The 8-12% Loser, Gap Down Filter

Because of my limited day trades, I'm always looking for ways to develop good swing trades.

Swing trades are tough because essentially you're trying to predict what's going to happen overnight.

I mentioned "Red Floaters" after big run-ups, which can make for good shorts into the close.

However, just as you can use Johnny Vento's gap stat calculator for the probability of a gap-up, you can also use it for the probability of a gap-down.

So here's the idea behind this filter:

1. We want stocks that have decreased 8-12% during the day. We can't have it less than that because there's too many stocks that will decrease 5% or so. We don't want it more than that because a stock that loses 15-20% has a good chance of bouncing the next day.
2. We want stocks that have a past history of gapping down at least 70% of the time they've lost 8% of their value.
3. We want reasonably liquid stocks (trading > 100000 shares)
4. We want stocks that are not tied strongly to the overall market. We don't want a stock that is falling just because the market is falling. We want a stock that is falling on its own. Thus, it needs to have a Dow green/stock red correlation of more than 45%.

So here's the code. Please note that you need the "Advanced Stockfetcher" which is a bit more expensive than the standard. The standard Stockfetcher doesn't let you use complicated screens like this.

set {drop_threshold, 0.08} /* 8% drop */

set {OpCl , open - close}

set {OpClPct , OpCl / close}

set {met_down_threshold_Yesterday , count(OpClPct 1 day ago >= drop_threshold,1)}

set {gapped_up_today , count(open > close 1 day ago,1)}

set {gapped_down_today , count(open <>

set {met_down_threshold_then_gapped_up_today , met_down_threshold_yesterday * gapped_up_today}

set {met_down_threshold_then_gapped_down_today , met_down_threshold_yesterday * gapped_down_today}

set {met_down_threshold , count(met_down_threshold_yesterday >= 1,100)}

set {down_then_gapped_up , count(met_down_threshold_then_gapped_up_today >= 1,100)} set {down_then_gapped_down , count(met_down_threshold_then_gapped_down_today >= 1,100)}

set{gap_down_ratio, down_then_gapped_down / met_down_threshold}

set{gap_down_percent, gap_down_ratio * 100}

add column met_down_threshold

add column down_then_gapped_up

add column down_then_gapped_down a

dd column gap_down_percent

set{close_ind, ind(^DJI, close)}

set{open_ind, ind(^DJI, open)}

set{ind_diff, close_ind - open_ind}

set{ind_up, count(ind_diff > 0, 1)}

set{ind_down, count(ind_diff <= 0, 1)}

set{stock_diff, close - open}

set{stock_up, count(stock_diff > 0, 1)}

set{stock_down, count(stock_diff <= 0, 1)}

set{indup2, ind_up + 1}

set{inddown2, ind_down + 1}

set{stockup2, stock_up * 2}

set{stockdown2, stock_down * 2}

set{correlate_up, stockup2 + indup2}

set{correlate_down, stockdown2 + inddown2}

set{both_up, count(correlate_up equals 4, 100)}

set{both_down, count(correlate_down equals 4, 100)}

set{indupstockdown, count(correlate_up equals 2, 100)}

set{indownstockup, count(correlate_down equals 2, 100)}

set{indexupdays, both_up + indupstockdown}

set{indexdowndays, both_down + indownstockup}

set{indexupstockup, both_up / indexupdays}

set{indexupstockdown, 1 - indexupstockup}

set{indexdownstockdown, both_down / indexdowndays}

set{indexdownstockup, 1 - indexdownstockdown}

set{indexup_stockup, indexupstockup*100}

set{indexup_stockdown, indexupstockdown*100}

set{indexdown_stockdown, indexdownstockdown*100}

set{indexdown_stockup, indexdownstockup * 100}

add column indexup_stockdown

Open greater than Close

and OpClPct greater than drop_threshold

and OpClPct less than 0.12

and Volume greater than 100000

and Close less than 20

and Close greater than 1

and met_down_threshold greater than 10

and gap_down_percent greater than 70

and indexup_stockdown above 60

sort by column 8 descending

I ran this using Stockfetcher's "Backtest" function. I backtested it for the past 2 years. Only 19 stocks came up. The average return was 4.49%, and 12 out of 19 trades were winning trades. The reward/risk ratio was 3.65. The average max profit was 8.25%. A stop loss of 4% was triggered only 33% of the time. The average drawdown was -7.4%.

That's pretty good for a Stockfetcher backtest. It's a pretty stringent filter and you won't get many plays with it, but you'll get an occasional one.

EDITED TO ADD: I tweaked it so that the indexup_stockdown variable had to be more than 60. The win ratio improved to 75% of the trades as winning trades, an R/R ratio of 5.19 and an average profit of 6.14% (average max profit of 6.36%). Only 8 trades came up on the filter over 2 years, however.

HSNI came up on this filter yesterday. If you had shorted HSNI into the close near $2.80 yesterday, you would've had a 40% return on your position if you exited today at $2.