25.4K
DERESULTS at 00:53. I wrote code to rank stocks by how “good” their price action would be TODAY, and it was very accurate. Here’s exactly how it works.
First, let’s just define “good price action” as large true range as a % of the stock price. That means a big move from low of day to high of day, adjusted for price, so we’re not biased toward expensive stocks.
We start by pulling a year of historical price data for each symbol. Then, we compute the true range percentage for each day, and fit an AR(1) model to that series. Put simply, we’re modeling tomorrow’s price action based on today’s — because true range is autocorrelated. That’s not just my opinion — it’s a stylized fact in financial econometrics.
The model learns:
• The baseline volatility personality of each stock
• How much yesterday’s movement affects today’s
• And how to adjust for random noise based on forecasted volatility
The simple model predicted 4.01% true range on AMD, and we got 4.00%.
Tesla: Predicted 3.51%, got 3.08%.
NVIDIA: Forecasted 3.07%, saw 2.06%.
Apple: Forecasted 1.78%, got 1.6%.
Microsoft: Forecasted 1.56%, got 0.59%.
These weren’t cherry-picked after the fact. This is real modeling using historical data only, and the predictions were made last night, WELL before today’s market open.
This is just the beginning — we can do way better than using only %TR to define “good price action.” But this already works.
If you want to learn real quant finance instead of buying into daytrading guru hype, follow me.
—
#stockmarket #pythontrading #volatilityforecasting #quantedge #optionsflow #backtest #daytradingmyths #garch #financialmodeling #marketvolatility #traderlife #statisticalarbitrage #optionstrading #predictivemodeling #truevolatility #priceactiontrading #stockforecast #quantitativefinance #dataanalytics #algoedge
@deltatrendtrading
