TradingView Deep Backtesting: coverage, limits and comparability

Compare three runs with identical parameters

This diagnostic experiment separates the calculation start from the dates on which orders are allowed. It is a reproducible procedure, not a published performance result. Use standard daily candles for COINBASE:BTCUSD and the script below. Confirm that your available history includes 2024 and 2025 before continuing.

1. Keep the symbol, exchange, timeframe, candle type, source code and strategy properties fixed. The example uses EMA lengths 20 and 100, fixed size 0.01, initial capital 100,000, commission 0.1% and slippage of one tick. These are controlled example inputs, not recommended execution assumptions.
2. Leave the entry inputs at January 1, 2025 through January 1, 2026 UTC, with the end excluded. The EMA calculations run on every available bar, even while entries are blocked.
3. Run A, B and C below without editing the inputs. Refresh each report after its range changes. Save each report and its trade list with the run label.
4. Compare first entry time, first exit time, total closed trades, net P&L and open-position status. Locate the earliest divergent trade before interpreting a difference in aggregate profit.

Pine Script v6 comparison strategy

//@version=6
strategy("Deep Backtesting comparison", overlay=true,
     initial_capital=100000, default_qty_type=strategy.fixed,
     default_qty_value=0.01, commission_type=strategy.commission.percent,
     commission_value=0.1, slippage=1, pyramiding=0)
fastLength = input.int(20, "Fast EMA", minval=1)
slowLength = input.int(100, "Slow EMA", minval=2)
entryStart = input.time(timestamp("01 Jan 2025 00:00 +0000"), "Entry start")
entryEnd = input.time(timestamp("01 Jan 2026 00:00 +0000"), "Entry end")
fast = ta.ema(close, fastLength)
slow = ta.ema(close, slowLength)
crossUp = ta.crossover(fast, slow)
crossDown = ta.crossunder(fast, slow)
entryWindow = time >= entryStart and time < entryEnd
if entryWindow and crossUp
    strategy.entry("Long", strategy.long)
if crossDown or time >= entryEnd
    strategy.close("Long")
plot(fast, "Fast EMA", color=color.blue)
plot(slow, "Slow EMA", color=color.orange)

Market orders use the broker emulator's default next-available-tick fills. The entry window gates order creation, not the fill timestamp. At the end boundary the script requests a close; the report extends to January 10 so a subsequent daily bar can fill it. Verify that the final position closed. The settings do not model every cost or execution constraint. Order execution documentation.

Interpret the difference without inventing a result

TradingView documents EMA and RMA dependence on the calculation start. Run C gives the averages pre-entry history; run B starts them at the entry boundary. Earlier initialization can change a crossover. A changed entry can then affect later positions, even after the averages have become close. Matching results are also possible: this experiment does not require a mismatch. Why start dates matter.

For a 100-period EMA, the weight of an initial-state difference after 300 updates is (99/101)³⁰⁰, about 0.00248. A hypothetical initial difference of 10 therefore shrinks to about 0.0248 if subsequent closes are identical. This arithmetic illustrates decay, not a universal warm-up length or a guarantee of matching trades.

If A has no 2025 bars, mark it unavailable instead of calling the result zero performance. If B and C disagree, inspect their first trades and settings. If they agree, retain that observation without concluding that every strategy is insensitive to its start date.