Yes… this style is also the bread & butter of equity pods at hedge funds like millennium, citadel, baly, etc.
Problem is that all prices are nonlinear. I tried this whole analysis for a while making the assumption that price was linear over very short periods and if you could estimate that period length, then you could use trend lines, linear regression, etc to determine trade exits, breakouts blah blah. Short version of results: garbage. It still sounds so good though, I'm enamored by the idea, but it's just false.
Yep! We like to say “stock prices are non-stationary”. You improve your analysis if you looked at log chg of price instead of absolute price. The old technical indicators are based upon this (oscillators, RSI, bollinger bands, etc) but don’t come calibrated properly. You actually generate a decent signal of momentum by using cross-sectional and/or a timeseries of rate of change. Bollinger bands on price chg are decent because they scale properly, etc. Again, not really a trading signal, but at least can corroborate a story. The sad truth is that a lot of “technical analysts” are really just math and stats illiterate. :/
Actually the burden of proof is that they do work not that they don't work. So you're in the clear. Raises the question is there a real study with performance results that support trend lines as useful?
Damn dude, does that really work? What frequency data do you use? I'm going ro backtest the f*** out of that simple ass sh*t and then write you a bot. Not sarcastic.
You've got too many posts to truly believe trendline is the only stat used to quantify stock trades in absence of fundamentals.
Definitely not knocking the longshort analysis, in fact much respect, but if I personally had to go through all that to decide to enter a trade, I would definitely not make it. Exhausting!
Dude, cmon, that's not how probability works. If a distribution has high probability of a negative low impact event and low probability of positive high impact events, totally reasonable for over all impact to be positive. Hit rate is basically irrelevant without knowing the impact of the event.
I mentioned previously that hitrate was in absolute terms. The hitrate of your signal absolutely needs to be 50%+. In stats terms, if your variable has low r2, what exactly is the use of it? And if you say “it improves multi-factor performance” then the question stands for the multi-factor signal you’re using. Also, in reality you’re not accounting for transaction costs and b/a incorporate the probability distribution of (at least) the base rate. For example, if your point was true then you should be a trillionaire by buying 0dte call options every day. You’re not because the price of the call option includes the market implied distribution, which means your ability to spot a “high reward low risk” or “low reward high risk” opportunity based upon price is gonna be less than 50/50.