In one paragraph
The desk took over 600 trading strategies published by their authors on TradingView, compiled each one to Python, and ran all of them across 13 liquid coins on hourly, 4-hour and daily bars from October 2021 to June 2026, on Bybit spot data: 47,242 backtests. 599 strategies produced a valid result. 273 were gross-profitable on at least 12 of their coin-timeframe cells and were replayed cell by cell, long-only, with a 0.12 percent per side fee charged on every fill and six credibility gates applied to every cell. 22 cleared the robustness floor that the desk set before looking at the results. The method is below, then the aggregate results, then the table of every strategy, which you can download and cite.
The universe
Each page in the strategy library is one of these strategies, read from its code. Filed by what the code does:
The raw screen: how often a published strategy makes money at all
The first pass ran every strategy as published, at zero commission, on every cell it could run on. The question was simply how many cells came out gross-profitable. The median strategy was gross-profitable on 40.5% of the cells it traded; the distribution:
Zero commission flatters everything, which is why this pass was only a filter: a strategy that cannot make money before fees on at least 12 cells was not replayed.
After fees: the 273 that were replayed
Every cell of every candidate was replayed long-only with the 0.12 percent per side fee charged on each fill. Of the 267 candidates in this table, 234 had at least one net-positive cell after fees, the median candidate was net-positive on 16.0% of its cells, and only 6.7% of candidates were net-positive on more than half of them. Fees are where most published strategies stop being strategies. The medians below are taken, as the method says, over a candidate's credible cells where it has any and otherwise over its net-positive cells, so they describe a candidate's good cells, not its average one; the median of those medians was +3.5%:
A cell only counted as credible if it passed every gate at once: at least 50 trades, net-positive after fees, a profit factor between 1.05 and 30, a drawdown between 3 and 100 percent, no single trade worth more than half the profit, and a mark-to-market drawdown no more than four times the closed-trade one. Credible cells per candidate:
Where the after-fee cells were positive, by timeframe
The share of candidate cells that were net-positive after fees, by the bar size they ran on. Slower bars pay less in fees per unit of return, and it shows:
What the strategies are built on
Read from the compiled code of the 614 strategies in the library, the indicators that appear most often. Each has its own page, with the calculation, how it is read, where it fails and the strategies that use it:
- Exponential moving average (EMA): what it measures and where it misleads 325 strategies
- Simple moving average (SMA): the plainest average, and why it still runs half the strategies 312 strategies
- ATR (average true range): the volatility number behind most stops 295 strategies
- RSI (relative strength index): the formula, the 70/30 rule, and why it misleads in trends 189 strategies
- Standard deviation in trading: the volatility measure inside Bollinger Bands and z-scores 110 strategies
- Donchian Channels: the highest high and lowest low, and the breakout rule they define 105 strategies
- Pivot points: swing pivots and floor-trader pivots, two different things with one name 94 strategies
- ADX and the directional movement index: how strong the trend is, not which way 80 strategies
Method, in full
- Universe. Over 600 community strategies published as Pine Script on TradingView, compiled to Python with PyneComp and run under PyneCore, a re-implementation of the Pine execution model.
- Data. Bybit spot 1-minute candles for 13 liquid coins, aggregated to 1h, 4h and 1D (sub-hourly bars for some cells), October 2021 to June 2026.
- Raw screen. Every strategy × coin × timeframe cell, zero commission, as published. 47,242 runs; 599 strategies with at least one valid result.
- Breadth filter. Gross-profitable on at least 12 cells, so that a single favourable coin could not carry a strategy. 599 became 273.
- Replay after fees. Each candidate's trade logs replayed long-only with 0.12 percent per side applied to every fill; the same arithmetic produces every after-fee figure on this site.
- Credibility gates. All of the gates listed above, applied per cell. Medians are taken over a candidate's credible cells when it has any, otherwise over its net-positive cells.
- Robustness floor. At least 4 credible cells, at least 8 net-positive cells, median Calmar of 1 or more; set before the results were read. 22 candidates cleared it. What the desk did with those is the desk's own and is not in this dataset.
- Limits. One venue's data; one window that contains one full cycle and one down year; long-only replay; compilation can lose a detail of a script. Nothing here is a forecast or a recommendation.
Download and cite
Two tables, under CC BY 4.0: use them, quote them, redraw them, with a link back to this page.
- strategies.csv 614 rows: the raw-screen and after-fee figures for every strategy in the library, with its page
- indicators.csv 31 rows: how many strategies use each indicator
Suggested citation:
CopyOrders (2026). The TradingView strategy screen: over 600 community strategies, 47,242 backtests on crypto. https://zerobit-site-g094.onrender.com/research/the-screen/
Every figure on this page is a backtest: hypothetical, not a forecast, and not a recommendation to trade any strategy. Strategy names are their authors' and the authors are not affiliated with the desk.