Tax loss selling screen

Does anyone know how I can include only the dates December 15 through January 15 of each year for a screen (not a port/sim) in order to do a backtest?

Thanks for any help. Doug

Hi Doug, I haven’t used this, but maybe this works:

(Month =12 and MonthDay >= 15) OR (Month =1 and MonthDay <= 15)

I just did a quick run and it seems like it only selects stocks during those dates

Thanks, SJ, yes, that seems to work. I appreciate the help! I had just read that Morgan’s quant guru is recommending under-owned, small cap, heavily shorted, value laggards for the year-end/January effect.

In addition to December selling pressure on big % losers for tax advantage, your post reminds me to also consider that diminished December selling might be expected in big % winners possibly contributing to a runnup in price into yearend - then to be possibly followed by pent up selling pressure on previous year winners in January. I’m going to check my highest ranked positions for stocks fitting both these scenarios. Thanks for the tip!

I thought I’d provide an update on my tax-loss selling screen lessons. Caveat: these are all generalizations based on my specific screens that are all in-sample backtests based on only 15 years of data. This year could/will be totally different than history.

Sweeping generalizations:

The performance, historically, has resulted in big excess returns, but with above average drawdowns.

When screening for the worst performing stocks (that might bounce back), I found using a performance period significantly less than one-year the best.

Low priced stocks with relatively high volume backtest best.

The two weeks before and one week after the new year are by far the best performing weeks (Starting this Monday). The last three weeks in January are okay, but with huge potential drawdowns (Probably best to exit Jan. 8). February tends to be really bad for these stocks (reversion to the mean?).

Most fundamental factors were of no help (or a negative), except to remove the really “bottom of the barrel” companies (dayslate >1, bad accruals, etc, etc.).

Hope this might help someone.
Doug

Doug - given that you are targeting the really bad stocks, I suggest that before you start trading this, you might want to inspect the realized trades to see how many stocks are sold with “no Rank”. The issue is that stocks may be delisted but the trade is registered as the last price, not “0”. This tends to make the trading system look better than it should.

Thanks for the feedback. Is that an issue if I am using a screen and not a ranking system? I thought the 123 database did not have “survivorship bias”.

It is probably an issue either way. It isn’t a question of survivorship bias. It is a case of the software not being able to determine whether a stock has gone bankrupt or delisted for some other reason. The software solution was to assign the last trade price as the price for the sale of the stock. If the stock is delisted then you might be able to recover money by selling it OTC. Or you may lose your investment. It is probably not a major issue, I just suggest that you inspect the trades to make sure you you don’t have a lot of stocks dropping out of the universe. This could be the case if you are focusing on unwanted stocks… they may be unwanted for a reason.

SteveA

Thanks, Steve, I appreciate your detailed explanation. That issue was unknown to me. I’ll definitely research the realized trades. The good news is that I’m only looking at stocks that trade $300,000+ a day of value for a three-week window. I doubt many companies get delisted during the holidays, but it will be interesting to find out.

Quick update on results for this year. Caveat: small sample size ahead!

My tax-loss selling screen returned 20 percentage points in excess of the Russell 2000 over the period I recommend (Dec. 15 to Jan. 8), in-line with my in-sample 15-year backtest results.

My main lessons (in addition to those in my prior post):

Expect a huge variance in performance between stocks.

A modest number of stocks accounted for the excess returns and there was no clear rationale for those stocks that popped. In other words, invest equally across all candidates regardless of how “ugly” some appear!

20pp excess over the course of 3-4 wks? Wow - that’s super. Congrats on the success of the strat - and the guts to put money behind it!

Thanks, SJ! Next year I will have more confidence (and more guts) to allocate a larger amount of money. It is a fun strategy with some crazy companies.

Another caveat that I forgot to mention earlier: Logically, it makes sense that this strategy works best after a very strong year like 2019–more investors looking to offset gains.

fuddeld around for two hours, first results: