Well, coding is easy but interpretation isn't.
From 15+ years of full-time trading (and building & backtesting strategies), I've learned GENERALLY:
1. Strat building & backtesting depends on the cleanliness of the data - I can't vouch for TS's intraday data on cryptos which trade across various crytpo exchanges & networks generally.
2. Shorter timeframes for something like this are often LESS PROFITABLE, higher-cost, and more demanding to trade.
Are you going to stay awake at your trading-station 24/7 to execute EVERY TRADE as it's presented, when tests show max-consecutive winners=4 and max-consecutive losers=18? Unlikely. Or do you intend to automate the system, and TRUST IT to not go down while you're asleep, potentially costing you $1000s?
3. You have to make assumptions on what fills you can count on, etc. This code says, "if triggered, enter on next bar's open at market", which is considered a conservative coding technique.
4. In my experience, the biggest wins and easiest trades are those that you are comfortable sitting in for many days if not weeks - I still trade garbage pennystocks intraday but that's a different discussion.
Long-story short: I've run the same tests over your suggested timeframes but I didn't see what i would consider a worthwhile improvement overall.
News 
Market Data 
Discover 
