Quickstart

This is the shortest honest path from a clean install to a strategy running in paper mode. It walks the real pipeline in order: get data, backtest a builtin strategy, run it through the gauntlet in the lab, then promote it to paper. Every UI route named here is real.

The order is deliberate. A strategy cannot reach paper without surviving the robustness battery, and the battery cannot run without data. There are no shortcuts, and that is the point — the discipline is the product.

Forven is a research tool. Backtest and paper results describe the past and a simulation; they are not predictive, and nothing here is financial advice. Paper trading (optionally validated on the HyperLiquid testnet) is the supported mode — no real capital is at risk during this walkthrough.

Before you start

You need:

• Forven cloned and running. See Install if you have not done that yet (clone the repo, run start_all.ps1 on Windows or start_all.sh on macOS/Linux).
• The web UI open in your browser at http://127.0.0.1:5173.
• A few minutes — the gauntlet runs real walk-forward analysis, so it is not instant.

A note on names. The public site uses simplified words; the app uses the real pipeline stages. This guide uses the real names:

• screen is quick_screen — the overfitting triage gate.
• gauntlet is gauntlet — the robustness battery.
• candidate is paper — paper trading on the live feed with no real orders. We call it paper.

What you'll build

By the end you will have one bundled strategy (for example rsi_momentum) that has a backtest on a real dataset, a robustness verdict from the gauntlet, and a session running in paper mode that you can watch on the live feed. No real money moves at any point.

Steps

1. Get market data (/data)

Open the Data Manager at /data. Backtests need OHLCV candles, so populate at least one dataset before anything else.

1. Click Download Data (or switch to the Datasets tab).
2. Choose a source: CCXT/Binance, Binance Vision, Polygon, Yahoo, or a CSV upload.
3. Specify a symbol and timeframe — for example BTC/USDT on 1h — and a date range.
4. Click Fetch to queue the download, and watch the data-fetch status banner.
5. Switch to the Maintenance tab, open the Backfill Plan, and click Catch up now if there are gaps to fill.

What you'll see: the Overview tab updates its KPIs (datasets count, total rows, latest download), and the dataset appears in the coverage matrix with a quality score. Backfill is asynchronous — counts may lag a few seconds while the plan re-scans.

For the full data workflow, sources, and the coverage matrix, see Managing market data.

2. Backtest a builtin strategy (/backtest/new)

Open the Backtest Studio at /backtest/new. This is where a strategy meets history.

1. Pick a strategy source. For your first run, select a prebuilt strategy from the list — Forven ships 75+ builtins (rsi_momentum, bollinger, keltner, macd, ema_cross, vwap, and more).
2. Set the symbol, timeframe, and date range. Match the dataset you just downloaded.
3. Edit parameters if you want, or leave the defaults.
4. Click Preview to dry-run the signals without executing.
5. Click Submit to queue the backtest.

What you'll see: the result opens with an equity chart, a trades table, and a metrics panel. The engine splits your data 70/30 into in-sample (IS) and out-of-sample (OOS), applies leverage (default 3x), fees (default 4.5 bps), and slippage (default 2.0 bps) as round-trip costs, and computes Sharpe, Sortino, max drawdown, profit factor, win rate, and per-side / per-regime breakdowns.

Read the OOS numbers, not the IS numbers. In-sample metrics are optimistic by construction — they describe data the signals were tuned on. OOS is the unseen tail and the only honest read. Metrics also carry reliability flags: Sharpe is flagged unreliable below 20 trades, and annualized return is flagged below a 3-month window, so a glittering number on a 25-day run is not real.

For the studio's three modes (prebuilt / visual / custom code) and every metric, see Backtesting a strategy and Metrics.

3. Run the gauntlet in the lab (/lab)

A single backtest is not evidence. Open the Strategy Lab at /lab, find your strategy, and open its detail page at /lab/strategy/[id].

1. From the strategy detail, submit it to the Gauntlet.
2. The gauntlet runs the robustness battery: walk-forward analysis, Monte Carlo, parameter jitter, cost stress, and regime split.
3. Optionally optimize parameters first — grid search samples up to 200 combinations (Latin Hypercube) and returns the top 5 by fitness, which you can then validate with walk-forward.
4. Read the robustness verdict in the optimizations tab and the PromotionReadiness checklist.

What you'll see: walk-forward returns a verdict of PASS or FAIL. By default it runs 5 folds at 70% IS / 30% OOS per fold, and passes only if aggregate OOS Sharpe degrades less than 50% from IS and there are at least 5 OOS trades. Degradation above 50% is the signature of overfitting — the strategy learned the training set, not the market. A FAIL here is the system doing its job. Kill it and move on.

For the full battery and what each test stresses, see The gauntlet. For the lab as a development hub, see The strategy lab.

4. Promote to paper

If — and only if — the readiness checklist is satisfied, promote.

1. On the strategy detail page, click Promote.
2. Provide a reason if this is a manual override.
3. The promotion is queued as an approval in /approval.
4. An operator reviews and approves it; the strategy transitions to the paper stage.

What you'll see: the promotion gate reads fitness from policy. A composite fitness score of >= 60 makes a strategy paper-eligible (the score weights Sharpe 30%, win rate 20%, profit factor 20%, max drawdown 15%, and trade count 15%). The lean paper gate is intentionally looser than the strict paper-to-live gate — paper costs nothing but information, so the bar to enter it is lower than the bar to risk capital.

Once promoted, watch the session at /paper-trades. Paper mode runs your strategy against the current price feed with no real orders sent — same signals as live, zero capital. Paper PnL will not reconcile with any exchange because nothing was routed there; that is expected.

For the simulation model and session controls, see Paper trading.

What happens next

Paper is not the finish line — it is the proving ground. A strategy earns its way toward live consideration by accumulating honest paper PnL over time. Paper (optionally validated on the HyperLiquid testnet) is the supported mode; there is no supported live/mainnet path, so you can run this loop freely while the safety rails stay on.

The autonomous side of Forven — the research daemon inventing and testing hypotheses on its own — does the same loop at scale without you driving each click. You have just done by hand what the system does continuously.

Caveats

• Data first, always. A backtest on a thin or gappy dataset produces confident nonsense. Check the coverage matrix before you trust a result.
• OOS is the truth. Treat in-sample metrics as a sanity check, never as a forecast.
• Short windows lie. Reliability flags exist because annualized figures explode on tiny samples. Respect the flags.
• A FAIL is a feature. Most strategies do not survive the gauntlet. That filter is the entire value.

Related

• Managing market data
• Backtesting a strategy
• The strategy lab
• Paper trading