6 months ago I learned about algo-trading and for the past months I've been working pretty much full-time. I have developed a strategy using a combination of several indicators, with a novel exit strategy (something I have at least never seen anyone doing, its using an indicator that is not designed all to be used as an exit). My entry is also using an indicator which is quite unknown (I had to code it from academic papers research with the help from my brother who is a mathematician/software dev).
I trade only in intervals of 4h and sometimes 6h (just to double check) I only trade cryptos.
As you may know with only being 6 months at this I am still an amateur, but I wanted to share some data with you guys just so I could hear what you think, my main fear is overfitting although I have tried to use 80% of sample data against 20% out of sample with very similar results, but it doesn't feels reassuring for some reason
As an extra test I have checked across multiple crypto-ticks using a (one size fits all settings) and while results vary considerably there is absolutely no losers (no matter what tick) every coin that I pick (tested on 12 of them) is a winner (some more, some less, with the absolutely worse coin having a -45% Drawdown and 300% profit in the past 2.5 years or so with the average being around 600% and 20% Drawdown)
I have optimized BTC/USD with changing parameter settings (changes in parameters dont vary drastically from each other but I do imagine even the slightest change might lead to overfitting).
My back tests (with optimization to each coin) for BTC/ETH In the past 5 years (2015-2020) are as follows1)BTC
In case you wonder there is no repainting, I am using tradingview to backtest but I am certain the results are exactly what I would achieve when doing it manually. I have not yet tested the strategy in real life data
Are there any steps you would follow to try and spot overfitting, any tips or suggestions ? I truly don't know what next steps I should follow to try and identity if I am onto something or I am just fooling myself.
Does it helps if trading intervals are at 4h-6h and that there is a standard configuration that I can get to work across with most coins ? (Of course not with the above results, but still with promising ones)
Submitted October 17, 2020 at 06:58PM by thepwnman33