> Unlike simple averages or betting odds, DeepShot uses Exponentially Weighted Moving Averages (EWMA) to capture recent form and momentum
This is a lot of buzzwords to describe what I'm pretty sure is either very standard analysis technique in the field, or else known to be problematic for some reason or other.
> highlighting the key statistical differences between teams so you can see why the model favors one side
This is effectively just debug output and similarly doesn't need to be puffed up like that.
> or just curious whether an algorithm can outsmart Vegas
If it could, why are you here advertising the project rather than doing so yourself?
Hey, thanks for the comment — I totally get where you’re coming from. Let me clarify a bit what Deepshot actually tries to do and why I built it.
The project isn’t meant to “beat Vegas” or make betting calls — it’s an analytical tool that explores whether a model can numerically describe which team is favored to win based purely on data.
The EWMA part isn’t buzzword fluff: it’s a deliberate choice. Through a lot of testing, I found that using an exponentially weighted window of 25 games gave the most stable signal, minimizing error between predicted and actual outcomes. In practice, it captures a team’s momentum — how it’s been performing recently — better than simple averages or rolling means.
Highlighting the key statistical differences (say, +5% in rebounding or turnover rate) isn’t “puffing up debug output”; it’s a way to help visualize why the model leans toward one side. The NBA is an extremely competitive environment, and even small statistical gaps can meaningfully shift game outcomes — that’s what I wanted to surface.
As for the project itself — I’m not trying to sell it or claim it beats bookmakers. I’m sharing it because I’m 20, still learning, and I wanted to build something unique and interactive, not just another command-line model spitting numbers. Deepshot’s goal is to make basketball data exploration fun, transparent, and open to improvement by others who might want to contribute ideas or tweaks.
In short — it’s not about betting or buzzwords, it’s about learning, experimenting, and hopefully getting feedback from people who care about sports analytics as much as I do.
> Unlike simple averages or betting odds, DeepShot uses Exponentially Weighted Moving Averages (EWMA) to capture recent form and momentum
This is a lot of buzzwords to describe what I'm pretty sure is either very standard analysis technique in the field, or else known to be problematic for some reason or other.
> highlighting the key statistical differences between teams so you can see why the model favors one side
This is effectively just debug output and similarly doesn't need to be puffed up like that.
> or just curious whether an algorithm can outsmart Vegas
If it could, why are you here advertising the project rather than doing so yourself?
Hey, thanks for the comment — I totally get where you’re coming from. Let me clarify a bit what Deepshot actually tries to do and why I built it. The project isn’t meant to “beat Vegas” or make betting calls — it’s an analytical tool that explores whether a model can numerically describe which team is favored to win based purely on data. The EWMA part isn’t buzzword fluff: it’s a deliberate choice. Through a lot of testing, I found that using an exponentially weighted window of 25 games gave the most stable signal, minimizing error between predicted and actual outcomes. In practice, it captures a team’s momentum — how it’s been performing recently — better than simple averages or rolling means. Highlighting the key statistical differences (say, +5% in rebounding or turnover rate) isn’t “puffing up debug output”; it’s a way to help visualize why the model leans toward one side. The NBA is an extremely competitive environment, and even small statistical gaps can meaningfully shift game outcomes — that’s what I wanted to surface. As for the project itself — I’m not trying to sell it or claim it beats bookmakers. I’m sharing it because I’m 20, still learning, and I wanted to build something unique and interactive, not just another command-line model spitting numbers. Deepshot’s goal is to make basketball data exploration fun, transparent, and open to improvement by others who might want to contribute ideas or tweaks. In short — it’s not about betting or buzzwords, it’s about learning, experimenting, and hopefully getting feedback from people who care about sports analytics as much as I do.