Agree with the "edge internally", by which I assume you're referring to using your own privately generated data to make trading decisions (as opposed to public data). This is a Bayesian/path dependency way of thinking about the markets (older terms might be cybernetics/feedback control). No, this does not necessarily mean equity curve trading, but it might share some similar features/aspects with it. By equity curve trading I mean using an overlay, the kind of work Ernest Chan is doing lately with PredictNow AI (there's a Youtube channel).
New traders need to know how to code (C and python is a good combo). That's because most of the volume in the markets these days are just algos trading with other algos and your brain doing discretionary trading is definitely not going to beat them.
Focus on strategies with positive skew (the largest winning trade is vastly larger than the absolute value of the largest losing trade). You want large blowout winning trades and your largest losses should be relatively tiny in comparison.
All your strategies should be taker only. Don't waste any time trying to earn an extra tick or two by using maker orders and spending huge amounts of time on cost/latency reduction. Maker strats have limited capacity, you can't scale them up. If your ES average profit per contract trading 1 contract is only 4 ticks, that's what, like 50 contracts max capacity? To scale up you need lots of uncorrelated strategies and/or strategies with high average profit per contract.
Once out of sample is not good enough. You need to test the out of sample results out of sample as well. You need to verify if you just got lucky or if your algo does have predictive power.
