Stop paying for subsidies that do not drive retention.
Template switching, new-player templates, cumulative and daily compensation, suppression, balance rescue. You once tuned these tables by hand and changed them after incidents. We price each interval by the retention it buys and recalibrate the full system in a loop.
Your problem
The ledger does not reconcile
Compensation adds, suppression subtracts, and net subsidy is a single number no one can break down by table, interval, or tag.
Parameters set by feel
Trigger probabilities, adjustment bounds, and suppression ranges were set manually once, then changed only when GGR went out of control or jackpots hit back to back.
Tags based on five spins
Segmentation looks only at the last five bets. Rescue timing is guesswork. No one has measured whether any of it actually moves D2.
What you get
Cost breakdown by table
Break net subsidy down by compensation interval, suppression range, new-player interval, and tag, with anomalies and ineffective rows listed.
A price for every interval
For each compensation and suppression interval, calculate the retention bought per unit cost: first from simulation, then from split tests after launch.
Recalibrated tables
A full configuration set with change log and rationale; tested in rollout before touching real money, then retuned monthly.
A ledger for operators
What the control system spent and what it bought, presented in the columns operators already read.
Evidence
769 runs: tables decide cost, players decide retention
In a simulation of a live dynamic control system, configuration explained the cost ledger, while player populations explained 96–98% of retention variance. The two were almost uncorrelated.
Read the study802 runs: the population decides the ledger direction
Across ten combinations of behavior-mechanism switches, chasing alone moved realized RTP from 92.7% to 103.6%. Jackpot suppression is absolute money; who is playing determines whether the system spends or profits.
Read the studyThe loop itself
Observe, Propose, Verify, Apply, Monitor, and the maker proposing a change is never the checker scoring it. Loop Engineering for Self-Improving Slot Agents.
Read the paperFirst step
- 1
Send us your configuration tables, or run them here
The browser version uploads nothing. If you send tables, we handle them under NDA and never use them across customers.
- 2
Receive a health-check report in two weeks
Cost breakdown by table, anomaly list, savings potential, and a recommended recalibration plan.
- 3
Recalibration and monthly tuning
A recalibrated system completed on rollout within two months, then retuned monthly as the player mix changes.
Common studio questions
Start with your next decision
Bring a table.Tell us your goal.
Want to reduce pre-submission iteration, evaluate a compensation budget, or validate an AI-generated design? Tell us which metric you want to optimize, and bring a PAR sheet, weight table, or draft. Together we’ll define the evaluation scope, deliverables, and next step.