A historical backtest, a paper-trading exercise and a prospective forward observation answer different questions. None alone demonstrates that a strategy will make money, and none should be described as a live trade unless an actual order was submitted and its execution independently recorded. Treat the sequence as a way to expose assumptions one at a time, not a staircase that guarantees readiness.
Backtest: what would these rules have done on these records?
A backtest applies specified rules to historical data. It is useful for checking whether the rules execute as intended, estimating results under a declared model, finding data errors and identifying obvious failure cases. The result depends on what the dataset contains and on assumptions about signal timing, entry, exit, costs, funding and order execution.
For example, if a rule needs the final close of an hourly candle, that information is not available at the candle’s open. Entering earlier in the same bar can introduce look-ahead bias. A backtest that uses hourly OHLC records cannot prove a one-second queue fill, even if its output reports precise decimal returns. Synthetic data can test software behavior, but it cannot stand in for observed exchange history. Record the provider, instrument identifier, timestamps, coverage, missing intervals, rule version and every material fill assumption.
Use a chronological evaluation window that was not used to select settings, and keep track if you later examine it. Multiple parameter searches, asset choices and date ranges create more opportunities to find a chance result; Bailey and colleagues’ backtest-overfitting study on the publisher’s page explains why repeatedly selecting the strongest candidate can make a backtest look more persuasive than its unseen evidence warrants. Report the baseline, trade count, drawdown and cost sensitivity alongside the attractive metric. A clean holdout helps answer a narrow question about reused history; it does not ensure that future conditions resemble the past.
Paper trading: what does a practice workflow reveal?
Paper trading records hypothetical decisions without sending an order that changes market exposure. Depending on the tool, it can help reveal whether a participant follows written rules, whether the interface supports the intended workflow and how a chosen fill model affects the ledger. A simulated order may be filled at a candle price, a live quote, or another model. Those are different forms of evidence and should be named. As one current platform example, TradingView describes Paper Trading as practice using simulated money; its separate Broker Emulator documentation lists configurable commission, slippage, limit-fill and execution-delay assumptions for strategy tests. The platform’s settings describe its own simulator and backtest behavior, not fills on an exchange.
If the simulation assumes that every order fills at the observed mid-price, it may omit spread, queue priority, partial execution, rejected orders, latency and market impact. A stop can trigger without executing at its chosen level. A paper account may omit venue-specific funding, margin tiers, fees or operational failures. Record the simulated price source, timestamp, costs, missed or partial fills if modeled, and gaps between the intended process and observed behavior. Do not call a hypothetical fill a venue fill.
Paper results also remain a limited sample. A short period with few decisions says little about rare events, changing regimes or long-horizon exposure. Changing the rule during practice creates a new version. Preserve the old record so that learning does not rewrite what the original version would have done.
Forward observation: what happens after the rules are frozen?
A forward observation applies a frozen hypothesis to data that arrives later than the research period. It reduces direct reuse of that later segment for parameter selection only if you do not keep consulting it to revise the rules. Log the signal time, available market observation, decision, assumed entry convention, costs and reason for standing aside. Decide in advance how long to observe and what evidence would count as a failure.
Forward observation is not automatically paper trading: it might record signals without simulating orders. Paper trading can be conducted on historical replay or prospective data. Neither phrase, by itself, says whether real-time quotes, realistic fees or actual execution are involved. Name the actual process. A prospective paper ledger using an observed quote is stronger evidence about timing than a retrospective candle fill, but still cannot prove that a live order would have filled at that price.
Live execution is another claim entirely
A real-fill claim requires evidence of an actual submitted order and its venue response, including order identity, timestamps, quantity, status, executed prices, fees and relevant settlement records. A simulation, a screenshot of a signal or a successful software test cannot supply that evidence. Live results may also include operational, funding, liquidity and account-specific effects absent from a research model.
For the current SuperQuant product, historical backtests and manual paper-practice workflows run locally against hourly inputs. Bundled histories are synthetic, and the app is not a real exchange execution venue. Public observations, if used, must be labeled separately from simulated orders. The application does not establish actual venue fills through its paper ledger.
Keep one evidence record across stages
Save the hypothesis and rule version before each stage. Preserve the data range, time source, price source, cost assumptions, skipped decisions, changed parameters and review date. State what each stage did not test. When moving from historical analysis to prospective practice, freeze a version first. When moving from practice to any real venue, reassess the contract, current fees, order behavior, permissions and personal risk independently; a favorable simulation is not an instruction to proceed.
The backtest audit checklist helps record historical assumptions. The crypto trading journal separates the plan from what happened, and the trading plan template provides a place for explicit failure criteria. Every stage can end with “stop” or “collect better evidence.” This is general education, not financial advice or a promise of performance.
Store the evidence stage as a field in the record, not as one broad “tested” label. For each observation, preserve the data-arrival time, signal time, decision, simulated or submitted order state, fill source and cost source. A complete record can establish what the process did without establishing a profitable outcome.