Sour day because a bunch of negative news coverage. Haven't been following live but that's how it looks like post-fact. And of course correction at parabola top like this was kind of expected.
NQ stop blown fast for minor loss, trying again a bit later. Leaving China position (much wider stop/lower trade size) open to next week. Will cut if it starts heading downhill by much tonight though (needing to be awake for other reasons).
Bleh crap China data with predictable results on my equity positions (stopped). Jobs report & stuff at 14:30 my time. Closing gold for 400 EUR profit before then as it's better to see what happens.
Overall, since I'm basically on the way to going backwards another month I think it's worth going ahead an implementing automated stop loss sizing since this is the thing that is ruining a lot of my trades as pointed out already in January. The inconsistent stop sizing is making is way harder to assess my trading. Currently I'm considering training an ML model to predict the stop size (in %) on all my successful trades, then having to scale positions to match to ensure constant risk. Might need to use weighting so that shorter term trades have higher weight, of course the easiest way to get a large win rate is having huge stops/risk... Potentially training another model on all my failed trades and only entering when that prediction is meaningfully different from the successful stop size prediction. The trouble is, if I was to update said models based on my latter trades based on them, there will be feedback that supervised learning typically poorly handles. (Because predicting stop sizes from predicted stop sizes converges to some BS number.) But hey, a finished model is a finished model and I won't necessarily have to update it. The question is just how representative my trades thus far will be to all my future trades. There is one problem here and it's that I have had very varying time frames in mind on trades on this account, which I can't (without heavily assuming I can do it live) add as an ML input feature. However, if automated stop loss sizing forces me into a consistent profitable behavior and time frame then that's a good thing. "ML" here might simply be linear regression or KNN lookup, we will see (should first look at what input features are relevant). Have quite few data points so the simpler the better. Maybe I can oversample into additional synthetic data points.
Buy China A50 x 2 @ 12119 stop 11961 World is not going to end until next week. Not thinking here is the new downtrend either.
Got a new aquarium two weeks ago and have been spending significant time installing it. Find it helps with the PTSD runs from trading; seeing the fish doing their own thing is very calming and allows switching focus from trading effectively.
Regret not being home in the afternoon to buy more before run up before open, but anyway... NAV 19579