Who I am
My name is Alexander Graziano. I'm a senior at Drew University with a double major in Finance and Economics and minors in Data Science and Law, Justice, and Society. I'm not a professional developer. Most of the ideas behind this site are ones I've had for years, but the barrier to actually building them was real. When tools like Codex and Claude Code came around, I finally had a way to take what was in my head and put it on a screen. The combination of those tools, years of ideas, and a lot of late nights got the site to where it is now. Frontend and backend code are different animals, and synthesizing the two into something readable is harder than it looks. There's a ton more I want to add. Between internships, school, social life, and being a one-man team, things take time.
Why baseball
I've been a baseball fan as long as I can remember. I always thought sites like Baseball Reference and Savant were great. The issue I kept running into was that each of them did one specific thing really well, and to put a full picture together I was bouncing between five tabs. My idea was to make something that captured all the different facets of baseball in one place. Standings, projections, leaders, profiles, daily slate, betting markets, prospects, the optimization lab. Not the deepest version of every single thing on day one, but an all-in-one starting point that gets better the longer it runs.
Why betting too
Studying finance and data science made me curious about whether analysis could carry into daily props honestly. There's a real cost to that. APIs aren't free, backtesting takes time and money, and a single broken pipeline can quietly invalidate a week of work. I was hesitant about the betting side because there's a stigma where the baseball nerd crowd often wants nothing to do with sports betting, and a lot of the betting crowd doesn't care about the underlying analysis. That's why I split the social presence into two accounts. @Mithrandir_base is the general baseball account with daily player cards. @Mithrandir_bets is the betting account. You can follow the side you want, or both. The site itself shows both layers because I think they reinforce each other, but you don't have to engage with both.
How rigor works here
The site is built around a three-lane methodology. Palantir is the stable, projection-anchored lane. Fangorn is a nonlinear lane that finds interactions a simpler model misses. Valinor is the calibration lane that checks whether the probabilities are behaving like probabilities. Three lanes matter because a metric can look right for the wrong reason. When the lanes agree, that's useful context. When they disagree, that's also useful context. The goal isn't to force consensus. The goal is to make disagreement visible.
Every Mithrandir+ metric goes through a validation framework before it ships publicly. Carleton-style split-half reliability, Pemstein-Dolinar stabilization, year-over-year stickiness, RMSE against a baseline. Thresholds are written down before validation runs. Outcomes are either PROMOTE, HUMAN-JUDGMENT-NEEDED, or DO-NOT-PROMOTE. Stuff+ passed. SEAGER+ Plate Discipline Index passed after being reframed. Two-Strike Adjustment did not pass, so it stayed parked. The point is that I tell you what shipped and what didn't, instead of hiding the failures.
What is still being built
Three Lanes Beta shipped. Total bases now has full three-lane coverage with CatBoost as Fangorn and LightGBM + beta calibration as Valinor. Hits and Pitcher Outs have two of three lanes with Random Forest as Fangorn. Valinor refinement for those markets is on the post-launch roadmap. Strikeouts has one lane while I work on a Fangorn architecture that exceeds the accuracy benchmark and a calibration approach that improves on Palantir. Futures stays Palantir-only because historical award-market snapshots are structurally different from daily props. The framework is the same across all markets: I ship what passes validation gates, and I say what does not.
Feedback
There's a feedback box at the bottom of every page. I read every message. If something is wrong, missing, ugly, confusing, or just an idea you think would make the site better, send it. That's how the site has improved through the entire build, and that's how it keeps getting better.