Why an indicator on the live feed differs from one on history

[Fairman]

Why an indicator on the live feed differs from one on history

An indicator drawn on historical bars is complete. The last bar on a live chart is still forming. Its values change with each tick.

Some indicators repaint. A signal appears while the bar is open and disappears when it closes. The historical chart shows only the signals that survived.

A trader watching live sees more signals than the history suggests, and many of them vanish. The backtest on history looks cleaner than the live experience.

Testing an indicator for repainting is straightforward: record its signals live for a few days and compare with how they appear on the chart afterwards.

Indicators that use only completed bars do not repaint. Those that use the current bar might.

Some platforms also recalculate indicators after reconnecting, redrawing recent bars with slightly different data. A chart that looks different after a restart is a hint that the indicator or the history has changed. Saving a screenshot of live signals alongside the timestamp creates a record that the later chart cannot overwrite.

Does the indicator in use repaint, and was the live signal log ever compared with the history?

[Shadow Trader]

A signal appears on the chart, a trade is taken, and later the indicator's history shows the signal was never there, or was in a different place. This is repainting, and it's a quiet disaster for anyone relying on indicator signals.

Some indicators recompute their entire history with each new bar or tick. Zigzag-style tools are a classic example: the last leg changes until a new swing is confirmed. Others use future bars in their calculation, which is only possible in historical data.

Looking at a chart afterward, the signals look perfect. The indicator seems to call every turning point. In real time, those same signals were unstable or absent.

The test is simple. Run the indicator live on a demo for a day, take screenshots at each signal, and compare them with the chart at the end of the day. Any difference is repainting.

Even non-repainting indicators can mislead at the tick level, since a moving average of the current forming bar changes every tick. Using closed bars only avoids it, with a delay as the price.

Which repainting indicator looked best in a backtest and turned out worst live?

[Fairman]

An indicator can show one value on the live chart and a different value for the same bar when you look back later. The most common reason is repainting, where the indicator recalculates past values as new data arrives. Some do it by design, like certain zigzag or fractal indicators that can only confirm a swing after later bars form. Others repaint because of coding choices, such as using the current bar's values in calculations for earlier bars. On a historical chart, repainting indicators look remarkably accurate, because every signal shown is the final version, not what you would have seen at the time.

[Amanhasnoname]

A few years ago I was convinced I'd found a great reversal indicator, because on history it called nearly every turn within a pip or two. When I put it on a live demo for a week, almost nothing matched. Half the arrows appeared and vanished within a minute, and the ones that stayed often showed up three or four candles after the turn. It was repainting, and the history had simply kept the survivors.

The test I use now is the one you describe, with one addition. I take a screenshot at every signal with the clock visible, and I run a script that saves the indicator buffer values on each new bar to a file. At the end of the week I can compare the saved values with the redrawn chart and see exactly which bars changed.

If the signals use only closed bars, I accept the delay of one candle. It costs me some entries, but I'd rather trade what I can actually see in real time than what looked good afterwards.