The same 14 strategies, each run twice on the same daily data from 2016 to 2026: once with its short side on, once long only. Long only made more in 34 of 36 pairs and had the smaller worst drawdown in 33. The short side paid in 2018 and 2022, the years the index fell, and cost money in the rest
Each pair is one Pine v6 script from the strategy library on one symbol and the same history, run once with its short side switched on and once long only. Long only made more in 34 of 36 pairs, made money in 35 of 36 against 17 of 36 with the short side, survived its held-out third in 29 of 36 against 13, and had the smaller worst drawdown in 33 of 36. The median pair gave long only 78.6 more points of return.
The 2 exceptions: RSI(2) 10/90, 200-day trend, 5-day exit on QQQ, +32.1% with the short side against +29.9% long only; EMA pullback 20/50 on SPY, +1.9% with the short side against -7.4% long only.
SPY returned +277.8% over the period. Shorting the dips of an index that kept rising is where the money went: the 1,769 short trades inside these runs won 30.0% of the time and returned -1.20% each on average, while the 1,884 long trades won 51.1% and returned +2.13%.
Each result page's lead run, with its short side on and long only, 4 January 2016 to 23 October 2026.
14 strategies, SPY daily data. With the short side on, 4 of 14 made money; long only, 13 of 14 did. Holding SPY itself returned +277.8%, more than any of them.
The trades inside the runs traded both ways, short and long apart.
| Trades | Count | Won | Average trade after costs |
|---|---|---|---|
| Short trades | 1,769 | 30.0% | -1.20% |
| Long trades | 1,884 | 51.1% | +2.13% |
The two-way versions also traded about twice as often, a median of 94 trades against 50, and were in the market a median 60% of the time against 37%. Every extra trade paid the commission and the slippage, and the median worst drawdown went from 21.5% long only to 41.2%: the short side added losses in the rallies more than it cushioned the falls.
In how many of the 36 pairs the version that also shorted returned more, year by year.
| Year | SPY | QQQ | Short side did better | Short trades won | Average short trade |
|---|---|---|---|---|---|
| 2016 | +11.2% | +8.1% | 6 of 36 | 28.1% | -1.30% |
| 2017 | +18.5% | +30.3% | 1 of 36 | 18.9% | -1.73% |
| 2018 | -6.9% | -2.7% | 21 of 36 | 38.0% | -0.56% |
| 2019 | +28.6% | +37.3% | 2 of 36 | 23.1% | -1.23% |
| 2020 | +15.1% | +45.1% | 5 of 36 | 29.9% | -1.61% |
| 2021 | +28.8% | +28.6% | 3 of 36 | 25.2% | -1.60% |
| 2022 | -19.9% | -33.7% | 25 of 36 | 52.7% | +0.95% |
| 2023 | +24.8% | +54.8% | 3 of 36 | 25.9% | -1.68% |
| 2024 | +24.0% | +27.0% | 2 of 36 | 25.9% | -1.76% |
| 2025 | +16.6% | +20.4% | 7 of 36 | 27.7% | -1.78% |
| 2026 (to 23 October) | +11.2% | +14.9% | 4 of 36 | 27.9% | -1.32% |
The short side earned its place in 2018 (SPY -6.9%, better in 21 of 36 pairs) and 2022 (SPY -19.9%, better in 25 of 36 pairs). In every other year it did better in 7 of 36 pairs at most. Insurance that pays in the falling years has a price in the rising ones, and this decade had far more of the second.
Net return after costs over the whole period, both ways and long only.
| Strategy | Symbol | Long and short | Long only | Drawdown, both / long | Held-out, both / long |
|---|---|---|---|---|---|
| Bollinger Bands 20, 2.0 | QQQ | -40.7% | +3.4% | 52.9% / 22.5% | did not survive / survived |
| Bollinger Bands 20, 2.0 | SPY | -28.2% | +2.0% | 38.5% / 24.3% | did not survive / survived |
| Donchian breakout 20-day entry, 10-day exit | QQQ | -32.2% | +100.6% | 38.6% / 12.7% | survived / survived |
| Donchian breakout 20-day entry, 10-day exit | SPY | -30.0% | +24.8% | 39.8% / 17.4% | did not survive / survived |
| EMA crossover 9/21 | QQQ | +23.6% | +61.4% | 28.8% / 19.9% | decayed / survived |
| EMA crossover 9/21 | SPY | +11.7% | +58.2% | 19.4% / 13.6% | decayed / survived |
| EMA crossover 20/50 | SPY | +14.7% | +28.6% | 24.7% / 16.7% | survived / survived |
| EMA pullback 20/50 | QQQ | +10.5% | +13.6% | 31.1% / 32.5% | survived / survived |
| EMA pullback 20/50 | SPY | +1.9% | -7.4% | 13.7% / 16.9% | survived / did not survive |
| Golden cross 50/200 | AAPL | +35.9% | +280.6% | 65.7% / 31.7% | did not survive / too few trades to judge |
| Golden cross 50/200 | MSFT | +72.6% | +234.5% | 56.4% / 28.3% | did not survive / survived |
| Golden cross 50/200 | NVDA | +2259.6% | +4114.9% | 49.1% / 37.5% | too few trades to judge / too few trades to judge |
| Golden cross 50/200 | QQQ | +166.5% | +253.1% | 29.4% / 28.4% | too few trades to judge / too few trades to judge |
| Golden cross 50/200 | SPY | +20.1% | +95.4% | 42.0% / 34.2% | too few trades to judge / too few trades to judge |
| Keltner Channel 20, 2.0 | NVDA | +272.9% | +987.2% | 46.8% / 25.4% | survived / survived |
| Keltner Channel 20, 2.0 | QQQ | -42.7% | +51.5% | 50.1% / 16.9% | did not survive / survived |
| Keltner Channel 20, 2.0 | SPY | -34.9% | +12.8% | 47.2% / 17.6% | did not survive / survived |
| MACD 12/26/9 | QQQ | -34.3% | +47.6% | 52.7% / 29.0% | did not survive / survived |
| MACD 12/26/9 | SPY | -4.5% | +46.4% | 23.4% / 14.8% | did not survive / decayed |
| Mean reversion 20-day MA, 3% envelope | QQQ | -57.8% | +2.9% | 61.3% / 33.3% | did not survive / survived |
| Mean reversion 20-day MA, 3% envelope | SPY | -34.4% | +16.7% | 40.4% / 20.5% | did not survive / survived |
| Moving average 50-day EMA | SPY | -19.5% | +91.1% | 40.5% / 19.1% | survived / survived |
| Moving average 200-day SMA | SPY | +4.4% | +91.0% | 42.2% / 23.2% | survived / survived |
| Parabolic SAR 0.02 / 0.02 / 0.2 | SPY | -27.5% | +86.6% | 50.6% / 25.1% | did not survive / survived |
| RSI RSI 14, 30/70 | QQQ | -35.7% | +3.0% | 46.0% / 22.5% | did not survive / too few trades to judge |
| RSI RSI(2) 10/90, 200-day trend, 5-day exit | QQQ | +32.1% | +29.9% | 17.3% / 10.2% | survived / survived |
| RSI RSI 14, 30/70 | SPY | -30.4% | +14.1% | 34.3% / 14.9% | did not survive / survived |
| RSI RSI(2) 10/90, 200-day trend, 5-day exit | SPY | +20.3% | +25.5% | 12.2% / 13.7% | survived / survived |
| Stochastic 14, 3, 3, 80/20 | QQQ | -73.2% | +26.1% | 77.3% / 31.3% | did not survive / survived |
| Stochastic 14, 3, 3, 80/20 | SPY | -54.2% | +28.0% | 56.9% / 19.5% | did not survive / survived |
| SuperTrend ATR 10, factor 3 | NVDA | +66.6% | +3902.6% | 76.9% / 41.8% | decayed / survived |
| SuperTrend ATR 10, factor 3 | QQQ | -8.0% | +194.0% | 35.8% / 20.2% | survived / survived |
| SuperTrend ATR 10, factor 3 | SPY | -17.8% | +100.1% | 47.3% / 18.1% | survived / survived |
| Trailing stop 9/21 cross, 3 ATR trail | NVDA | -32.1% | +551.4% | 72.3% / 42.7% | did not survive / survived |
| Trailing stop 9/21 cross, 3 ATR trail | QQQ | +79.8% | +239.0% | 30.8% / 24.9% | survived / survived |
| Trailing stop 9/21 cross, 3 ATR trail | SPY | +79.8% | +147.4% | 24.0% / 17.3% | survived / survived |
The day-trading rules were run both ways too, on 5-minute data over the last 600 sessions. Every one of them lost money either way; long only lost less in 4 of 4. With hundreds of trades, 0.15% a round trip decides the result before the direction does.
| Strategy | Symbol | Long and short | Long only | Drawdown, both / long | Held-out, both / long |
|---|---|---|---|---|---|
| VWAP 2-candle confirm, 1.5 ATR stop | SPY | -98.3% | -85.8% | 98.3% / 85.9% | did not survive / did not survive |
| VWAP 2-candle confirm, 1.5 ATR stop | QQQ | -99.1% | -88.4% | 99.1% / 88.4% | did not survive / did not survive |
| Opening range breakout 09:30–10:00 range, out by 15:55 | SPY | -52.3% | -36.7% | 54.1% / 37.6% | did not survive / did not survive |
| Opening range breakout 09:30–10:00 range, out by 15:55 | QQQ | -51.1% | -28.4% | 57.1% / 31.9% | did not survive / did not survive |
Describe it in your own words or paste the script. The Free plan writes the Pine v6 script with the AI, runs it on any US stock or ETF with the same costs and the same held-out check, and shows every trade on the chart. No card.
Every pair comes from a strategy result page, with every trade and the script as it ran. The same runs also show why a high win rate can still lose money.
In these backtests it was. Daily data, 4 January 2016 to 23 October 2026, the same rule on the same symbol with the same settings: the long-only version made more than the version that also shorted in 34 of 36 pairs, made money in 35 of 36 against 17 of 36, and had the smaller worst drawdown in 33 of 36. On the 14 strategies' lead runs on SPY, 4 of 14 made money traded both ways and 13 of 14 long only. That was a decade in which SPY returned +277.8%; in the years it fell, the answer changed.
Because the index kept rising. The 1,769 short trades inside the two-way runs won 30.0% of the time and returned -1.20% each on average; the 1,884 long trades won 51.1% and returned +2.13% each. The two-way versions also traded about twice as often (a median 94 trades against 50) and were in the market a median 60% of the time against 37%, paying costs on every extra trade.
In the years the index fell: 2018 (SPY -6.9%), when the version that also shorted returned more in 21 of 36 pairs, and 2022 (SPY -19.9%), when the version that also shorted returned more in 25 of 36 pairs. In 2022 the short trades won 52.7% and returned +0.95% each on average. In every other year the version that also shorted returned more in at most 7 of 36 pairs. A short side is insurance that paid in the falling years and cost money in the rising ones.
In 33 of 36 pairs it did. The median worst drawdown was 21.5% long only against 41.2% for the same rules traded both ways: the short side added losses in the rallies instead of cushioning the falls.
The last third of every run's history was judged on its own, with no fitting on it. The long-only version survived its held-out third in 29 of 36 pairs, the two-way version in 13 of 36. Survived means a positive Sharpe ratio on the held-out part at least half the fitted period's.
The day-trading rules lost money either way. On 5-minute data over the last 600 sessions, 4 pairs were run and 0 made money in either direction; the long-only version did better in 4 of 4. With hundreds of trades, costs decide those runs before the direction does.
With the free Alpha Charts tester's own engine, as on every result page: the library's Pine v6 scripts with their textbook settings, each run once with its short side on and once long only, on Alpaca daily data from 4 January 2016 to 23 October 2026, split-adjusted and not dividend-adjusted, one position at a time with the whole account, fills at the next candle's open, 0.05% commission and 0.025% slippage on every fill, the last third of the history held out. Nothing was tuned.
Every number here is the output of the Alpha Charts backtest engine on 25 October 2026, on Alpaca market data, for the scripts on the strategy result pages. Past results are not a forecast and nothing here is investment advice. TradingView and Pine Script belong to TradingView, Inc.; Alpha Charts is not affiliated with it.