Workstream 1 · Backtester outsource

Your backtester, run by Dvar, under your name

Dvar builds, hosts and runs backtesting for Dhan, Dhan Algos and Stratzy as a managed service with committed uptime and response times. The screens, the API and the models carry Dhan's name. Dvar stays in the background.

One engine, any brand

The same result card, embedded in three products. Only the skin and the model name change; the engine, the run hash and the charges table are shared, so a strategy shows the same numbers everywhere.

Colours are placeholders for the demo, not any company's brand guidelines.

Dhan · BacktestDaily · NSE

Buy TCS above the 10 EMA, sell when it comes back, with a 5% stop.

Ends at

Trades

Worst fall

Rule written by Artham Rule-1 · numbers by the engine · run hash on every result

What Dvar delivers

Layer	What you get	In this demo	To build
Engine	Deterministic backtest engine as an API, a browser SDK and an embeddable screen in your design. Every result carries a run hash and the engine, data and charges versions.	Daily bars, long-only NSE delivery, EMA, SMA and RSI rules	Minute bars, F&O (lot sizes, expiries, margin, chains), shorting, more indicators
Data	Ingest of your own historical prices, corporate actions, holidays and a versioned table of your charges.	Demo prices only	Your data or a licensed vendor; nightly refresh with reconciliation
Verification	Five checks on every listed strategy and a nightly re-run of the whole catalogue.	10 sample strategies (Verify tab)	Dhan Algos and Stratzy listings, verdict history, alerts on change
Plain words	A model turns a sentence into a strict rule; the engine does all the maths.	Working, with a built-in reader as fallback	Hindi and Hinglish, F&O vocabulary, your own branded model
Hosting	Daily runs in the trader's browser; heavy runs on workers in AWS Mumbai or your own cloud account.	Demo hosting	Production workers, queue, result storage, audit log
Operations	On-call, incident reports, monthly service review, runbooks, full source access for you.	None	Agreed in the contract

One call from your app

The proposed API. The plain-words step and the engine sit behind one endpoint; your app never sees Dvar's name.

POST https://backtest.<your-domain>/v1/runs
Authorization: Bearer <key>

{
  "text": "Buy TCS above the 10 EMA, sell when it comes back, with a 5% stop",
  "capital_inr": 100000,
  "from": "2019-01-01",
  "charges": "dhan-delivery-2026-10"
}

200 OK
{
  "rule": { "universe": ["TCS"], "entry": "close > ema(10)", "exit": "close < ema(10) or stop 5%" },
  "rule_model": "Artham Rule-1",
  "end_value_inr": "<engine output>",
  "closed_trades": "<engine output>",
  "run_hash": "sha256:…",
  "versions": { "engine": "1.4.0", "data": "nse-daily-2026-10-21", "charges": "dhan-delivery-2026-10" }
}

Custom models you can show as yours

We post-train small models on Dhan's data and tasks, and Dhan ships them under its own name, for example as members of the Artham family. Raise owns the weights trained on its data; Dvar does not reuse them for anyone else. Each ships with a model card and an eval report in Raise's name. The names below are placeholders.

Artham Rule-1

Sentence to strategy rule

Short structured output a schema can check; the first job for a tuned Artham

Proof: Rule-writer eval on the Post-training tab

Artham Explain-1

Plain-words summary of a run

Written only from engine output, so every figure is checkable

Proof: Figures-from-source check, the same one the assistant uses

Artham Support-1

Support intent and answers

Shared with Dhan Assist; high volume, Hinglish heavy

Proof: The release gate on the Evals tab

Proposed service levels

Starting points for negotiation, not commitments yet.

Same input gives the same output	100% of runs, checked by run hash
Daily-bar single-stock backtest, 95th percentile	Under 5 seconds
Minute-bar or F&O backtest, 95th percentile	Under 60 seconds (to confirm once data is sized)
Service uptime	99.5% in market hours
Data refresh	Prices and charges updated by 7:00 am each trading day
Incident response	30 minutes in market hours, next business day otherwise

How it rolls out

Weeks 0–2
Set-up
Data licence and access, 20 real strategy definitions, where it runs
Weeks 2–6
Verified backtests
Every daily-bar listing re-verified on your data, nightly re-run
Weeks 6–12
Backtester inside Dhan
Plain-words backtester on one surface, plus Artham Rule-1
Weeks 12–20
Intraday and F&O
Minute data, options and margin rules
Week 20 on
Managed service
Service levels in force, monthly review

The API, service levels, model names and timings on this tab are proposals from the project plan, not a running service. The engine itself is real: it runs on the Try it and Verify tabs in your browser.