Quantor
Trading · 10 min read

PAPER mode trading — the complete guide for risk-free crypto bot testing

Quantor · Solo-founded · · ← All posts

Most retail traders who burn their first account do it in the same way: they read about a strategy, type it into a bot product, click Start LIVE, and find out the strategy had a bug six hours later when their position is 14% underwater. They didn't lose money because the strategy was wrong. They lost money because they ran an untested binary against a real exchange account.

Quantor refuses to let that happen by default. Every new bot is provisioned in PAPER mode. Real candles from Binance feed the strategy, real signals fire, real entries and exits get recorded — but no order ever reaches the exchange API. The position is virtual. The P&L is virtual. The fees, the slippage, the timing — all simulated against the live tape.

This post is the long version of what that actually means: what PAPER simulates faithfully, what it deliberately approximates, how to read the resulting P&L without lying to yourself, and the seven-item checklist that decides when a strategy has earned the right to touch real money.

What "PAPER mode" actually is

Three terms get used interchangeably in crypto-bot marketing, and they mean three different things:

  • Backtest — feed historical candles into the strategy offline. You see what would have happened from Jan 2024 to Dec 2024. The market is not reacting to your trades because they never happened in real time. Useful for filtering strategies that are obviously broken; dangerous as a stand-alone signal because it's too easy to overfit to a known past.
  • Paper trade / forward test — feed live candles into the strategy as they arrive. Entries and exits are virtual but timed exactly when they would have fired in production. You can't peek at the future because the future hasn't happened yet.
  • Demo account — exchange-provided sandbox where you place "fake" orders against a synthetic order book. Useful for testing API wiring but the prices often diverge from mainnet.

Quantor's PAPER is the second one. The strategy gets the same live OHLC stream a LIVE bot would see — same Binance feed, same indicator state, same regime tags. When the strategy says BUY, the bot records a virtual entry at the candle close price minus a configurable slippage haircut. When it says SELL, virtual exit at close price plus haircut. Round-trip fees come straight from the active exchange's published taker rate.

The result is a P&L curve that tracks exactly what the strategy would have done in production — under the simplifying assumption that the strategy didn't move the market. For retail-sized notional (the kind a $19/mo plan will ever run), that assumption is reliably true on majors like BTC/USDT and SOL/USDT. For micro-cap tokens with thin books, it's false; we don't allow those in the PAPER allow-set either, so it never matters.

What PAPER simulates faithfully

Every component a LIVE bot uses is identical in PAPER. Specifically:

  • Strategy code — same JAR, same Java class, same parameters. The toggle is one boolean on the bot instance. There is no separate "PAPER strategy" branch with mock implementations.
  • Market data — same Binance kline stream that LIVE bots consume. Identical sample rate, identical history depth.
  • Regime classifier — same MarketRegimeDetector. If we'd refuse to start a LIVE bot in DANGEROUS regime, PAPER refuses the same way. See the regime-detector post for the exact mechanism.
  • 14 risk gates — every one of the pre-trade checks fires in PAPER. Symbol allowlist, notional ceiling, kill switch, plan permissions. Detail: the 14 risk gates post. The only check that differs is EXCHANGE_KEY_VERIFIED — PAPER doesn't need keys.
  • Decision tape — every BUY / SELL / HOLD decision is written to bot_decisions with the full reasoning. You can replay any bot's actions candle-by- candle six months later.
  • Audit chain — PAPER trades land in the same tamper-evident audit_chain table as LIVE ones. The chain doesn't distinguish virtual money from real money — it just records what the system did.

What PAPER deliberately approximates

Faithfulness has limits. The point of PAPER isn't to clone reality perfectly; it's to expose strategy bugs cheaply. We accept these simplifications:

  • Slippage = fixed haircut. We don't model the actual order book depth at the moment of decision. Real LIVE orders eat into the bid/ask spread; we charge a flat 8 basis points per leg to approximate. On BTC/USDT that's roughly right. On illiquid pairs (which, again, we don't allow) it would be optimistic.
  • Fills are instantaneous. Real LIVE orders may sit on the book for tens of milliseconds before executing. PAPER assumes fill at the candle close, which is close enough for 1-minute and 5-minute strategies but will overstate performance for sub-second scalpers. Quantor doesn't host sub-second strategies — minimum interval is 1m — so this rarely bites.
  • No partial fills. PAPER fills the whole notional or nothing. LIVE can leave a residual position if the book runs out; the residual stays open and the strategy has to handle it.
  • No exchange downtime. PAPER assumes Binance is always reachable. In LIVE, an API outage during a planned exit means the position stays open longer than the strategy intended. We can't simulate this honestly without simulating real failure injection, so we don't.
  • No funding-rate impact. Quantor only supports spot trading right now, so this is mostly moot. When perpetuals ship, PAPER will need a funding-cost model on top.

The cumulative effect of these simplifications: PAPER tends to overstate performance by a small amount — typically 1–3% per year on a strategy that trades a few times a day. If your PAPER curve is flat-to-slightly-down, you should expect LIVE to be worse. If your PAPER curve is +20% annualised, expect LIVE around +15% in normal conditions. That's the realistic gap, and it's why a PAPER curve barely above breakeven is a red flag, not a "close enough" signal.

How to read a PAPER P&L honestly

The dashboard shows a few numbers per PAPER bot. The temptation is to look at the headline P&L and stop there. Don't. Read in this order:

  1. Trade count. Fewer than 30 closed trades is statistically meaningless. A strategy with 4 trades and 100% win-rate tells you nothing — random walks win 50% of the time on a four-trade sample one in eight attempts. Run PAPER until you have at least 30 closed round-trips, ideally 100+.
  2. Maximum drawdown. How far did the virtual equity sink from its peak? If max DD is bigger than you could emotionally tolerate at 1× notional, either the strategy isn't yours or the notional needs to drop. A strategy you'd panic-stop in real life is not a strategy you should run live.
  3. Profit factor. Gross wins divided by gross losses. Above 1.5 is a positive signal; below 1.2 means you're surviving on noise. Anything above 3 on retail data is suspicious — overfitting somewhere.
  4. Sharpe ratio. Annualised return divided by annualised volatility. We compute it 365-day, zero-risk- free-rate. Above 1.0 is good for retail crypto. Above 2.0 is exceptional; verify the strategy isn't peeking at the future before trusting it.
  5. vs. buy-and-hold edge. Did the strategy beat just holding the same asset over the same period? If BTC went up 40% and your bot did +35% with daily trades, you under-performed the asset and added complexity. The public performance dashboard shows this comparison explicitly so users can see whether the runner bots are adding alpha or just trading for the sake of trading.

Anything else (total P&L, win-rate alone) is a vanity metric. A 70% win-rate strategy that loses 5× as much on losing trades as it wins on winning ones is a losing strategy with good optics.

The seven-item PAPER-to-LIVE checklist

You should not move a strategy from PAPER to LIVE until all seven are true:

  1. At least 30 closed trades in PAPER (90+ is better).
  2. At least 60 days of forward time. Running PAPER for one weekend says nothing about how the strategy behaves in different regimes. You want to see at least one full regime cycle (calm → volatile → calm).
  3. Max drawdown you'd accept on real money. If PAPER drew down 12% at one point and 12% on $1000 real money would have you closing the laptop in a panic — the strategy isn't ready, or the notional is wrong.
  4. Profit factor > 1.3 on the full sample. Below that you're trading for fees.
  5. Positive edge vs buy-and-hold in at least two of the last three months. If the bot trades all the time but underperforms HODL, the strategy is paying you less than just sitting on the asset would.
  6. You understand why each loser lost. Open the decision tape for the three biggest losing trades. Read the reasoning. If the loss makes sense in the strategy's logic (e.g. "VOLATILE regime, entered late, stopped out on noise") — fine. If the loss makes no sense, the strategy has a bug.
  7. You'd be comfortable explaining this strategy to a friend. If you can't articulate in two sentences what it does and why it should make money, you will second-guess it under stress and override it badly.

Missing even one of these? Run PAPER for another month. The exchange will still be there. Your account won't be if you skip ahead.

Why we ship PAPER as the default

Most retail bot products treat PAPER as a toggle some power users find buried in settings. Quantor inverts that — LIVE is the toggle, PAPER is the floor. A new bot can't be LIVE on day one even if the user clicks every button trying. The flow forces you through the 14-gate risk check before LIVE will activate, and one of those gates is HAS_SUFFICIENT_PAPER_HISTORY: the bot must have run PAPER long enough to produce a meaningful sample.

The friction is intentional. It's the same logic as the self-custody stance — architecture should prevent the failure modes users will reach for under stress. "Just click LIVE and see what happens" is the most common path to a blown account. We make it require deliberate effort instead.

What this means for the closed beta

Through 2026-07-01 — the closed-beta window — every new Quantor account starts on the FREE plan with one PAPER bot. That's enough to validate one strategy end-to-end across the seven-item checklist above. When public launch lands and paid plans turn on, beta participants keep 50% off any plan forever + first month free, and the existing PAPER history carries forward — you don't restart the clock when you upgrade.

The four runner bots powering the public performance dashboard are also PAPER-mode. Their numbers are the most honest possible benchmark for what a real strategy looks like in our environment — same code path, same risk gates, same regime tags, same fee model as a user-owned bot. The first signed monthly snapshot lands 2026-07-01 with one full month of real data.

Quantor PAPER is the cheapest education in algorithmic trading you can buy. It costs zero, takes thirty days, and saves you the $4,200 lesson the founder learned the expensive way. Apply for the closed beta — 100 seats, FREE PAPER for everyone, perks for the first wave.