QuantilanSimulated results, no real orders. A backtest is not a live track record. Past results do not guarantee future results. Nothing here is financial advice.
1. What is covered
Two different methods are used, and they are not mixed:
- Crypto strategies (Q-Mix, BB, RSI, IMP): a multi-year backtest on Binance USDⓈ-M perpetual futures, 2022 to 2026. Sections 2 to 7.
- Tokenized assets (Q-RWA, Q-ETF, IMP-RWA): a replay of the current rules on Binance TradFi perpetuals since each contract's listing. They have too little history for a multi-year test. Section 8.
The Signal Atlas app also runs public virtual accounts that trade the same rules forward in time. They are described in section 9.
2. How the simulation runs
The test is sequential: it walks forward through 15-minute candles, one bar at a time, using only what was known at that bar. A signal is produced on a closed bar and filled at the open of the next M15 bar, never at the price that triggered it. Positions, stops and take-profit levels are tracked bar by bar, and the account balance is updated by every closed trade.
The backtest runs on the same production engine that generates the live signals, not on a separate research copy of the rules.
3. Costs: fees and slippage
Every published result includes adverse execution cost. It is not a perfect fill at the signal level.
| Component | Rate | How it is applied |
|---|---|---|
| Taker fee | 0.05% | Per side: on entry and on exit |
| Entry slippage | 0.05% | Market fill versus the open of the next M15 bar after a closed-bar signal |
| Stop / take-profit slippage | 0.03% | Against the trade, relative to the stop or take level (not the mid price) |
| Funding | not modeled | An explicit assumption: funding payments are not included |
For reference, live telemetry over 30 days showed about 0.035% entry and 0.034% exit slippage, so the model is slightly conservative.
4. Position sizing and risk
Every strategy trades with 5× leverage. BB and RSI use isolated margin per coin, so one coin never draws on another coin's margin, and BB, RSI and IMP run with an adaptive risk controller.
| Strategy | Universe | Margin per position | Loss at the stop |
|---|---|---|---|
| Q-Mix | 19 coins on one shared account, Q-INDEX 4H rule | 5% of the balance | about 2% of the account (5% × 5× × the 8% stop) |
| BB (BearBull) | 16 coins | 5% | 0.5–0.75% of the account |
| RSI (Advanced RSI) | 18 coins | 5% | 0.5–0.75% of the account |
| IMP (Impulse Retest) | BTC, SOL, ETH | 40% / 30% / 10% | about 6% / 3.75% / 1.75% of the account |
Each strategy has its own account that starts at $10,000. A multi-strategy portfolio therefore means several separate accounts, one per strategy, not one account running everything.
5. Annual reset
Each year is a separate run that starts again from $10,000 on 1 January. Profits are not carried into the next year, so the yearly figures never compound. Two consequences follow:
- Yearly returns can be compared with each other, because every year has the same starting base.
- A sum of years is not the result of an account held for five years. It models a book that is reset to the same size each year.
Passive benchmarks are measured on the same convention. Over 31 Dec 2021 to 12 Oct 2026, the sum of yearly returns was about +199% for spot Bitcoin and +66% for the S&P 500 Total Return index. A buy-and-hold with no annual reset compounds to about +77% for Bitcoin and +75% for the S&P 500 TR. Bitcoin's yearly-sum figure is higher because the 2022 fall of about 64% shrinks the base for the later gains. The reset figures are the fair comparison for a reset strategy.
6. What the metrics mean
| Metric | Definition |
|---|---|
| Trades | Number of closed trades in the selected year |
| Win rate | Share of closed trades that finished in profit |
| Profit factor | Gross profit divided by gross loss. Above 1.0 the strategy earned more than it lost |
| Max drawdown | The largest fall of the account from a peak to a following low, inside that year |
A high win rate says little on its own: a strategy can win often and still lose money if the losers are larger. Read win rate together with profit factor and drawdown.
7. Coverage and known limits
- 2022 uses fewer coins. Not every coin existed on the exchange then. A coin trades only from its listing, and only after it has 12 days of candles. The full coin lists apply from 2023. The number of coins trading in each year is shown under each chart. Year 2021 is not published.
- Coin tiers are the average daily turnover over the last 365 days on Binance USDⓈ-M: Tier 1 at or above $1B a day, Tier 2 from $100M to $1B, Tier 3 below $100M.
- IMP is not a clean out-of-sample test. Its parameters were chosen by the worst year, with 2022 and 2026 inside the selection as extreme regimes. It also trades rarely, about 50 to 66 times a year, so its statistics rest on a much smaller sample than BB or RSI.
- The current year is partial and partly live. 2026 runs to the date shown under the chart (25 Oct 2026 at the last update). Up to 30 Sep the figures are a backtest. Since 1 Oct 2026 the strategies run in a live test on virtual accounts, so results from that date are no longer a backtest. The charts mark the switch with a dashed line.
- Funding is not modeled (section 3).
8. Tokenized assets: a replay, not a backtest
Tokenized U.S. and Korean stocks and leveraged ETFs trade on Binance TradFi perpetuals, which are new. The stocks start in February and the leveraged ETFs in May, so there is no multi-year history. For these strategies the current rules are replayed on Binance prices from each contract's listing to the date under the chart. Q-RWA and Q-ETF use split-adjusted prices. These are virtual accounts, and the section is marked Experimental on purpose: a few months of data cannot show how a rule behaves across different market regimes.
9. Public virtual accounts and the live test
Inside Signal Atlas, the strategies trade on public virtual accounts. Each decision is logged with its time, price and reason, so anyone can follow an account forward in time and compare it with the backtest. No real orders are placed.
Since 1 October 2026 the strategies run in a live test on these accounts. Results from that date are forward results, not a backtest, and the charts mark the switch with a dashed line.
10. What a simulation cannot show
- Live fills can differ from the modeled slippage, especially in fast markets or on thin coins.
- Exchange outages, API limits and delayed signals are not part of the simulation.
- Funding payments and any change in exchange fees are not included.
- Markets change. A rule that worked in 2022 to 2026 may not work later.
This is why results are shown with their assumptions, why the tokenized-asset section is marked as experimental, and why the trading agent lets you set your own risk limits on your own exchange account.
See the numbers this article describes in the Strategies section of the home page. Open Year by year on any card to read a single year.