Before analyzing any specific situation, anchor to base rates. "How often do air strikes escalate to ground invasions?" Start with the historical frequency, then adjust for specifics.

Check for: Availability Heuristic, Confirmation Bias, Catastrophe Bias, Recency Bias, Anchoring, Favorite-Longshot Bias. Markets are routinely distorted by all of these simultaneously.

Read exact resolution criteria. Assess information quality. Generate probability as a range (never false-precision). Pre-commit to what specific developments would change your estimate.

Superforecasters average a Brier Score of ~0.15 vs ~0.37 for random guessing. A well-calibrated 70% prediction resolves correctly ~70% of the time. The favorite-longshot bias means: low-probability contracts win less often than priced; high-probability contracts win more often.

To bet against something. "Fade YES" means take the NO side; "don't fade the market" means don't bet against the market's price — it's probably right.

The gap between what a contract is really worth and what it costs. If Marty thinks a market is 75% likely and it trades at 62¢, that's a 13-point edge on YES. No edge = fairly priced = the disciplined move is to pass.

How often things like this have happened historically, before considering today's specifics. "Markets like this resolve YES 37% of the time" is a base rate — it's the starting anchor every analysis adjusts from.

The accuracy grade for predictions: (forecast − outcome)². 0 is perfect, a coin flip scores 0.25, lower is better. It punishes confident wrongness hardest — saying 90% on something that doesn't happen hurts much more than saying 60%.

The house's cut. Sportsbooks hide it inside the odds (-110 both ways). Prediction markets show it as the small gap between the buy and sell price — usually 1–3¢ — plus a small exchange fee.

The fine print deciding whether a contract pays YES or NO. Two markets on the same event can settle differently because of one clause (extra time counting, a snapshot date, which agency's number is used). Marty reads these on every market.

Whether your confidence matches reality. A calibrated forecaster's 70% calls come true about 70% of the time. Being calibrated matters more than being loud — it's the whole point of tracking every call publicly.

The probability bar is the whole story at a glance: the teal band is Marty's range (where he thinks the truth lives), the tick is the market's current price, and the third marker is the base rate (how often things like this historically happen). When the tick sits inside the band, the market is fairly priced — pass. When it sits outside, that gap is the edge.

The bias chips show which mental traps are distorting this particular market; the reference class shows the history the estimate is anchored to; the sources are clickable so you can verify everything yourself.

Paste the market's URL instead of describing it — Marty pulls the live price and the exact resolution rules, which makes the analysis sharper for the same query. Follow-up questions inside a thread are the cheapest way to dig deeper ("why is the base rate so low?", "what would change your mind?") — interrogate the verdict before spending a fresh analysis on a new market.

Every analysis ends with update triggers — the specific developments that would change the estimate. The workflow: ☆ star the market, set a price alert near a level that matters, and when the alert fires (or a trigger happens in the news), hit Update on your watchlist. Marty re-runs the analysis in the original thread, against his own earlier call. Re-analyzing on a schedule wastes queries; re-analyzing on triggers is how the method is meant to work.

A thin market is one where almost nobody is trading — a few thousand dollars of volume, or less. Three problems follow. The price means less: a liquid market's price is the verdict of thousands of informed traders; a thin market's price might be two people's opinions, so a "mispricing" may just be an absent crowd rather than a real edge. You move the price against yourself: with few orders on the book, even a modest buy pushes the price up as you fill — you pay more than the quote you saw. Getting out is expensive: selling early only works if someone's there to buy; in a thin market you may face a wide spread or no takers, turning a "sell anytime" contract into a hold-to-resolution position.

Marty's analyses now see each market's volume and will flag thin ones. When he does, treat any claimed edge with extra suspicion and size accordingly — or pass.

Marty is calibrated, not psychic. A 70% call loses 30% of the time — that's not a bug, it's what 70% means. He's weakest where everyone is: very long horizons, thin markets where one trader moves the price, and events with no meaningful reference class. He will tell you when there's no edge, and passing on those markets is the discipline that makes the scorecard work. Nothing here is financial advice — Marty estimates probabilities; what you do with them is yours.