A working model of India's criminal courts that asks one question of every rule in the Sentencing Guidelines and the Satyug Bill: does it make telling the truth cheaper than lying and waiting? It prices the accused's choice, the witness's, the co-conspirator's and the complainant's, and follows them through the court queue.
| Today | With Satyug | |
|---|---|---|
| Tell the truthadmit and plead | A discount no one can calculate: there are no sentencing guidelines. | A published discount, known in advance, larger the earlier you plead.The door that pays |
| Stay silentmake the State prove it | Your right. | Still your right, and never held against you. |
| Lie and waitdeny, delay, let witnesses turn | Almost free. Perjury is rarely prosecuted, witnesses turn, and delay wears the case down.The door that pays | A real, quick, separate punishment, added on top. The recorded statement survives a hostile witness. |
| So the accused | denies and waits. 99.9% of cases go to full trial. | pleads early. The case ends in months, not years. |
The simulator is open to consultation participants. Enter the password you were given.
Start with the simulator, or read how the model was built first.
The problem
India's courts hold 5.9 crore pending cases.1 About one criminal case in a thousand ends in a plea of guilt;2 in comparable systems, nine in ten do,3 and settle ten to twenty times faster.
The usual answer is more judges. The Economic Survey's own model holds behaviour fixed and asks how many judges it takes to clear the queue. It never asks why almost every case goes to a full trial.
The answer is incentives. With no published sentencing scale, the reward for pleading cannot be calculated; with perjury almost never prosecuted, lying costs nothing; and every year of delay makes conviction less likely. Under those rules, denial is the rational move, and the queue sustains itself.
What the model finds
These figures are not typed in. They are computed from the model each time this page loads, with the same code the simulator runs.
of cases end in a plea under today's rules, against 0.11% observed: the model reproduces the status quo with no target set.
of all cases end in a truthful plea under the full package, and 88% of convictions.
is the longest trial wait at which truth still beats delay. Beyond it, waiting wins again, which is why the docket reforms matter.
wait for trial on the fresh list by year six, against 8.4 years if nothing changes.
Why the numbers hold
The model was built to be audited by people who do not take it on trust: judges, draftsmen, economists and sceptics. Each of these pages answers a different objection, and all of them are open to read without the password.
Sentencing commissions in the US, Minnesota and England model every rule before adopting it. Cartel leniency and auction design show where mechanism design has worked. The method note sets out the precedent, with sources.
Read the method and precedentEach equation the model runs, in the order it runs them, with a worked example at the scenario you have set and an explanation on every symbol.
See the mathematicsEach input is marked statutory, empirical or assumed, with its source. The status quo is checked against published plea, clearance and conviction rates without being fitted to them.
Check the assumptionsThe headline is reported as a range across the bounds of every assumed constant. A heat map shows where truth wins and where delay wins; a tornado chart shows which assumption matters most.
Test the robustnessAll 75 items of the amendment schedule, the sections of the BNS, BNSS and BSA they amend, linked to the official text on India Code, and what the model does with each.
Browse the levers and lawA plea of not guilty never raises the sentence; custody cannot coerce a plea; police cannot start a perjury case alone. Each safeguard is checked on every run and shown on the simulator.
See the safeguards auditIn good company
The draft Guidelines ask the Sentencing Authority to apply game-theoretic modelling to plea bargaining, witness cooperation and pendency (s.18(1)(a)(v)). This simulator is a working prototype of that duty, built on methods with a track record.
Listed as precedent for the method. None of these bodies has reviewed or endorsed this simulator.
Inside the model
Every page reads the same scenario. Change a setting on the simulator and the docket, the prisons and the mathematics follow.