Gaming revenue
Formula
Stakes − winnings
Behaviour forecasting · Live inference
Understand the pattern.
Anticipate what follows.
Forecast deposits and activity, inspect uncertainty, and identify what needs a closer look.
Connecting to inference API…
Behaviour forecast
TimesFM 2.5Generate a history to see observations, forecast and its interval.
Point to the chart, or focus it and use the arrow keys to inspect a period.
ObservedMedian forecastQ10–Q90
Q10–Q90 is a nominal central 80% prediction interval, not accuracy. Forecasts are not a diagnosis.
Observed changes · historical reconstruction
Seven historical predictions, each excluding its target day or week. Weekly checks describe forecast quality and do not trigger RG rules.
| Period starts | Actual | Q90 | Excess floor | Above |
|---|
Operational outlook
Selected metricExplore future payment volume and activity. Compare the forecast against simple historical baselines.
For infrequent deposits, daily medians can be close to zero. Their sum is not expected cash flow, and a low median alone does not predict churn.
Does the forecast add value?
Seven reconstructed forecasts of one day or one week on synthetic history. This small diagnostic is not evidence of commercial effectiveness.
Forecast by period · exportable values
| Period starts | Q10 | Median | Q90 |
|---|
Cohort economics
30-day scenario · EURA separate financial scenario for players acquired in one period. Enter cohort totals to calculate revenue and acquisition economics.
Revenue after agreed deductions
Formula
GGR − bonuses − taxes − content
Acquisition cost per player
Formula
All acquisition spend / cohort size
Per active player · 30 days
Formula
NGR / active cohort players
Scenario contribution per acquired player
Formula
(NGR − payment − servicing) / cohort size × months
Illustrative inputs. Edit the assumptions below.
Edit financial inputs
Use the same cohort and 30-day observation window. LTV assumes the current monthly contribution continues for the entered number of months; no churn model or discounting is applied.
How to interpret these metrics
NGR depends on accounting policy. This scenario deducts bonuses, gaming taxes and content fees; payment and servicing costs are deducted next to calculate contribution. ARPU uses active players, while CAC and scenario LTV use all acquired cohort players. Fixed OPEX is not included. These figures are calculated from your inputs, not predicted by TimesFM.
Reference: stakes and winnings in GGY ↗Review recommendation
Decision LayerAwaiting evidence
An explainable policy combines eligible evidence. It does not create another risk probability.
Why this recommendation?
FARO · profile context
1.0.0FARO provides separate behavioural context and does not adjust the TimesFM forecast. The event mapping currently covers 4 of 60 base features. Technical scores are excluded from player-review rules until mapping verification is complete.
—
What influenced the model score
Contributions are in raw margin units, not percentage points. Missing values can also affect the model output.
Inputs, quality & provenance
—Public demo accepts only a content-equivalent JSON roundtrip of your own server-generated dataset. Operator data requires a private installation and approved policy.
180 FARO inputs · schema from the artifact
| Feature | Value | Status |
|---|
Versions & reproducibility
How to read this demo
Events, not a hand-drawn chart
Presets generate deposits, bets, winnings and sessions. Actual daily series are built from these synthetic events, including zero-activity days and missing sources.
Future activity and uncertainty
TimesFM 2.5 forecasts four metrics by day or by complete week. The demo supports up to 1,024 days; long horizons are experiments. Weekly sums are forecast directly, not formed by adding daily quantiles. Q10–Q90 shows model uncertainty.
One cohort. Explicit assumptions.
GGR, NGR, CAC and ARPU are calculated from the financial inputs. LTV extrapolates monthly contribution under an editable assumption; it does not predict retention or convert deposits into revenue.
Compare against a simple baseline
Inspect historical forecast errors and interval coverage below the chart. Commercial value needs a longer holdout on operator data. FARO remains a technical demonstration while feature mapping is incomplete.
Inference source & model notices ↗
Commercial evaluation: compare forecast error, interval coverage and operational decisions on a time-based holdout. A deposit forecast is not a revenue forecast. RG evidence is used for player protection.