Gamblitude predictive models
A forward-looking view of player value.
Over the next 30 days.
Estimate the sportsbook GGR each eligible player is expected to generate. Compare future revenue potential with recent performance to support review, segmentation and CRM planning.
01 / What the model tells you
Player GGR Prediction estimates the sportsbook gross gaming revenue a player is expected to generate over the next 30 days. It produces a numerical revenue estimate, refreshed daily for the eligible player population.
Use the forecast alongside recorded GGR to compare players and investigate changing potential.
02 / History and future potential
Recent GGR describes realised performance. The forecast asks a different question: what revenue is expected over the coming 30 days? Compare both views to understand which players warrant a closer look.
The historical view
A recent GGR ranking shows what players have generated. A short period can also reflect unusual wins or losses, or activity that is now changing. Review that context before assuming the same pattern will continue.
The scoring point
The model combines selected attributes from your feature store with patterns learned from your historical data. Eligible players receive a refreshed estimate each day, based on the data available at that scoring point.
The forward-looking view
Compare the 30-day estimate with recent performance. Differences help your team decide which players and cohorts to investigate. Once the period has elapsed, evaluate it against the GGR actually recorded.
Each daily forecast covers a new, overlapping 30-day period.
03 / From a forecast to a decision
A revenue estimate gives your team another basis for prioritisation. Use it with activity, lifecycle stage and the player context relevant to the decision.
VIP & commercial teams
Identify players with high expected GGR outside your existing VIP group. Apply your programme’s eligibility requirements and player protection checks before deciding on the next step.
CRM teams
Combine forecast bands with activity, lifecycle stage and other criteria. Use that context to prioritise team attention and design a campaign test.
Analytics teams
Compare predicted GGR with recent performance. Investigate the players and cohorts behind meaningful differences, then follow what happens in the forecast period.
04 / A broad feature store
Gamblitude’s feature store brings together attributes describing how players arrive, engage, transact and use your products. Machine learning draws on this breadth to identify combinations of signals associated with future value.
The families below show the range available for model development. Each model uses a selection suited to your data, prediction target and validation results.
Session frequency and duration, return patterns and changes in engagement over time.
Product mix, content preferences and the way players engage across the available offering.
Channel, acquisition source and cohort context add another dimension to observed behaviour.
Deposit and withdrawal behaviour, transaction frequency and patterns across the player lifecycle.
Historical GGR, stakes and winnings provide context on the scale and shape of player activity.
Player tenure, activity timing, recency and evolving patterns across observation periods.
05 / Individually built for every operator
Every operator has a different player base, market and product mix. Each Gamblitude machine learning model is built individually and trained on that operator’s historical data.
Feature selection, training and evaluation are shaped around your operation, with the aim of achieving the strongest predictive performance your data can support.
Learn from your customers’ activity and how their revenue develops over time.
Build around the context in which your players engage, from acquisition channels to the products they use.
Align the target, currency and settlement treatment so forecasts and recorded outcomes can be compared consistently.
06 / Two ways to evaluate the forecast
The ability to prioritise players and the precision of each revenue estimate are different questions. Evaluate both against later outcomes and a relevant baseline.
For the same player population and review capacity, compare the model’s ranking with a baseline such as recent GGR. Examine how much subsequent GGR is represented by each selected group.
Compare predicted and realised GGR across cohorts and activity levels. Understand where estimates tend to run high or low. Individual player results can be volatile, so useful ranking and precise monetary estimates should be assessed separately.
The practical details
The forecast, its interpretation and how it fits your operation.
It produces a numerical estimate of the sportsbook GGR an eligible player is expected to generate over the next 30 days. The forecast is refreshed daily and can be used in player Attributes, Segments and dashboards.
This model has a defined 30-day horizon. Lifetime value covers a longer relationship and requires its own time horizon, assumptions and definition of revenue or profit.
GGR, NGR and profit are different measures. This model forecasts GGR. Bonus costs, taxes and other expenses require separate treatment according to your reporting definitions and the decision being made.
Sportsbook outcomes and player activity vary. A model may rank players usefully while its monetary estimate for an individual player remains uncertain. Assess forecast error as well as ranking quality, using the same GGR definition and outcome period.
Each daily estimate covers the next 30 days. Those forecast windows overlap. Do not add successive daily forecasts together to create a monthly revenue figure.
The eligible population is defined during model setup around the activity history and data available. Agree the inclusion criteria and check scoring coverage before using the results in operational Segments.
Early VIP Detection estimates whether a new player will meet an agreed high-value criterion. Player GGR Prediction estimates an amount of future revenue for eligible players with an activity history.
Player GGR Prediction and Player Churn Prediction provide different inputs for review. They can help define Segments together, but do not establish who will respond to an intervention. See the Predictive Models overview for the wider model family.
The GGR forecast provides value context. Choosing an incentive requires additional rules or modelling and evidence about the effect of the incentive itself. A predicted revenue amount is not a recommended spending limit.
Player GGR Prediction by Gamblitude
Show us how you rank players and which decisions depend on that view.
Let’s discuss where a 30-day GGR forecast could help.