Size is the number everyone stares at. Velocity is the number that actually pays.
Velocity = percent change divided by minutes elapsed. Simple math, ruthless filter. A stock up 5% in 10 minutes has a velocity of 0.50. A stock up 8% over three hours has a velocity of 0.04. The second one looks bigger. The first one is the one you want.
Why the faster move wins
A slow move is a crowd deciding. A fast move is an imbalance. When buyers hit the tape so hard that price squeezes 4% in 15 minutes, somebody knows something — or a big book is being worked. Either way, you're seeing institutional urgency in real time.
The eight-percent move spread across three hours? It had time to absorb sellers, attract shorts, and give everyone a chance to fade it. By the time you see it, the move is priced in.
Velocity inside GRIPS
Velocity drives the G dimension of the GRIPS score — that's 20 of the 100 points. It's also the #3 feature by importance in the machine-learning model that powers the P dimension (0.163 feature importance, trained on ~14,000 labeled trades). The model agrees with the simple math: when a move is fast, the next bar is more likely to follow.
Time of day matters
Here's the subtlety: velocity of 0.30%/min at 9:37 AM is a different animal than velocity of 0.30%/min at 2:14 PM.
Morning velocity usually means institutional accumulation hitting the market. Capital is getting deployed. Afternoon velocity often means a news pop fading into the close, or a gamma squeeze with no legs. The ML model weighs time-of-day explicitly, which is why High-Conviction alerts fire only between 9:30 and 11:30 AM ET.
The rule Squintz uses
To clear the HC gate on velocity alone, a stock has to post ≥ 0.20%/min. That's the floor. Combined with a move of at least 8% and a window of 45 minutes or less, it's a filter that only about one in fifty raw signals actually clears.
Fast, clean, early. That's the signal. Everything else is decoration.