It is well-known that a doji at a downtrend may indicate a trend reversal, i.e. we should consider a long entry if we notice it. On the other hand, there is no guarantee and the doji may be just a short stop before a sell-off.
Therefore, I used another strategy: I state that there must be a bullish action on the day after a doji. It means that the day closes above the high of the day with a doji, as well as the candle has a long bullish body (“long” means the longest of the past 3 days excluding the doji). This is depicted in the figure below:
Alternatively, we can wait an extra day and only enter if also the second day closes above the previous high. This would result in a lower income per trade because we enter later, but such trades may be safer.
The results of backtesting in years 2000-2017 for the Russell 3000 stocks are collected in the table below. The first column (“unconfirmed”) refers to the first case where we enter as shown in the figure above. The second column (“confirmed”) is the second scenario where we wait one extra day.
|Total net profit||$88292||$19837|
|Number of trades||1109||266|
|Profitable trades [%]||55.00%||65.04%|
|Average trade [%]||0.85%||0.75%|
|Number of days held||20||24|
The first conclusion is that both scenarios lead to positive results. The first scenario is more risky; see the higher maximum drawdown but the total profit is much larger due to a high number of trades. Also, an average trade is greater for the first scenario, as mentioned previously.