Experiment with dojiPercent to change the number of "dojis" detected. I will add doji sub-type indication if it appears to be meaningful.
By default, it plots the 5m (red), 15m (orange), 1h (yellow), 4h (green), and chart (cyan). If the chart timeframe is any of the configurable ones, the chart copy won't be drawn. (I might reverse that, so that cyan is always drawn.)
Since doji points are somewhat sparse, and the lookback is short (default 10), the EMA's make drastic corrections toward new indecision. (I'm not convinced the EMA's are useful and/or relevant.)
This works on any timeframe, but seems to work best on the 1D. (5m is somewhat irrelevant on the 1D, so there are tweaks to be made.)
Dojis from a timeframe are corrections to a doji trend from a higher timeframe.
Red corrects to orange, corrects to yellow, corrects to green.
If the chart timeframe is > 4h, the others will correct to cyan.
Otherwise, cyan will fit in-between the adjacent timeframes.
Multiple indecision candles within a short timespan forming sharp peaks indicate retests, backtests, rejections, and bounces off of support/resistance.
With a correct larger-timeframe channel, one would expect lower-timeframe indecision at/along typical levels.
Although the doji's have unpredictable wicks, the dots printed by this indicator do not. Matched with volume, they reveal the prices where the most violent battles between bulls and bears took place, and are likely to take place, again.
1) Put trends on the longest segments, then look for confluence along them, and/or near the intersections.
2) Use lower-timeframe doji trends to estimate the direction of the higher-timeframe doji trends, before they become detectable to Doji Trender. Confirm by looking for confluence where those trends intersect with horizontal support/resistance, this indicator, and/or others.
3) Notice that multiple legs on the same trend line are close to parallel, if not colinear.
4) Notice that many of the doji segments point toward (very-distant) future dojis.
5) Drop horizontal lines on the dots where we previously reversed, and find confluence in VRVP when we revisit them.
6) Create parallel (fib/whatever) channels that more-closely match MM's intent. The segments one uses to set the angle of the channel, and those used to align the channel, vertically, are not always the same:
a) Match the channel slope to as many doji slopes as possible, considering every trend.
b) Figure out where the channel actually belongs, re-considering every trend.
In true TradingView spirit, the author of this script has published it open-source, so traders can understand and verify it. Cheers to the author! You may use it for free, but reuse of this code in a publication is governed by House Rules. You can favorite it to use it on a chart.