Some more analysis of my process Friday:
- I was expecting a retracement of up to 900 USD from peak. I was expecting a not too bad day given positive mood from MSFT report.
- I was monitoring account NAV throughout day. Looked fine and as expected until market open.
- Checking back maybe one hour after open, I was already 500 down from peak IIRC.
- Checking at before 6pm local time, I was down 900 from peak. This is where I would have added if I didn't prioritize my GFs stuff...
- Price from there went up to 700 from peak (which on the position was 300 unrealized). Because I am in the blind while doing this, this calmed me down.
- Checking back one hour before close it had started to look quite bad at 1200 from peak. Made a counter trend bet, which I think was fine on its own.
- As it turned out, market continued dropping in afterhours trading. So now down 2000 due to this algo from peak.
The reason I am shell shocked is not because of algo peforming worse than expected (it isn't, aside from warm up issue). Rather, it's from doubling account and then using up 3/7 of the YTD realized profit in a day. Again, with the kind of leverage I'm doing, 2.5x on NQ, I need to tolerate this. It can get way worse in a day than this, obviously. The principle of being able to sleep at night is a good one (I feel good now, but I definitely wasn't happy earlier).
Currently, shutting down algo requires logging into an external site and I can't do it from phone. This is a risk I am aware of. The trouble is, shutting down algo blind is basically pointless since I don't have any context. And why be so dependent on discretionary actions anyway... that means I have no edge.
The bigger picture:
- With this kind of leverage, timing is crucial. I am late into the race and betting offensively. I'm doing something comparable on a Swedish margin acc where I am buying equities after wide market dips and then gradually selling them off (because I don't have time to immediately buy more as I spend most attention on other accs), but thus are avoiding these huge drawdowns.
- My main motivation for (until end of July) increasing prices is institutions releveraging into US indices in a feedback loop. Higher volatility throws a spanner in the wheel here and risks causing a feedback loop in other direction until Fed intervenes.
Actionables:
- Monday strategy will be to set an upper limit on how bad this is allowed to get. It's the classic circumstance where it's much easier to stomach in backtest than in real life. 10% drawdown is reasonable for what I'm doing in normal cases and 20% in bad cases; this means I should shut up until being -$4500 (coincidentally where I'm starting to go red for year, do recall completely change of strategy in June and subsequent deposits in July, so this is less of a deal than it sounds, but as noted earlier it feels terrible psychologically).
- Hurry up fixing the warmup issue previously described in thread as currently it's clear that even if there was high volatility in more than a few days before starting algo, it's not in warmup window. That's why I am at max leverage now in fact (also because I'm greedy and letting this happen, and the algo logic which is supposed to keep my greed in check isn't helping when not warmed up).
- RESEARCH: Changing the leverage to lock in profits might is something I should look into. The trouble is I have experimented with this before and it sounds like an easier thing to do than it is. I don't think it will work, but at least check.
- RESARCH: It would be interesting to investigate market wide outperformance/underperformance based on high profile large cap companies post earnings report afterhours trading after a function of time. Is it ever reasonable to expect positive MSFT report to give any edge that much later? I bet someone already did this though, so best to dig for papers on e.g. Quantocracy.
Fixed a bug so I am happy. Although the quality of realtime data stream is disconcerting.