TradingView Super Scalper Strategy: Best Settings, Backtest
The short version. This is a TradingView Super Scalper strategy backtest. VinayKumarKV's "Super Scalper - 5 Min 15 Min" has about 7,400 boosts. I ported it to Python and ran it on Exness tick data for seven markets, with real spread and commission, from January 2020 to September 2026.
- As a scalper, it loses. On M5 and M15, the two timeframes in its name, none of the 14 market/timeframe cells got above PF 1.01 after costs with the default long + short setting. Gold M15: PF 0.95, 4,753 trades, −1,354 USD per 0.01 lot.
- Put the same code on H1 and it's the steadiest result I've had in this series. Gold H1: PF 1.31, +3,409 USD per 0.01 lot, 1,116 trades, 778 max drawdown. After swap, PF 1.27. Buy and hold made +2,588, with a drawdown of 1,600.
- It held on data I never looked at. On Dukascopy gold from 2004 to 2019, the same H1 setup made PF 1.10 after swap, and it beat 99.5% of randomly shifted copies of itself.
- USDJPY M30 and H1 also worked (PF 1.11 and 1.18). EURUSD, GBPUSD and NZDUSD lost.
- The stop loss and take profit in the code are never used, and neither is the golden cross. The part doing the work is a double RSI filter.
My verdict: don't use it to scalp. On gold H1 with the default settings it's the one I'd keep watching. The edge is thin, though, and it swings hard: one year −39%, another +54%.
What is the Super Scalper strategy on TradingView?
It's an open-source strategy published on 22 March 2022, shown on a NIFTY 5-minute chart. The author's description says it "works better in 5 and 15 Mins time frame", suggests backtesting with 1:2 reward to risk, and says to use other alerts to take profit or trail the stop. The current version (19 May 2022) switched from one RSI to two "to increase the profitable percentage."
Here's the part of the code that trades:
atr_slen = ta.wma(ta.tr(true), 14) // "ATR Smoothing" = WMA by default
upper_band = atr_slen * 1 + close
lower_band = close - atr_slen * 1
longCondition = open < lower_band
shortCondition = open > upper_band
RSILong = ta.rsi(close, 25) > ta.rsi(close, 100)
RSIShort = ta.rsi(close, 25) < ta.rsi(close, 100)
if (longCondition)
stopLoss = low - atr * 2
takeProfit = high + atr * 5
strategy.entry("long", strategy.long, when = RSILong)
if (shortCondition)
stopLoss = high + atr * 2
takeProfit = low - atr * 5
strategy.entry("short", strategy.short, when = RSIShort)
Three things about this aren't obvious from the chart.
"Open below the lower band" means a big green candle. The bands are built from the close of the same bar. So open < close − ATR is just close − open > ATR: the candle closed up by more than its average range. The "Buy" label looks like buying a dip under a band. It's actually buying after a strong up bar. That's momentum, not mean reversion.
The stop loss and take profit do nothing. stopLoss and takeProfit are calculated and never passed to strategy.exit. The only exit is the opposite signal, so it's a stop-and-reverse system that's in the market about 98% of the time.
The golden cross is decoration. EMA 21 and 65 crosses get labels on the chart and never touch an order. To be fair, the author says so in the release notes.

That's a typical stretch on gold H1. The longs carried most of the rally from 3,800 to 4,350, the short flips in between mostly cost a little, and the last sell on 21 October caught the start of the drop.
Does Super Scalper repaint?
No. Everything uses the bar's own open and close plus past bars. The signal is final on the bar close, and I fill it at the next bar's open, the same as TradingView's tester with default settings.
How I tested it
- Port: Python, signals on the close, fills at the next open, stop-and-reverse like
strategy.entry. Default settings: WMA of true range 14, multiplier 1, RSI 25 and 100. - Data: Exness Raw Spread ticks for XAUUSD, USDJPY, EURUSD, GBPUSD, NZDUSD and US500 from January 2020 to September 2026 (US500 to August), BTCUSD from January 2024.
- Costs: buys fill at the real ask on the bar's first tick, sells at the bid, plus 3.5 USD per lot per side for forex and gold.
- Swap: included for gold (Exness Raw rates, scaled to the gold price at the time). Not included for the others.
- Older data: Dukascopy M1 via Tickstory, 2004 to 2026, for gold, USDJPY, EURUSD, GBPUSD, NZDUSD and the S&P 500.
- Size: fixed, no compounding. Gold in USD per 0.01 lot (1 oz), forex in pips per lot, US500 in index points, Bitcoin in USD per 1 BTC.
The results

XAUUSD, Jan 2020 – Sep 2026, before swap. Buy and hold: +2,588, max drawdown 1,600.
| Timeframe | Trades | PF | Net | Max DD | Win rate | Long net | Short net | Long only: PF / net / DD |
|---|---|---|---|---|---|---|---|---|
| M5 | 14,490 | 0.93 | −3,010 | 4,081 | 30% | −212 | −2,798 | 0.99 / −212 / 1,467 |
| M15 | 4,753 | 0.95 | −1,354 | 2,057 | 31% | +612 | −1,966 | 1.05 / +612 / 1,022 |
| M30 | 2,266 | 1.23 | +3,746 | 589 | 34% | +3,166 | +579 | 1.42 / +3,166 / 532 |
| H1 | 1,116 | 1.31 | +3,409 | 778 | 35% | +2,999 | +411 | 1.60 / +2,999 / 532 |
| H4 | 329 | 1.16 | +982 | 1,016 | 35% | +1,802 | −819 | 1.79 / +1,802 / 846 |
USD per 0.01 lot. With swap: M30 1.21, H1 1.27 (+3,016), H1 long only 1.51.
USDJPY, Jan 2020 – Sep 2026. Buy and hold: +4,845 pips.
| Timeframe | Trades | PF | Net | Max DD | Win rate | Long net | Short net | Long only: PF / net / DD |
|---|---|---|---|---|---|---|---|---|
| M5 | 15,549 | 0.83 | −22,503 | 22,807 | 28% | −8,832 | −13,671 | 0.86 / −8,832 / 9,044 |
| M15 | 4,871 | 1.00 | +193 | 2,499 | 31% | +2,521 | −2,328 | 1.08 / +2,521 / 1,552 |
| M30 | 2,334 | 1.11 | +5,343 | 1,493 | 34% | +5,104 | +239 | 1.24 / +5,104 / 950 |
| H1 | 1,153 | 1.18 | +5,817 | 1,923 | 35% | +5,298 | +519 | 1.37 / +5,298 / 1,553 |
| H4 | 332 | 0.74 | −5,925 | 7,500 | 27% | −429 | −5,496 | 0.96 / −429 / 2,900 |
Pips per lot, swap not included.
The other five markets, long + short:
| Market | Timeframe | Trades | PF | Net | Max DD | Long only PF / net |
|---|---|---|---|---|---|---|
| EURUSD | M5 | 16,260 | 0.80 | −19,131 | 19,182 | 0.80 / −9,404 |
| M15 | 5,136 | 0.88 | −6,587 | 6,673 | 0.88 / −3,146 | |
| M30 | 2,604 | 0.91 | −3,325 | 4,753 | 0.92 / −1,525 | |
| H1 | 1,342 | 1.00 | −95 | 3,023 | 1.01 / +150 | |
| H4 | 325 | 0.83 | −2,249 | 4,068 | 0.85 / −967 | |
| GBPUSD | M5 | 16,159 | 0.80 | −24,697 | 25,398 | 0.80 / −12,239 |
| M15 | 5,218 | 0.86 | −10,229 | 10,548 | 0.86 / −5,046 | |
| M30 | 2,591 | 0.86 | −7,186 | 9,323 | 0.86 / −3,471 | |
| H1 | 1,346 | 0.92 | −2,892 | 6,711 | 0.92 / −1,282 | |
| H4 | 351 | 0.79 | −3,995 | 6,375 | 0.84 / −1,581 | |
| NZDUSD | M5 | 17,417 | 0.66 | −30,819 | 30,981 | 0.66 / −15,812 |
| M15 | 5,268 | 0.79 | −9,935 | 10,221 | 0.77 / −5,353 | |
| M30 | 2,606 | 0.84 | −5,062 | 5,601 | 0.82 / −2,908 | |
| H1 | 1,243 | 0.92 | −1,728 | 3,069 | 0.89 / −1,195 | |
| H4 | 277 | 1.11 | +1,037 | 1,177 | 1.08 / +352 | |
| US500 | M5 | 14,300 | 0.98 | −1,784 | 3,788 | 1.04 / +1,315 |
| M15 | 4,693 | 0.97 | −1,456 | 3,355 | 1.07 / +1,481 | |
| M30 | 2,258 | 0.97 | −1,026 | 2,236 | 1.12 / +1,691 | |
| H1 | 1,216 | 1.01 | +217 | 2,067 | 1.21 / +2,277 | |
| H4 | 314 | 0.96 | −477 | 2,408 | 1.53 / +2,260 | |
| BTCUSD* | M5 | 7,706 | 0.93 | −116,995 | 124,451 | 0.95 / −38,110 |
| M15 | 2,565 | 0.97 | −31,183 | 70,453 | 1.01 / +3,607 | |
| M30 | 1,264 | 0.99 | −4,798 | 71,587 | 1.05 / +17,547 | |
| H1 | 653 | 1.03 | +12,462 | 46,893 | 1.10 / +24,584 | |
| H4 | 171 | 1.02 | +5,525 | 88,666 | 1.21 / +23,228 |
Forex in pips per lot, US500 in index points, BTCUSD in USD per 1 BTC (*January 2024 – September 2026). Buy and hold: EURUSD +305, GBPUSD +215, NZDUSD −820, US500 +4,425, BTCUSD +41,200. No swap.
The long-only column on US500 and Bitcoin looks fine until you compare it with buy and hold: US500 H4 long only made +2,260 points, holding the index made +4,425. That's an index going up, not the strategy.

Why does a "5 Min 15 Min" scalper only work on H1?
I expected the answer to be spread. It's only part of it. On gold M15 the average trade loses 0.12 USD per 0.01 lot before costs, and spread plus commission add another 0.16. So the signal itself has nothing on M15, and costs just make it worse.
On H1 the average trade makes +3.21 before costs. Costs are still about 0.16, which barely dents it.
The difference is the RSI filter. On M15 it barely helps (gold goes from 0.88 without it to 0.95 with it, still losing). On H1 it's most of the edge. More on that below.
Is the gold H1 result real?
I tried to break it four ways, and it held each time.
Random timing test. I took the strategy's own long/short sequence, slid it to a random point in time, and kept every trade's length. 200 times. Gold H1 beat 98.5% of the shifted copies (p = 0.015), gold M30 99.5%. USDJPY M30 and H1 passed too. Gold M15, gold H4, every other forex pair, US500 and Bitcoin didn't.
Moving the candle open. H1 candles normally start on the hour. Starting them at :15, :30 or :45 instead gave PF 1.12 to 1.13. Lower than 1.31, still positive. Gold M30 fell to 1.04–1.05 when I did the same, so I trust H1 more.
Settings. 60 combinations of ATR length (7 to 28), multiplier (0.6 to 1.5) and RSI pairs (14/50, 25/100, 50/200): 56 of 60 had PF above 1 on gold H1, and 70% were profitable in both halves of the period.
Older data. This is the one that matters. I picked "gold H1" by looking at 2020–2026, so I went back to Dukascopy gold from 2004 to 2019:

| Gold, long + short | 2004–2019 PF (after swap) | 2020–2026 PF | Random timing test, 2004–2019 |
|---|---|---|---|
| M15 | 0.90 | 0.91 | |
| M30 | 0.97 | 1.16 | beats 96% |
| H1 | 1.10 (+85% of price, 2,577 trades) | 1.27 | beats 99.5% |
| H4 | 0.98 | 1.04 | beats 89.5% |
Gold H1 kept its edge for 16 years the backtest never saw. M15 lost almost the entire 22 years. USDJPY did the same on old data: M30 PF 1.05 and H1 1.11 from 2004 to 2019, both beating 97% of random copies.
It also held after the author published the current version. On Exness gold H1, trades before 19 May 2022 made PF 1.46 (354 trades) and trades after made 1.27 (762 trades). USDJPY H1 went from 1.28 to 1.16.
What are the best Super Scalper settings?
The defaults, on H1. A March 2022 update added the choice of RMA, SMA, EMA or WMA smoothing and told people to "select a moving average which is giving highest profits." It doesn't matter much: on gold H1 all four gave PF 1.24 to 1.31, on USDJPY H1 1.16 to 1.22. Picking whichever did best on your chart is just fitting noise.
Then I took each layer away to see which one matters:

- Remove the double RSI filter and gold H1 drops from 1.31 to 1.01. Gold M30 drops from 1.23 to 0.97. The big-candle rule on its own is roughly break-even.
- Switch on the stop and target the author calculated (2 ATR beyond the signal candle, target 5 ATR) and gold H1 drops to 1.16, USDJPY H1 to 1.15. On M15 it doesn't rescue anything. I checked those exits minute by minute.
So the 2022 change to a double RSI really did make it better, at least on gold and yen. The stop-loss code would make it worse if anyone wired it up.
Gold H1 by year
| Exness gold H1 | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 | 2026 (to Sep) |
|---|---|---|---|---|---|---|---|
| Trades | 123 | 170 | 172 | 201 | 175 | 157 | 118 |
| PF | 1.69 | 1.24 | 1.00 | 0.83 | 1.32 | 1.63 | 1.32 |
| Net | +611 | +227 | +4 | −227 | +407 | +1,284 | +1,103 |
Less lopsided than most of this series (MACD ReLoaded made 75% of its gold profit in 2026 alone). Still, 2022–2023 went nowhere. Over 2004–2026 on Dukascopy, H1 was positive in 14 of 23 years, with swings from −39% of price (2006) to +54% (2008). In 2013, when gold fell 28%, it made +29%, almost all of it from shorts (+30.5% short, −1.5% long). A short side that pays its way is rare in these tests.
Risk structure
- No stop loss. A position only closes on the opposite signal. A wrong-way trade can run for days.
- Win rate is low: 35% on gold H1. Profits come from a few long rides.
- The ten worst gold H1 trades were all in 2026, when gold moved 100+ USD in a day. USD per 0.01 lot, before swap:
| Entry | Exit | Side | Entry → exit price | Loss | Held |
|---|---|---|---|---|---|
| 23 Jan 2026 09:00 | 26 Jan 15:00 | Sell | 4,913.9 → 5,096.4 | −182.58 | 78 h |
| 2 Jun 2026 05:00 | 5 Jun 13:00 | Buy | 4,517.3 → 4,406.4 | −110.96 | 80 h |
| 9 Mar 2026 00:00 | 10 Mar 02:00 | Sell | 5,076.2 → 5,184.3 | −108.21 | 26 h |
| 26 Jan 2026 19:00 | 27 Jan 21:00 | Sell | 5,069.1 → 5,176.7 | −107.72 | 26 h |
| 17 Apr 2026 13:00 | 21 Apr 15:00 | Buy | 4,846.3 → 4,739.2 | −107.21 | 98 h |
| 1 May 2026 14:00 | 4 May 16:00 | Buy | 4,625.3 → 4,521.6 | −103.71 | 74 h |
| 4 Feb 2026 01:00 | 4 Feb 16:00 | Buy | 5,015.8 → 4,920.7 | −95.17 | 15 h |
| 15 Jan 2026 14:00 | 20 Jan 04:00 | Sell | 4,588.5 → 4,680.2 | −91.78 | 110 h |
| 21 Jan 2026 20:00 | 22 Jan 16:00 | Sell | 4,781.3 → 4,871.9 | −90.63 | 20 h |
| 20 Mar 2026 08:00 | 20 Mar 14:00 | Buy | 4,706.4 → 4,616.3 | −90.16 | 6 h |
At current gold prices, one bad trade costs around 180 USD per 0.01 lot. Size for that.
- What isn't a red flag: no martingale, no grid, no averaging down.
Who should use it?
If you trade gold or USDJPY and want a simple mechanical system on H1, Super Scalper with the default settings is worth paper trading. It doesn't repaint, the logic is three lines, and it held on 16 years of data it had never seen. Expect long flat stretches, a 35% win rate, and years like 2023.
Don't run it on M5 or M15, whatever the name says. And don't wire up the stop loss in the code.
What I couldn't verify
- TradingView's own report. My numbers come from my port, not TradingView's tester.
- NIFTY. The author's chart is the Indian NIFTY 50 index on 5 minutes. I don't have NIFTY tick data, so I can't say how it did there.
- Forex swap. Not included for USDJPY or the other pairs. While US rates sit above Japan's, longs usually earn swap and shorts pay it, so on a two-sided system the effect partly cancels. I haven't measured it.
- Live fills. These are bar-open fills with real spread, not live execution. On H1 a few seconds of delay matters less than on M5, but I haven't measured it.
Bottom line
The TradingView Super Scalper strategy doesn't repaint and, as a scalper, never got above PF 1.01 on any market I tested. Move it to H1 and gold made PF 1.27 after swap from 2020 to 2026 and 1.10 from 2004 to 2019, which makes it the most durable result in this series so far. Thin and bumpy, but real enough to keep watching.
This is part of a series where I port the most-boosted open-source TradingView strategies and test them on broker ticks with real costs. For a strategy that held on Bitcoin but faded each year, see the MACD ReLoaded backtest; for one where a chart setting creates fake profits, the UT Bot backtest.
I'm not affiliated with TradingView or VinayKumarKV. Past results, including backtests, don't guarantee future results.
Comments
No comments yet. Questions about the settings, the data or the numbers are welcome.
Log in or create a free account to comment.