Best indicator for day trading: compare the rules

One fixed intraday sample

These eight five-minute bars are hypothetical, not a historical market or a performance record. Times label bar opens; each value is available only after that bar closes. The deliberately short three-bar lengths make the arithmetic visible and are not recommended trading settings. Prices are arbitrary units and volume is an assumed quantity.

• EMA(3): use close, seed with the first close, then average the current close and previous EMA with equal weights. This is a declared demonstration seed; earlier chart history would change the values.
• VWAP: start at 09:30 and accumulate ((high + low + close) / 3) × volume, divided by cumulative volume. Do not reset within the sample.
• RSI(3): seed average gains and losses with the first three close-to-close changes. Then update each average as (2 × previous average + current gain or loss) / 3; calculate 100 − 100 / (1 + average gain / average loss).
• ATR(3): true range is the largest of high − low, absolute high − previous close and absolute low − previous close. Use high − low on the first bar; seed ATR with the first three ranges, then use (2 × previous ATR + current range) / 3.

What each tool says about the same bars

Values below are calculated from the stated inputs and rounded only for display. "Not ready" means the seed does not yet have enough observations. None of these readings includes an order or a fill.

• At the close of the 09:50 bar, price is below both EMA and VWAP while RSI remains above 50. Requiring all three to agree would delay a bearish decision relative to using the price references alone.
• At the close of the 09:55 bar, price is above VWAP but below EMA. The two averages have different weights and memory, so a disagreement does not establish that either calculation is wrong.
• The final bar closes lower while ATR rises from about 0.70 to 0.77. The larger range describes volatility, not a separate sell signal.

Compare strategy variants on a fixed historical sample

The following is an unexecuted research protocol. Use NYSE Arca SPY five-minute standard candles, regular sessions only, America/New_York time, and one recorded data feed and adjustment setting. Use January 2 through March 31, 2025 for development, then April 1 through June 30, 2025 as an untouched check. Keep holidays and shortened sessions in the record; exclude shortened sessions from this protocol before seeing their results. If those dates are unavailable, declare a new range before testing any variant.

• Load at least 200 regular-session warm-up bars before the start. Keep EMA(20), RSI(14) and ATR(14) continuous across sessions; reset VWAP at the first regular-session bar. Use close for EMA/RSI and (high + low + close) / 3 for VWAP.
• Common entry event: while flat, require the completed close to exceed the previous bar high. Evaluate signals on bars closing from 10:00 through 15:30. Submit a long market order for the next bar open; allow one position and no more than three entries per session.
• Use a fixed 10-share research position in every version. Exit at the next bar open after a close below the previous bar low, or after the bar ending at 15:55. Do not enter outside the window or carry a position overnight.
• First compare the baseline with each filter separately. Record net result, maximum drawdown, trade count, average net trade and exposure on development and untouched periods. This is a comparison design, not a position-sizing recommendation.

Account for delay and overlapping signals

Adding EMA, MACD and several momentum oscillators does not create independent evidence by counting votes. They can reuse the same price moves. After testing individual additions, combine only the candidates that answer a distinct question, and compare the combination with its simpler predecessor on a fresh period.

• A five-minute bar-close decision waits for that bar to finish. Shorter indicator lengths do not remove this confirmation delay.
• An unfinished hourly filter can change even after a five-minute bar has closed. Decide whether the strategy uses the last completed hourly bar and accept the resulting delay.
• Pivot-based divergence needs later bars to confirm the swing. Record the signal at confirmation time rather than at an earlier, back-plotted pivot.
• Retain the timestamps of accepted and rejected entry events. This reveals whether an added filter reduces losses, misses moves or merely lowers the trade count.

Include costs and define when not to trade

Commission, spread and slippage can outweigh a small intraday move. For an illustrative 100-share round trip, assume $0.01 spread, $0.005 slippage per share on each side and $1 commission per order. Total friction is $4, or $0.04 per share. A gross $0.05 move leaves $0.01 per share before other charges. These are assumptions, not a broker quote or a forecast.

• For the fixed test, record the broker commission schedule and a declared spread/slippage assumption before selecting parameters. TradingView exposes commission and slippage settings; an OHLC bar does not tell you the actual bid-ask spread.
• Run the same variants again with doubled execution-cost assumptions. If average net trade becomes nonpositive, treat the apparent advantage as too sensitive to the chosen costs for deployment.
• Skip a signal when required data is missing, indicators are not ready, or the session/reset definition is uncertain. Do not silently substitute a different feed or extended-hours VWAP.
• For an ATR-stop implementation, define a maximum allowed spread relative to the stop distance before live observation, for example 10%. If historical bid-ask data is unavailable, mark this filter as untested rather than claiming the candle backtest applied it.
• For this example, skip new entries if price crossed the session VWAP at least four times in the previous six completed bars. This is a hypothetical chop filter to test separately, not a validated threshold; keep its rejected events in the log.

Turn the comparison into adjustable rules

Use Pineify AI Coding Agent to draft the chosen entry, exit, session and cost rules as a Pine Script strategy. The Visual Editor is another route for combining supported indicators and conditions. After validating the script in TradingView, Strategy Optimizer can vary exposed inputs and collect results. Set cost assumptions before optimization, reserve untouched data and inspect the generated orders; an optimized backtest does not establish future returns.

Scope and risk

This guide is educational and is not investment advice. The eight-bar example is synthetic, and the historical protocol has not been executed for this page. It establishes no indicator ranking, win rate or expected return. Intraday trading can lose money, and simulated orders can differ from actual fills.