Rating of releases by the number of launches by players
1) What is "start" and what metrics to count
Launch-The user's game opening event (not to be confused with the first spin). For correctness, consider only sessions with minimum retention: ≥10 seconds or ≥1 spin.
TL (Total Launches) - total number of launches for the period. Reflects the interest and stickiness of the interface, but is sensitive to spam.
UL (Unique Launchers) - the number of unique players who opened the game during the period. Main coverage indicator.
LPU (Launches per User) = TL / UL. Shows how often a slot is returned within a period.
New vs Returning - shares of new and returning players.
R7/R30 - return after 7/30 days among those who launched a slot in D0-D3.
ASL (Avg Session Length) - average session length (minutes or backs).
Demo/Real split - shares of "demo" and "for money." Be sure to post.
2) Data sources
Internal casino analytics: raw event logs, most reliable.
Aggregators/platforms: summarize telemetry for many operators (unification of identifiers is needed).
Provider panels: Give UL/TL on their integration network.
Public indirect proxies: tournament tables for the new product, views/online from streamers, "popular" feeds. Use only as auxiliary signals.
3) Periods and normalization
Observation windows: 7D (hype after release), 30D (medium-term interest), 90D (stability).
Release maturity: compare games with the same "age" (for example, D1-D30 from the launch date).
Smoothing: Exponential (EWMA) for TL and UL to keep ratings from jumping from one-off promos.
Seasonality: adjust peaks (holidays, tournament weeks) with coefficients or take them to "off-rating" campaigns.
4) Antifraud and data cleaning
Exclude auto clicks and spin-free ping runs.
Frequency filter: max X runs per minute per device/IP.
The exception to the "farms" of working out bonuses: strange patterns of ASL≈0, LPU is abnormally high at zero rates.
Split Demo/Real money at event and report level.
Deduplication - single player on multiple devices - hashed-ID + device graph is solved.
5) Basic rating formula (30-day window)
Normalize metrics to z-score or min-max within the novelty pool for the period. Example of scales:
6) Rating segmentation (do subtasks)
By region/currency: AU/AUD, EU/EUR, CA/CAD, etc.
By device: Mobile Portrait/Mobile Landscape/Desktop.
By game type: Megaways/Clusters/Hold & Win/Classic
By traffic source: organic, promo, tournament.
Each segment has its own rating and its own winner; the overall "world" top is only useful for PR.
7) How to read the player's rating
Top 10 by UL (30D) - lists of "what most play." Suitable for quick selection.
LPU and ASL: high LPU at average UL = niche "magnet" (often launched by those who have entered).
R7: if high, the slot "lives" longer than one evening.
Demo/Real split: a big bias in the demo is a sign of hype or a complex economy; worth starting with a free test.
8) How to use casino/venue rating
Showcase: raise UL (7D) leaders in banners, UL (30D) in "recommended," UL growth in "on the rise."
Promo: Give freespins where UL is high but LPU is low (motivate second session).
Purchases/tournament schedule: put events on games with positive TL-growth.
RTP version control: tops "sag" with lowered configurations - check game cards.
9) Constraints and traps
Disparate definitions of "run" across sources. Fix the SLA of the definition: "≥1 spin or ≥10 sec."
Stocks and streams distort TL - take them into a separate layer and remove the impact.
Novelty shift: In the first week, almost any release is at the top, so compare "age to date."
Multi-account/bots: without anti-fraud, the rating is invalid.
Demo ≠ revenue: launch popularity is not equal to profitability for the player or casino.
10) Recommended structure of the rating table (30-day window)
11) Mini calculation process (practice per quarter)
1. Collect raw events from the last 90 days and mark releases D0-D90.
2. Apply filters: "≥10 sec or ≥1 spin," anti-bots, deduplication.
3. Count UL, TL, LPU, R7, ASL separately for Demo and Real.
4. Normalize the metrics in the novelty pool, calculate the Score by weights.
5. Build three ratings: 7D, 30D, 90D + stability table (coefficient of variation of rank).
6. Separately release segments: AU (AUD), Mobile-Portrait, High-Vol.
7. Fix the method: definitions, windows, weights, cut date - so that the reports are comparable.
12) Interpretation cases
High UL, low LPU: massive curiosity, weak "binding" - it makes sense to try in the demo.
Medium UL, high LPU and R7: niche, but strong release - a good candidate "for the main slot" for long sessions.
Fast TL-growth at low UL: the game was "rocked" by promo - check the conditions and RTP configuration.
Sagging R7 with UL growth: aggressive hype; without bonuses, the retention is weak - suitable for a short test.
Result
Rating by the number of launches is useful if you count it according to a single definition of "launch," separate demo/real sessions, remove the noise of the promo and compare slots of the same "age." For the player, he answers the question "what is really launched," for the casino - where to focus the showcase and promo. Use three slices (7/30/90 days), segmentation (region/device) and a transparent Score formula - this is how the rating will become a working tool for choosing new products, and not a hype showcase.
Launch-The user's game opening event (not to be confused with the first spin). For correctness, consider only sessions with minimum retention: ≥10 seconds or ≥1 spin.
TL (Total Launches) - total number of launches for the period. Reflects the interest and stickiness of the interface, but is sensitive to spam.
UL (Unique Launchers) - the number of unique players who opened the game during the period. Main coverage indicator.
LPU (Launches per User) = TL / UL. Shows how often a slot is returned within a period.
New vs Returning - shares of new and returning players.
R7/R30 - return after 7/30 days among those who launched a slot in D0-D3.
ASL (Avg Session Length) - average session length (minutes or backs).
Demo/Real split - shares of "demo" and "for money." Be sure to post.
2) Data sources
Internal casino analytics: raw event logs, most reliable.
Aggregators/platforms: summarize telemetry for many operators (unification of identifiers is needed).
Provider panels: Give UL/TL on their integration network.
Public indirect proxies: tournament tables for the new product, views/online from streamers, "popular" feeds. Use only as auxiliary signals.
3) Periods and normalization
Observation windows: 7D (hype after release), 30D (medium-term interest), 90D (stability).
Release maturity: compare games with the same "age" (for example, D1-D30 from the launch date).
Smoothing: Exponential (EWMA) for TL and UL to keep ratings from jumping from one-off promos.
Seasonality: adjust peaks (holidays, tournament weeks) with coefficients or take them to "off-rating" campaigns.
4) Antifraud and data cleaning
Exclude auto clicks and spin-free ping runs.
Frequency filter: max X runs per minute per device/IP.
The exception to the "farms" of working out bonuses: strange patterns of ASL≈0, LPU is abnormally high at zero rates.
Split Demo/Real money at event and report level.
Deduplication - single player on multiple devices - hashed-ID + device graph is solved.
5) Basic rating formula (30-day window)
Normalize metrics to z-score or min-max within the novelty pool for the period. Example of scales:
- Score = 0. 55·UL\_norm + 0. 25·TL\_growth\_norm + 0. 15·LPU\_norm + 0. 05·R7\_norm
- where TL\_ growth is the increase in TL 30D vs previous 30D (or vs first week) to encourage those gaining momentum.
- Alternatives: enable ASL\_ norm (0. 05–0. 10) if depth of sessions is important.
6) Rating segmentation (do subtasks)
By region/currency: AU/AUD, EU/EUR, CA/CAD, etc.
By device: Mobile Portrait/Mobile Landscape/Desktop.
By game type: Megaways/Clusters/Hold & Win/Classic
By traffic source: organic, promo, tournament.
Each segment has its own rating and its own winner; the overall "world" top is only useful for PR.
7) How to read the player's rating
Top 10 by UL (30D) - lists of "what most play." Suitable for quick selection.
LPU and ASL: high LPU at average UL = niche "magnet" (often launched by those who have entered).
R7: if high, the slot "lives" longer than one evening.
Demo/Real split: a big bias in the demo is a sign of hype or a complex economy; worth starting with a free test.
8) How to use casino/venue rating
Showcase: raise UL (7D) leaders in banners, UL (30D) in "recommended," UL growth in "on the rise."
Promo: Give freespins where UL is high but LPU is low (motivate second session).
Purchases/tournament schedule: put events on games with positive TL-growth.
RTP version control: tops "sag" with lowered configurations - check game cards.
9) Constraints and traps
Disparate definitions of "run" across sources. Fix the SLA of the definition: "≥1 spin or ≥10 sec."
Stocks and streams distort TL - take them into a separate layer and remove the impact.
Novelty shift: In the first week, almost any release is at the top, so compare "age to date."
Multi-account/bots: without anti-fraud, the rating is invalid.
Demo ≠ revenue: launch popularity is not equal to profitability for the player or casino.
10) Recommended structure of the rating table (30-day window)
| Parameter | Description |
|---|---|
| Rank | Score position |
| Game/Provider | Name and Studio |
| Release date | D0 |
| UL (30D) | Unique Players |
| TL (30D) | Total Runs |
| LPU | TL/UL |
| R7/R30 | Return |
| ASL | Average Session Duration |
| Demo/Real | Shares% |
| TL Growth | Gain to Last Window |
| Top Geo/Device | Top Region and Device |
| Rank change | Δ positions vs previous report |
| Notes | Promo/Tournaments/Features |
11) Mini calculation process (practice per quarter)
1. Collect raw events from the last 90 days and mark releases D0-D90.
2. Apply filters: "≥10 sec or ≥1 spin," anti-bots, deduplication.
3. Count UL, TL, LPU, R7, ASL separately for Demo and Real.
4. Normalize the metrics in the novelty pool, calculate the Score by weights.
5. Build three ratings: 7D, 30D, 90D + stability table (coefficient of variation of rank).
6. Separately release segments: AU (AUD), Mobile-Portrait, High-Vol.
7. Fix the method: definitions, windows, weights, cut date - so that the reports are comparable.
12) Interpretation cases
High UL, low LPU: massive curiosity, weak "binding" - it makes sense to try in the demo.
Medium UL, high LPU and R7: niche, but strong release - a good candidate "for the main slot" for long sessions.
Fast TL-growth at low UL: the game was "rocked" by promo - check the conditions and RTP configuration.
Sagging R7 with UL growth: aggressive hype; without bonuses, the retention is weak - suitable for a short test.
Result
Rating by the number of launches is useful if you count it according to a single definition of "launch," separate demo/real sessions, remove the noise of the promo and compare slots of the same "age." For the player, he answers the question "what is really launched," for the casino - where to focus the showcase and promo. Use three slices (7/30/90 days), segmentation (region/device) and a transparent Score formula - this is how the rating will become a working tool for choosing new products, and not a hype showcase.
