Hi folks, an update from me on my $50K weekly gap strategy:
In relation to this strategy I have been pre-occupied with an issue I have mentioned a couple of times now in previous posts, namely the number of valid trades being triggered according to this strategy. To illustrate, I will share a couple of graphs from my 20 year backtest history for this strategy.
Firstly, the number of valid trades triggered during each month of the 20 year history:
Looking at the graph it appears as is if over the 20 year backtest period the average number of valid trades per month increases slightly - I put this down to surviroship bias, since the backtest was only performed on current ASX300 members, not all historical ASX300 members - hence as some of the current but newer stocks in the index were listed throughout the last 20 years, the number of trades has edged up.
But nevertheless, as the above graph shows, the highest number of valid trades triggered in any single month is 30, which is back in only 2014. In comparison, in the last 4 weeks or single month since I started trading this strategy live, it has triggered 35 valid trades - 5 more than the highest month in the last 20 years.
The pattern is even more interesting when I look at the 3 versions of gap setup in the strategy - the v2, v5 & v6. As per the current trades table below, the number of each type that has been triggered is v2 = 23, v5 = 11, and v6 = 1 (in the backtest period all 3 produced a roughly equal number of valid trades).
Now when I look at the history of valid v2 gap trades, the graph looks as follows:
What the above graph shows is that the average number of valid v2 gap trades per month is less than 5, and the maximum in any one month over the last 20 years is 14. Yet in the last month the strategy has triggered 23 valid v2 gap trades.
This is definitely the kind of red flag I look for when trading a systematic strategy - when live trading breaks 20 year records within the first few months of live trading. Looks suspicious and hence I have been focused on this apparent change in market behaviour (i.e. how is it so coincidental that the market makes such a dramatic change in behaviour the first month I am trading a new strategy???).
Well, I have reviewed my backtest data over and over several times, and can't find anything which would suggest I have done my sums wrong. So the market does seem to be behaving differently (rather than me having wrong backtest figures) - but is this a good or bad thing?
The results so far now included 21 closed trades, whose results are as follows (including strategy vs actual):
So the results so far are not great, basically just treading below water, but variances have put an almost $700 dent in the result to date.
So while it is tempting to conclude that the market has changed behaviour (the very month I start trading) with an extraordinarily high number of valid trades, when I look at the result of those trades (P/L) it is only just below breakeven and in fact the biggest contribution to the actual loss so far is execution variance (-$687) compared to the strategy loss (-$302).
When I analyse it further, I can see that 7 out of the 21 closed trades have variances due to delayed or missed entry because of the technical problem I have with my weekly screening process, and of the remaining 14 open trades delayed or missed entries affect 7 of those as well.
A secondary execution factor is higher than expected slippage on entries and exits because I have been executing manual limit orders rather than pre-entered contingent orders.
So while the market does seem to be behaving a lot more "gappy" than my 20 year backtest data shows, my analysis tells me that this doesn't appear to be acting "against" me. Rather, I have a couple of now known execution issues which are costing me money and which I need to fix. Thankfully now I finally have a fix for my weekly screening problem and I have started to execute trades this week using contingent rather than manual limit orders, so my focus now is to see over the coming weeks a significant improvement in strategy execution, and hopefully that will make a positive difference to results from the next month of trades (i.e. execution variance to strategy will be significantly reduced, I can't control what the market does...).
