What makes a crypto simulator different?
Crypto markets run continuously, quote differently on every venue, charge holding costs on perpetual contracts, and have far less history than currencies or equities. All four change what replayed practice means.
Continuous trading removes the session structure that organises practice in other markets. There is no open, no close and no weekend gap, so the habit of framing a strategy around a session — which carries a good deal of the structure in forex — has no direct equivalent. Practice sessions have to be bounded by something else, usually a fixed number of hours or a chosen window in a given timezone.
Venue fragmentation means the price you are replaying is one exchange's version of the market. Two exchanges can differ by enough to change whether a level was touched, which decides whether your stop was hit. A simulator rarely tells you which venue the data came from.
Perpetual futures add funding payments, which accrue while a position is open and are invisible on a price chart. A strategy holding positions for days can have a material part of its result determined by a cost the replay never shows.
Does crypto history go back far enough to practise on?
Most crypto instruments have a tradable history measured in a handful of years, and for anything outside the largest few, the record often does not cover a complete cycle of expansion and contraction.
This is a sharper constraint than it first appears. A strategy practised entirely across a rising market has never met a prolonged decline, and one practised across a single crash has met exactly one. In currencies you can reach back through several distinct regimes; in most crypto pairs you cannot, and no amount of replay creates history that does not exist.
There is also a survivorship problem specific to the asset class. The instruments available to replay today are the ones that still exist, and the delisted and abandoned tokens are not in the sample. Practising on the survivors builds intuitions calibrated to a population selected for having survived.
The honest response is to weight the conclusions accordingly and to be explicit about which regimes the practice covered, rather than treating a short history as if it were a long one.
What does a crypto simulator get wrong about execution?
Replay tools model crypto execution optimistically in three specific ways: spread, liquidity depth, and the behaviour of the venue during exactly the moments a strategy is most likely to trade.
Spreads on major pairs are tight most of the time and widen sharply during the fast moves that trigger breakout and momentum strategies. A simulator applying a fixed spread is therefore most wrong precisely when the strategy is most active, and the error runs in the flattering direction.
Depth is the second issue. A chart shows the last traded price and says nothing about how much size sat at it. For a practice session this is invisible; for anything beyond small retail size it is the difference between the replayed fill and the real one.
Venue behaviour under load is the third. Exchanges have historically slowed or restricted access during the largest moves, which means some of the trades a replay shows as available were not, in fact, available. No replay tool models this, and it cannot be inferred from the price series.
How should you practise a crypto strategy?
Practise on the instruments and venues you will actually trade, bound your sessions deliberately since the market provides no natural boundary, and record funding and fee assumptions alongside every trade.
Pick the venue first and take the data from it, because a strategy tested on one exchange's prints and traded on another is being graded on a different series than it will trade. Where a simulator does not disclose its source, treat any result that depends on precise levels with suspicion.
Bound sessions by time and stick to the boundary. Without an enforced close, replay sessions run until attention fails, and the trades taken in the last hour of a long session are the ones most likely to be the drifted rules rather than the written ones.
Write the cost assumptions into the log. Fee tier, expected spread, and funding rate for any perpetual position held overnight. A crypto strategy with a modest per-trade edge can be entirely consumed by costs that never appeared on the chart you practised against.
How does QuantParadox grade a crypto strategy?
QuantParadox includes crypto instruments in the same decade-scale minute-resolution archive used for forex, metals and indices, and grades crypto rules out-of-sample by the same default as everything else.
The properties that matter for this asset class are the ones the platform is built around: explicit cost modelling rather than assumed-away spread, minute resolution so that a candle containing both stop and target is resolved by looking inside it, and disclosure where coverage is thin rather than a silently shortened test.
That last point does real work in crypto specifically. Asking for ten years on an instrument that has four and receiving a result labelled as a decade test is a common and quiet failure, and the correct behaviour is to state the window actually covered and which end is missing.
The limitation worth naming: exchange fragmentation is real, and a graded result reflects the data series behind it. Funding costs on perpetuals and venue-specific depth remain outside what any historical price grading can settle.