Dr. Dean Flinters

PHD • 20+ Years • Statistician turned pro PM bettor • MLB/PGA/NBA

[Dr. Dean Flinters @Flinters_]

Kelly says 1.26 units. That is not a small number for an MLB line, where my model rarely finds enough gap to justify meaningful size.

The market is sharper here than NBA. Books price K props with real precision. When my model still shows edge after accounting for that, the K-rate signal against this opponent's contact profile is doing real work. The number is -110. I am taking it.

[Dr. Dean Flinters @Flinters_]

🚨 PLAY OF THE DAY 🚨
Peter Lambert O4.5 K (HOU vs LAA)

Looking at the inputs: my model has P(over) at 57.2% against a market-implied 52.4%, which is a gap worth taking seriously when the PA estimate of 23.4 gives him enough plate appearances to let the K-rate actually express itself. The 1σ spread of ±3.2 K is wide, but the projection lands at 5.3 against a line of 4.5, and that 0.8 K gap is what moves this from noise to signal.

My model has him at 5.3K. Taking him for 1.26u on Kalshi.

Model context in the next tweet.

[Dr. Dean Flinters @Flinters_]

Genuine question for the room: when a high-K/9 arm faces a lineup with a suppressed contact rate, does the strikeout distribution actually shift meaningfully, or are we just watching two independent signals that don't compound the way the intuition suggests?

[Dr. Dean Flinters @Flinters_]

MLB today. Peter Lambert, HOU vs LAA.

The market is pricing his strikeout line on something close to his unconditional K/9 distribution. My model is pricing a conditional distribution -- specifically, the one that obtains when a pitcher's stuff profile interacts with a particular opponent K%. Those are not the same number, and tonight they are 0.8 apart.

(I realize this is longer than strictly necessary, but the 0.8 is doing real work here.)

Play drops at 2:30 PM.

[Dr. Dean Flinters @Flinters_]

Practical rule: estimate your edge, apply Kelly, then ask whether you would be comfortable if your true edge is half what you estimated. If the answer is no, you are betting too much. Quarter Kelly is not timidity. It is the correct response to uncertainty about your own model. Save this one.

[Dr. Dean Flinters @Flinters_]

The Kelly criterion is the mathematically optimal bet sizing strategy. Most people who know this bet too much. Most people who haven't heard of it bet way too much. The gap between those two groups is the entire argument for why you need to understand this before you size a single bet.

[Dr. Dean Flinters @Flinters_]

I used half Kelly for the first eight months after leaving CMU. Lost less than my intuition-based phase (the $8,000 lesson), but still sized through two drawdowns I should not have taken. Moved to quarter Kelly in early 2023. The bankroll curve got smoother. Not because I got luckier. Because I stopped treating my probability estimates as facts instead of distributions.

[Dr. Dean Flinters @Flinters_]

Max Meyer U5.5. Closed 5.0. Covered.

Flat: +0.98u | Kelly: +1.48u

[Dr. Dean Flinters @Flinters_]

There have been, by my informal count, somewhere between fifteen and twenty-two "games of the year" designated in the current betting season, and the season is not over.

I want to sit with that for a moment.

The distribution of these designations is not, as you might hope, uniform across the calendar. It clusters. Heavily, reliably, in the two to four days following a winning run, which is exactly where you would expect it to cluster if the designation were assigned retroactively and dressed up as foresight. The tell is the timing. A play that wins on Tuesday becomes the game of the year by Thursday, once the outcome is known and the narrative has had time to set. I realize this is longer than strictly necessary, but the nuance is doing real work here: the question is not whether the play won. The question is whether it would have been designated game of the year had it lost. That question is never asked. The community has given it a pass, collectively, because asking it is uncomfortable and the play did win, and winning feels like it ought to mean something.

It does not mean what people think it means.

This is a specific form of hindsight bias, and it is not a minor epistemological error. It is the core error. If you cannot state your conviction threshold before the outcome is known, you do not have a conviction threshold. You have a highlights reel with a post-production team.

Designate your high-conviction plays before the game. Use a consistent threshold. Apply it uniformly across your record, not selectively to the ones that resolved in your favor. If everything is the game of the year, the phrase has zero information content, which, if you think about it, is a remarkably accurate description of most tout records.

[Dr. Dean Flinters @Flinters_]

39 plays. 22-17, +5.15u flat, +1.49u Kelly, 13.2% ROI. My model says this is promising. Statistics says this is 39 plays. Both of those things are true and they do not resolve each other.

[Dr. Dean Flinters @Flinters_]

The market doesn't misprice Shohei Ohtani because analysts lack data. It misprices him because they have too much of the wrong kind -- narrative accumulated over years, dressed up as a prior. The actual prior is invisible to them. That's the edge.

[Dr. Dean Flinters @Flinters_]

My model projected Meyer at exactly 5.5 this morning, which makes the under a coin flip with juice working against you -- except the juice is on the over, so the math still points the same direction it did at first pitch prep. Position unchanged.

[Dr. Dean Flinters @Flinters_]

The single most exploitable bias in public betting markets is not the favorite-longshot bias, not the recency heuristic, not even the frankly embarrassing tendency to bet teams that won last week on national television. It is regression to the mean, specifically the public's structural inability to believe in it.

Here is the mechanism. A player posts three exceptional games. The public, which processes sequential evidence as though each observation were independent evidence of a new, elevated true talent level, floods action onto the over. The book, which is not in the business of losing money to this crowd, moves the line to price the public's belief. The market is now pricing the hot streak, not the distribution.

Galton described this in 1886. One of my doctoral students wrote her qualifying exam on exactly this phenomenon, and her key finding was that the effect is not symmetric: public bettors anchor to recent highs far more aggressively than they update downward from recent lows, which means the overpricing after exceptional runs is reliably larger than the underpricing after poor ones. That asymmetry matters. It is where the edge lives.

The model does not watch highlight reels. It prices the long-run distribution, conditions on tonight's specific context, and asks whether the posted line reflects that distribution or whether it reflects three games that happened to go well.

Tonight's Max Meyer line is a mild version of this. The model has him at 5.5 K, which is exactly where the book landed, but the implied probability the book is assigning to the over is 52.4%. My model has P(over) at 41.9%. That gap, roughly ten percentage points, is not the book discovering something my model missed. That is a line that has drifted toward public enthusiasm and has not yet corrected.

I realize this is longer than strictly necessary, but the nuance is doing real work here. The point is not that the player is bad. The point is that the line is wrong, and the reason it is wrong is five thousand years of human beings overweighting what just happened.