SuperTrend (@SuperTrend76) header
SuperTrend (@SuperTrend76) profile picture

Systematic trader. Not a coder. Not a quant. Just a guy who built a machine. Full-time. Real money. Documenting the process

Florida, USA 🌴
Joined August 2022
Unless you are using TV to generate signals for you. Then you do need realtime data
A large number of @tradingview customers are non-professionals. In other words someone who is not a licensed financial expert or institutional trader—either an individual managing their own money or a non-finance worker using basic financial skills. Ignoring this and constantly putting new text and messing up end of day, end of week traders charts with (delayed data text) is clearly not understanding how to run a business. Or not knowing your client base needs. End of day trader/investors do not need real-time data. The guys that are trading with real-time data (large number) are already using Bloomberg and their companies are paying for the terminals.
34
I remember when the big argument was gold or bitcoin as the hedge not too long ago. So I tested $10K in each since January 2025: $GLD: $13,902 $SPY: $12,735 $BTC: $6,501 Gold is 31% off its high and still in front. It also beat $SPY overall since 2000.
80
This is what I think every day when I open X
Everytime I get worried that edge in trading is going to go away I just talk to normies for 20 minutes and realize if these guys ever hit the orderbooks it'll be free money
1
35
$NVDA is buying $10B of Anthropic stock apparently. Anthropic will use it to buy Nvidia chips. Next quarter Nvidia reports record revenue and the cycle continues...
94
The last time I watched money run from tech into boring stocks, I was working at a brokerage. It was the year 2000. From 1999 to the 2002 low, $XLK lost 82%. Staples' worst drawdown was 36%. I'm not saying this is 2000. Probably far from it, but just pointing out that boring isn't boring when it's working.
49
$MU closed back under its 200-day yesterday. It's 41% off its June high of $1,213 and 23% under its 50-day and still up about 150% this year. Wild
68
Everyone's saying Leopold's fund lost 92%, but that's not what a 13F shows. It's long US positions at quarter end, filed up to 45 days later. A drop can just mean he sold. And you're seeing it in November. This matters for backtesting 👇
1
41
Same problem shows up in other places: - earnings data dated to the quarter end, not the report date - index members using today's list - fundamentals that got restated later
1
7
This is what I do: every input gets the date it was actually public, and the signal trades the next bar after that. Strategies built on 13Fs or earnings data usually look a lot worse once you add the delay.
8
Every "this algo beat buy and hold on $NVDA" backtest has the same problem. Someone picked the stock after it went up 10x. Run the same rules on the 2021 leaders and on the names that got delisted, and most of them fall apart. That's why I test my rules on a whole universe like an ETF or screen and make sure to include the stocks that died.
59
$NVDA first week without Jensen: about $450B gone, a comeback rumor, a national champion title, and an earnings call tomorrow where nobody's sure who's talking. Who's buying?
1
73
This screams overfit, unless you can achieve similar results around the RSI settings neighbors
This quoted post is unavailable.
69
Hard to ignore
AI has completely changed my trading I’m not making any money But now when I lose it I have a bunch of custom dashboards telling me I’m a cunt
51
Most backtests buy at the same close that creates the signal, but when trading live, you can't always do that. I tested it on a simple $SPY RSI(2) dip buy since 2000. There were 203 trades with the same signals both ways. Filled at the signal close: 4.2% a year. Filled at the next open: 3.5% a year. That's about 0.09% a trade, and it adds up to a sixth of the return. If your system trades at the close, check you can actually have the signal a few minutes before it. If you can't, backtest it at the next open and see what's left.
82
$NVDA has earnings this coming Tuesday and the biggest catalyst this week is the NYT. Its going to be an awkward call for sure
79
4 for 4 failing means the checks work. Most traders on here would've went live with the first one.
I started building strategies again in November after spending about four months working on my backtesting engine. I’ve tested four ideas so far and all four have failed my robustness checks. The engine is doing its job by filtering out potentially overfitted strategies but it’s still frustrating spending days researching and developing an idea, only for it to fail validation. Mixed emotions. One half of me is proud that the engine is doing exactly what I built it to do, while the other half is disappointed that I haven’t managed to develop a single strategy that passes my checks so far.
1
87
Apparently, Dan Niles is buying the $NVDA dip and pointing to Apple after Steve Jobs stepped down, so I backtested $10K in $AAPL vs $QQQ from Aug 24, 2011, the day Jobs resigned: Sept 2012: $AAPL $18,744 Aug 2013: $AAPL $13,694, $QQQ $14,898 The rally happened, then $AAPL fell 44% from that peak by April 2013.
106