Most lists of technical indicators for swing trading hand you seven tools and a sentence of folklore for each one. No setting. No sample size. No cost assumption. Quantzee's own ATR sweep points at a different, more useful fact. Across 1,080 real-cost combinations on gold, silver and bitcoin, nothing was profitable on the 1-minute, 3-minute or 5-minute chart. Every single per-instrument winner landed on the 4-hour bar instead. That is not a day-trading result dressed up as something else. It is, almost by accident, an argument for swing trading specifically. This guide builds the case from that data instead of from opinion. It names which indicators carry real weight for a multi-day hold, and it says plainly where the evidence runs out.
According to the sweep published on our full sweep results page, 391 of 1,080 real-cost combinations — 36% — lost the entire notional under full reinvestment. We come back to that number later, because it is the uncomfortable half of the same dataset that produced the winners.
Fix the hold first — everything below is tied to it
Every one of the five pages we read while researching this piece makes the same mistake. It recommends a best indicator for swing trading without ever defining what "swing" means in bars, days or risk. We are not making that mistake here. For this guide, a swing trade is a position held 2 to 10 days, read on a 4-hour or daily chart. That is the definition wallstreetzen gestures at — "a few days to a couple of weeks" — without ever tying a setting to it. It is the definition bullishbears gets closest to naming outright, with a 3–6 day hold against a 9/20 EMA cross. Credit where it is due. Then we go further, with real backtest numbers behind the setting instead of a rule of thumb.
Pin that hold length before you touch a single indicator setting. An RSI period, a Bollinger width or a SuperTrend multiplier that works for a 2-day hold is not the same setting that works for a 10-day hold. Most of the swing-trading content ranking today never makes that distinction once, which is exactly why the settings it recommends rarely survive contact with real cost.
The sweep method — before the numbers, not after
Our data in this piece comes from one source: Quantzee's own parameter sweep, published in full on the site and reproducible from the raw grid. The methodology, as documented on the SuperTrend settings guide, runs as follows.
- ATR period swept across {7, 10, 14, 20, 30} — five values.
- Multiplier swept from 1.0 to 5.0 in 0.5 steps — nine values, 45 combinations per timeframe.
- Eight timeframes tested: 1m, 3m, 5m, 15m, 30m, 1h, 4h and 1d.
- Three instruments with complete historical data: gold (XAUUSD), silver (XAGUSD) and bitcoin (BTCUSD).
- Entries and exits filled at the next bar's open after the signal bar closes. No lookahead, no pyramiding.
- Real round-trip cost applied per instrument: 4 basis points on gold, 8 on silver, 15 on bitcoin.
- 1,080 real-cost rows in total, plus a 1,080-row zero-cost diagnostic baseline that is never tradable and exists only to isolate signal quality from cost.
Data windows matter here too. Gold and silver run January 2020 through October 2026 — close to 6.8 years of bars. Bitcoin's complete data starts only in December 2023, a 2.8-year sample against the metals' 6.8. Any comparison across instruments has to carry that gap, because a 2.8-year sample sees fewer market regimes than a 6.8-year one by definition, not by coincidence.
The negative result, stated as a negative
Here is the finding most swing-trading content would rather skip past. At the fast end of the grid, SuperTrend does not survive its own transaction costs. No setting — not one of the 45 ATR/multiplier pairs — produced a profitable real-cost result on 1-minute, 3-minute or 5-minute bars, on any of the three instruments. The full grid shows 391 of the 1,080 real-cost combinations landed at -100% total return, wiped out under reinvestment. Those losses cluster almost entirely on the fastest timeframes, where per-trade cost compounds hardest against a trend-following stop.
That is not a reason to distrust the indicator. It is a reason to distrust the timeframe it is being asked to work on. The same indicator, the same cost model and the same testing method turn up three outright winners once you move to the 4-hour chart. That chart also happens to sit inside the 2-to-10-day swing window we fixed above — which is the whole point of this page.
The three 4-hour winners — with their caveats intact
Per the sweep data, the top-performing real-cost SuperTrend configuration on each instrument, 4-hour timeframe, breaks down like this.
- Gold, ATR 30 / multiplier 2.5 — win rate 42.9%, profit factor 1.32, Sharpe 1.01, max drawdown -25.7%, 340 trades across roughly 2,484 days of data.
- Silver, ATR 30 / multiplier 3.5 — win rate 44.5%, profit factor 2.20, Sharpe 1.50, max drawdown -36.0%. Only 191 trades over the 6.8-year window, though — a thin sample by our own 100-trade floor, worth repeating as a caveat every single time this row gets quoted elsewhere.
- Bitcoin, ATR 30 / multiplier 5.0 — win rate 42.6%, profit factor 1.25, Sharpe 0.56, max drawdown -53.1%, 61 trades across a 2.8-year window that is roughly a third the length of the metals' sample.
Every one of those rows carries its own instrument, timeframe, cost model, trade count and window. That is the house rule we return to at the end of this piece. Quoting silver's 2.20 profit factor without also stating 191 trades, or quoting bitcoin's 42.6% win rate without stating its 2.8-year window, is the exact failure mode every other swing-indicator page falls into. A number gets borrowed. The context that made it checkable gets left behind.
Seven indicators, grouped by what they actually measure
Names are a worse organizing principle than inputs. RSI and Stochastic look unrelated by name. Mechanically they consume almost the same price series, and they frequently fire together as a result. Group by what feeds the formula instead of what the formula is called.
Trend — moving averages
A simple or exponential moving average, or a pair of them crossing, answers one question: which direction is price biased on this timeframe. On a 4-hour or daily chart, the 50/200 EMA cross is the standard swing-trend filter. It tells you the side to trade, not the day to enter. It cannot tell you momentum is fading, and it cannot tell you volume has dried up. It only knows direction, and nothing else.
Momentum — RSI and MACD
RSI and MACD both derive from the rate of change in price, just with different smoothing applied on top. For a swing hold, RSI's useful signal is not the textbook "70 is overbought, sell" rule. It is a pullback to the 40–50 zone inside an established uptrend — closer to a continuation entry than a reversal call, a distinction we cover in more depth for multi-day holds specifically in our standalone RSI guide. MACD's histogram crossing zero on the 4-hour chart lines up reasonably well with the hold window fixed earlier. Neither tool tells you anything about volatility or participation; both read price alone.
Volatility — ATR and Bollinger Bands
ATR is the backbone of the sweep above. It sizes the SuperTrend band, and separately it sizes stops — 2x ATR(14) is a common swing stop distance. Bollinger Bands measure the same underlying volatility, but visually, as a squeeze-and-expand cycle. That makes them useful for timing a breakout entry, not for predicting direction. Neither indicator has an opinion about whether the breakout it flags will hold once price clears the band.
Volume — On-Balance Volume
OBV is the one tool on this list that reads participation instead of price. It answers a different question entirely: is volume confirming the move, or is price advancing on thinning volume while fewer traders back it. That independence is exactly why, as the next section shows, OBV catches warnings that a pure price-momentum tool misses on the same chart, at the same time.
SuperTrend
SuperTrend is an ATR-banded trend-following overlay — the subject of the entire sweep above. On the 4-hour chart, with the ATR 30 settings shown, it is the one tool on this page carrying a measured, real-cost, multi-year result behind it, rather than a described tendency borrowed from a different timeframe.
Which of your three indicators to drop
RSI, Stochastic and Williams %R are, mechanically, one input series wearing three hats. All three normalize recent price momentum into a bounded oscillator, with different lookback and smoothing choices layered on top. Running all three on one chart rarely adds a second opinion. It adds three readings of the same underlying signal, dressed up as confirmation.
The useful comparison is not momentum against momentum. It is momentum against volume, where the inputs are actually independent of each other. Our team's 90-session independence test, run on an hourly chart in late 2025, compared a 14-period RSI against a 20-period CCI. Their overbought crossings overlapped within 1–2 candles on 19 of 22 occasions — roughly 86% agreement, which is exactly what you would expect from two indicators built on the same price input with different math wrapped around it. RSI against Chaikin Money Flow told a different story. Of 22 RSI overbought crossings, Chaikin Money Flow agreed on only 11 — a 50% agreement rate. The other 11 occasions were marked by volume already falling while price momentum still looked strong.
The volume tool flagged five early warnings that the price-momentum tool missed entirely in that window. Three of those five preceded a pullback of 1.5% or more within the next two trading sessions. Across the full 240-sample set behind that study, setups built on two independent-family signals held their initial direction past the next three candles 61% of the time, against only 44% for same-family signal stacks. That is the drop test in one line: if removing an indicator barely moves your signal count, it was never adding information. It was adding another vote from the same input, counted twice. Read the full breakdown on our confluence vs. correlation page.
Three swing stacks that hold up — and two that only look like they do
A single indicator is a filter, not a strategy. Here are three multi-indicator stacks built for the 2-to-10-day hold, each with an entry, a stop, an exit and — the piece most guides skip — a skip condition that keeps you out of the setup entirely when the evidence is thin.
Stack 1 — Trend-confirmed pullback
- Trend: 50/200 EMA on the daily chart. Only trade in the direction of the cross.
- Momentum: RSI(14) pulls back to 40–50 inside the established trend.
- Entry: RSI turns back up through 50 with price holding above the 50 EMA.
- Stop: below the most recent swing low, sized at no more than 2x ATR(14).
- Exit: prior swing high, or an RSI bearish divergence against a new price high.
- Skip condition: if the 50/200 EMA gap is inside 1 ATR, the trend read is too thin to trust. Stand aside instead of trading the chop.
Stack 2 — SuperTrend with a volume filter
- Trend: SuperTrend on the 4-hour chart, ATR period in the 20–30 range per the sweep above.
- Confirmation: OBV rising alongside the SuperTrend flip. A flip on falling OBV is a weaker signal.
- Entry: close on the SuperTrend side with OBV making a higher high in the same session.
- Stop: the SuperTrend line itself at entry. The indicator doubles as a trailing stop by design.
- Exit: the SuperTrend flip in the opposite direction.
- Skip condition: if OBV has been flat for the prior 10 bars, the flip has no participation evidence behind it. Skip it and wait for the next one.
Stack 3 — Volatility breakout
- Volatility: Bollinger Bands 20/2. Wait for a visible squeeze — band width near its 6-month low.
- Trend: SuperTrend for the direction of the resolution once it comes.
- Confirmation: volume above its 20-period average on the breakout bar itself.
- Entry: close outside the band in the SuperTrend direction, on above-average volume.
- Stop: the opposite band, or the 20-period basis line, whichever sits closer.
- Skip condition: a breakout on below-average volume. That is the single most common failed-breakout pattern, and it is visible before you enter, not just after the fact.
Two stacks that look sensible and are not
RSI + Stochastic + Williams %R "triple confirmation." As the independence test above shows, these three overlap on roughly the same input around 86% of the time with a correlated momentum partner. Three green lights from one underlying signal is not three independent opinions. It is one opinion, repeated three times and mistaken for consensus.
MACD crossover + RSI 70/30 reversal. This pairs a trend-continuation signal — MACD histogram crossing zero mid-trend — with a mean-reversion signal that fires at the opposite point in a trend's life cycle. In practice the two rarely agree at a useful moment. Traders end up overriding whichever one disagrees, which quietly defeats the purpose of having a rule in the first place.
Position sizing and risk — the part indicators never cover
None of the seven indicators above size a position or cap a loss. That is a separate decision, and skipping it is how a 42% win rate with a 1.3 profit factor — a perfectly survivable long-run edge — turns into an account-ending drawdown on a single bad week. Every stack above ties its stop to either a structural level (swing low, opposite band) or a volatility multiple (2x ATR), rather than a fixed dollar or point amount, because a fixed stop that is sized correctly for a quiet week is too tight for a volatile one and too loose for the next quiet one.
The gold row above carried a -25.7% max drawdown over 340 trades and still returned a positive total; the bitcoin row's -53.1% drawdown over only 61 trades is a different risk profile entirely, even though both rows cleared the sweep's profitability bar. Treating those two rows as interchangeable because they both "passed" the backtest is how a trader ends up sized for the gold row's risk while actually holding the bitcoin row's drawdown.
Does this transfer to stocks, forex or indices?
Honestly: we do not know, and we are not going to pretend otherwise. The sweep above covers gold, silver and bitcoin only, because those are the three instruments with complete, clean, tick-level history on the data vendor Quantzee's team used for this test. No equity, no index future and no forex pair appears anywhere in the 1,080-row grid. That is a real gap, not a rounding error, and it means the exact ATR 30 settings above should not be copied onto a stock chart and treated as validated.
What does carry over, in our view, is the shape of the result rather than the specific numbers. Transaction cost scales with how often a rule trades, and fast timeframes trade far more often than slow ones. That relationship holds regardless of instrument. A retail equity swing trader paying a wider effective spread than our 4-to-15-basis-point model should expect the same cost-versus-frequency tradeoff to push their own breakeven point toward slower timeframes too, even without a stock-specific sweep to confirm the exact crossover. Treat that as a reasoned expectation, not a measured result — the measured result is the metals-and-crypto grid above, and only that.
If you trade index futures or single-name equities, the honest move is to rerun this same ATR-period and multiplier grid on your own instrument and timeframe before trusting any of the specific numbers in this piece. The method transfers even where the numbers do not: fix a hold period first, sweep a real grid instead of guessing one setting, apply your actual round-trip cost, and require at least 100 trades before you believe the result.
Choosing a stack by account size and risk tolerance
The three stacks above are not ranked best-to-worst. They fit different situations. Stack 1, the trend-confirmed pullback, needs the least monitoring: a daily chart, checked once a session, is enough to catch the RSI pullback entries it is built around. That makes it the more practical choice for a trader who cannot watch a 4-hour chart intraday. Stack 2 requires closer attention to the 4-hour close itself, since the SuperTrend flip and OBV confirmation both need to be read at that specific candle, not approximated from a daily summary. Stack 3, the volatility breakout, demands the most patience of the three — the squeeze condition it waits for does not show up every week, and forcing an entry before the band has actually compressed defeats the setup's entire logic.
Account size interacts with the drawdown figures from the sweep in a way that is easy to skip past. A trader sized for the gold row's -25.7% max drawdown over 340 trades is sized very differently than one prepared for the bitcoin row's -53.1% drawdown over just 61 trades. Position size should scale down, not stay fixed, as the underlying instrument's historical drawdown widens — the sweep gives you that drawdown figure precisely so you can make that adjustment before the drawdown happens live, instead of after.
Why a swing win rate you read online is probably meaningless
Before trusting any win-rate figure for a swing setup — including the ones in this piece — run it against a control that contains no real trend at all. Our data from a random-walk test built exactly that control. Fifty synthetic price series, 1,000 bars each, calibrated to 14% annualised volatility with zero drift, meaning the data is engineered to contain nothing a trend-following rule should be able to catch.
Run a trend-following rule against those 50 series anyway, and 22 of 50 runs — 44% — still came back profitable. Fourteen of 50, 28%, cleared a marketing-style bar of a 40%+ win rate and a 1.3+ profit factor, on data built to contain no edge whatsoever. The median return across all 50 runs was -5.7%, which is the honest number. The +20.2% median among only the winning runs is the number that would get screenshotted and posted as proof of an edge, with the other 36 runs quietly left out of the frame. A 200-run robustness check came back close to the same picture: 49% positive, 20.5% clearing the marketing bar, confirming the original 50-run sample was not a fluke of small numbers.
The mechanism is simple. A rule with any trend-following component will occasionally catch a long, directionless drift inside a random walk, purely by chance. Visually, and statistically, that looks identical to catching a real trend on a real chart. Read the full methodology on our random-walk control page. The practical takeaway: a swing win rate with no sample size, no cost model and no random-walk control attached is exactly the kind of number this test shows up roughly a quarter of the time by pure chance alone.
The house rule: an indicator does not have a win rate
This is a rule we have published before, and it is worth repeating here, because it is the single organizing idea behind this entire page. A win rate belongs to a complete rule set — entry, stop, exit, skip condition, position size — never to an indicator sitting in isolation. Every number in this guide carries its instrument, its timeframe, its cost assumption, its sample size and its data window attached to it. A number arriving with none of those attached is not evidence. It is a claim, and the random-walk test above shows exactly how easily a claim like that shows up in data with no real signal in it at all.
That same discipline is why we are explicit about what this page does not have. No equity swing backtest — the sweep above covers gold, silver and bitcoin only, with no stocks and no index. No measured multi-year result for RSI, MACD or Bollinger Bands tested in isolation. No forward-tested result for any of the three stacks above. Where we have no measurement, we say so, rather than filling the gap with an assertion dressed up as one.
Where this guide stops — and where to look instead
If you are after sub-5-minute scalping settings, this is the wrong page by design. The sweep above shows those timeframes losing money against real cost on every setting tested. Our scalping indicators guide covers that territory with its own evidence instead of borrowing swing numbers that do not transfer across timeframes. For the full roundup of what Quantzee's toolkit tests across every timeframe, see our main indicators guide, where the swing section points back to this page for the full sweep detail. Traders running the SuperTrend side of this directly can see the live tool on the SuperTrend Pro page, and the broader signal-independence toolkit is covered on the AI Adaptive Quant Toolkit page.
For a regulator's baseline on what technical analysis can and cannot claim to do, the SEC's own investor education glossary is a useful outside reference. FINRA's writeup on momentum-based strategies covers the same ground from the regulator's side of the desk. Both are worth reading before trusting any indicator vendor's marketing claims, Quantzee's included.
Frequently Asked Questions
Last updated: October 23, 2026 — initial publication, built from Quantzee's published ATR sweep, random-walk control and signal-independence datasets.
