point in time at which the bookmaker priced the event, rather than to the real pricing process. Many factors affect bookmaker pricing:
Real information appears ➔ bookmaker adjustment ➔ professional players place bets ➔ market capital flow ➔ other bookmakers synchronize ➔ arbitrage capital enters ➔ bots follow the market ➔ pre-match risk control
If we believe that all financial investment value comes from mispricing, this instantaneous price may be the Value we need, but it may also not be. Under the premise that quantitative betting pursues long-term stability, handing an indicator over to randomness and luck is somewhat careless. Of course, I cannot be too absolute about this issue, because this approach is, after all, the mainstream practice in Value Betting.
What is dynamic odds (Odds Process)?
A: What we cannot avoid is that, for any match, the market odds follow a process like this:
t0 → t1 → t2 → t3 → t4 → t5 2.30→2.28→2.25→2.18→2.10→1.98 (Time Series)
In finance, this is called a Price Path rather than a Price. Each odds Time Series therefore has many dynamic characteristics, such as:
Velocity: 24 hours before kickoff: 2.30 ↓, 23 hours before kickoff: 2.25, down 0.05; Velocity = −0.05 / hour,
The velocity of different match markets can be completely different.
Acceleration: 24 hours before kickoff: 2.30 ↓, 20 hours before kickoff: 2.28 ↓, 16 hours before kickoff: 2.15, then it suddenly accelerates. This indicates that new information has entered the market.
Volatility: 2.10 → 2.15 → 2.08 → 2.18 → 2.05
This often means that market opinions are divided.
Jump: 2.25 → 2.25 → 2.25 → 1.98
A sudden jump.
Some markets maintain a one-sided downward trend, while others show a pattern of first falling, then rising, and finally falling again. Whether the trend itself can continue is an important analytical feature, and the synchronization of odds movements across different platforms has even higher research value.
For example,
Pinnacle odds move down from 2.05 to 2.00. One minute later, bet365 odds move from 2.08 to 2.02. Another minute later, 188 odds move down from 2.09 to 2.03.
This complete movement process clearly reflects the relationship between leaders and followers in odds movement. Many professional analysis teams focus on the sequence of odds changes across platforms, distinguish the source platform that adjusts first from platforms that merely copy the adjustment, and locate the true price source that holds first-hand market information. This analytical logic belongs to the field of Price Discovery.
The odds movement path can also be extended into multiple analytical dimensions:
First is Retracement,
for example, odds fall from 2.20 to 2.05 and then rise back to 2.15,
which can be used to judge whether the move is a false breakout;
Second is the number of oscillations, which counts the total frequency of back-and-forth odds movement;
Third is the extreme continuous trend, including the maximum number of consecutive upward moves,
for example, 7 consecutive rises,
and the maximum number of consecutive downward moves,
for example, 12 consecutive declines;
Fourth is maximum drawdown. The logic of this indicator is fully consistent with financial market analysis;
Fifth is range breakout judgment.
If odds stayed stable between 2.15 and 2.20 over the past 20 hours, and then directly dropped to 2.05,
this constitutes a Breakout signal.
On this basis, we can further analyze the linkage between Line Movement and odds. There is a critical detail here that is very easy to overlook: the handicap line level also adjusts at the same time. It is not only the odds that move independently. This type of linkage information has much higher analytical value than studying odds movement alone.
For example, the handicap changes from home team -0.25 to home team -0.5, and then to home team -0.75,
while the corresponding odds move from 2.05 down to 1.95 and then back up to 2.02.
If we only observe the odds values, we may mistakenly think that the overall market movement is very small. In reality, however, the handicap level has already crossed two levels. The essence of the change is that the market's implied expectation of the true strength gap between the two teams has changed.
For this reason, professional quantitative models do not analyze handicap and odds as two isolated datasets. Instead, based on goal prediction models such as Skellam, Poisson, and Dixon–Coles, they convert handicap and odds into a unified continuous implied-strength dimension, and then track the full evolution of that value over time.
How can dynamic odds be used for quantitative analysis?
A: Once odds are treated as a Series, many models from finance can be borrowed:
- Random Walk Translation: Random Walk Explanation: The next upward or downward move in asset prices is completely irregular and memoryless. Past movements cannot predict the future, and price changes are mutually independent.
- Brownian Motion Translation: Brownian Motion Explanation: The continuous-time version of Random Walk. It describes continuous small random fluctuations in asset prices and is a foundational model for options and price simulation.
- Hidden Markov Model（HMM） Translation: Hidden Markov Model Explanation: Hidden states exist but cannot be directly observed. They can only be inferred from visible price data. State transitions follow the Markov property, and the model is often used to identify market regimes.
- State Space Model Translation: State Space Model Explanation: The model is split into two equations: one describes the unobservable internal state, and the other describes the observed price data. It can handle noisy and dynamic time-series data.
- Regime Switching Translation: Regime Switching / State Switching Model Explanation: The market is divided into multiple operating states, such as oscillation, bull market, and bear market. The model can automatically identify the timing and probability of transitions between different states.
- Change Point Detection Translation: Change Point Detection Explanation: Algorithms designed to identify structural breaks in time-series data, locating the time points where odds, volatility, or returns suddenly change.
- Survival Analysis Translation: Survival Analysis Explanation: Originally used in lifetime statistics. In finance, it is used to estimate how long a certain market condition, spread, or handicap state can persist before reversal or breakout.
- Kalman Filter Translation: Kalman Filter Explanation: A recursive estimation algorithm paired with State Space Model. It filters market noise and continuously updates predictions of the implied market state in real time. Research on the connection between specific models and odds sequences, as well as their applications, will continue to be updated.