How far back you backtest is a strategy decision, not a technicality. Test over the last 12 hours and you learn how a wallet trades in one narrow slice of the market — which might be a slice that flatters it. Test over 30 days and you find out whether the result survives a change in conditions. The window you pick quietly decides whether you're measuring an edge or measuring luck.
This guide is about that choice: 12 hours versus 7 days versus 30 days, what each tells you, and how to read a result that only holds over a short stretch. It sits alongside how to read a copy-trading backtest; here the focus is the single lever of time.
The core trade-off: recency versus robustness
Every backtest window is a compromise between two things you want at once and can't fully have.
- Recency favours the short window. The last 12 hours are the market you're about to trade into — same tokens, same liquidity, same crowd. It's the most representative of right now.
- Robustness favours the long window. Thirty days span quiet days and frantic ones, a rally and a flush. A strategy that holds up across all of that is far more likely to keep working than one tuned to a single afternoon.
A short window is fresh but fragile; a long one is sturdier but older, and part of it may no longer resemble the market you'll copy into. You don't resolve this by picking a side — you use both ends for what each is good at.
What a 12-hour window tells you
Twelve hours is a mechanics check, not a verdict. It's the trial window in Soltrace — the last 12 hours, one wallet — and that's the right way to think about it even if you have a longer window available.
- It confirms the plumbing. Does the wallet trade often enough to copy? Do your filters let anything through, or screen out the whole strategy? Does the sizing produce positions that aren't dust? Twelve hours answers those fast.
- It's one micro-regime. Half a day is, at most, one mood of the market. Whatever happened in it — a run, a chop, a dead patch — is the only thing your strategy got tested against.
- It's easily luck. With few trades, a single fortunate entry or one avoided rug can flip the whole result. A great 12-hour curve is a reason to look closer, never a reason to fund.
Treat 12 hours as a sanity check on a single target: it tells you the strategy runs and roughly how it behaves, not whether that behaviour repeats.
What a 7-day window tells you
Seven days — the Pro window, up to 10 wallets — is where a backtest starts to be useful for decisions. A week usually contains more than one kind of session, so an entry that only worked in one narrow moment has room to show its other face.
- It's enough to tune. Change one exit rule, re-run, compare. Over a week the equity curves separate enough to show which change actually helped rather than reading noise.
- It's enough to compare a shortlist. With up to 10 wallets you can run the same rules across several candidates and rank them on the same seven days — a genuine like-for-like, not screenshots from different periods.
- It's still one week. This is the catch, and it's the big one. A week is short enough that if it happened to be a strong week, almost everything looks good in it.
Seven days turns a shortlist into a pick and rough rules into tuned ones. It's long enough to trust more than 12 hours, and short enough that you should still ask what kind of week you tested on.
What a 30-day window tells you
Thirty days — the Elite window, up to 250 wallets — is a screening and stress tool. A month almost always contains a mix: green stretches and red ones, high-volume days and dead ones — and that variety is the whole point.
- It screens at scale. With up to 250 wallets you can run a large list through the same rules and let the month sort them, instead of hand-checking each over a period too short to mean much.
- It spans conditions. This is where over-fitted strategies quietly fall apart. A rule set fitted to one good week often gives back its gains the moment the market turns — and 30 days is long enough to contain that turn.
- It's the toughest test your plan offers. A strategy that survives a month of mixed conditions has cleared a higher bar than one that only survived a hand-picked seven days.
Thirty days won't be as fresh as the last 12 hours, and some of what it contains is older market. That's the trade for robustness — and for screening a large list, it's usually worth it.
The over-fitting danger of a single week
Here's the failure mode the window choice really guards against. Tune a strategy against one week until the curve looks beautiful — tighter take-profit here, a filter loosened there — and you may not have found an edge at all. You may have fitted the rules to the exact wiggles of that week: its noise, not its signal.
The tell is fragility. A strategy hand-fitted to one short window looks flawless on that window and falls apart on any other. That's why the same tuning that dazzles over 7 days can disappoint over 30 — the longer window holds conditions the tuning never accounted for. The more you tune, the longer the window you should validate on, because tuning is the activity that manufactures results that don't generalise.
The regime problem: a bull week flatters everything
Markets move in regimes — stretches where almost anything works, and stretches where almost nothing does. If your window lands entirely inside a good regime, it will make a mediocre strategy look sharp, because the regime, not the strategy, did the work.
This is the strongest argument for the longest window you can run. A single 12-hour or 7-day sample can sit wholly inside one regime and tell you nothing about the other. A 30-day window is more likely to straddle a shift — and how a strategy behaves when the regime changes is what you need to know before funding it. If a result only holds inside one flavour of market, it isn't an edge; it's a weather report.
The data-availability reality
One honest constraint shapes all of this. Soltrace captures 1-second price data going forward, with 30-day retention — which has two consequences worth stating plainly.
- You can't test before capture began. There's no deep historical archive to reach into — very old market history simply isn't testable. The window is the recent past that's been recorded, not all of time.
- A brand-new wallet has little to test. If a wallet only started trading recently, there isn't yet a month of behaviour on record. Sometimes the right move is to add it, let data accrue, and backtest properly once the history exists.
Neither is a flaw to work around; it's the boundary of what an honest replay can claim. A backtest can only speak to conditions it captured.
From window to live
The practical guidance is simple. Sanity-check short — a 12-hour run to confirm the mechanics on one wallet. Decide on the longest window your plan allows — 7 days on Pro to tune and compare, 30 days on Elite to screen across conditions. And distrust a result that only holds over a single short window, because that's the shape luck and over-fitting both take.
Whatever window you settle on, the backtest replays the target's real on-chain trades under your exact rules — sizing (fixed SOL, fixed % of vault, proportional to target, proportional to portfolio) and exits (take-profit, stop-loss, trailing, scale-out, time-based, mirror sells) — priced on 1-second data across Raydium, Meteora, Orca and PumpSwap. It returns an equity curve, every fill and exit with its reason, max drawdown, the funnel, and a per-wallet breakdown. When a strategy holds up over the longest window you can throw at it, going live is short: you fund a non-custodial Solana PDA vault you own the entire time — the keeper places and sizes copies but can never withdraw — with keeper-signed execution across Jito, Nozomi, Astralane and AllenHark, optimistic on Pro and Elite, at a flat 0.85% per trade and no profit share. Compare the windows on the pricing page, and if you're still deciding whether a wallet is worth testing, see can you backtest a Solana wallet and how to backtest a Solana copy-trading strategy.
Ready to test across a real range of conditions? Try Soltrace backtesting and choose the longest window your plan allows before you fund a thing.
Copy trading does not guarantee profit, and crypto trading carries substantial risk of loss. A backtest describes the past under your rules; it is not a prediction. Nothing here is financial advice.